Fault discrimination and responsive processing method based on data and context

Abstract

To provide a method for discriminating a fault type in a continuous in vivo analyte monitoring system.SOLUTION: Systems and methods disclosed herein provide ways to discriminate fault types encountered in analyte sensors and systems, and further provide ways to process such discriminated faults responsively based on sensor data, clinical context information, and other data about a patient or an environment of the patient. The systems and methods thus employ clinical context in detecting and / or responding to errors or faults associated with an analyte sensor system, discriminating types of a fault and a root cause thereof, particularly as fault dynamics can appear similar to dynamics of physiological systems, emphasizing the importance of discriminating the fault, and providing appropriate responsive processing. Thus, the disclosed systems and methods take into account the context of the patient's health condition or state in determining how to respond to the fault.SELECTED DRAWING: Figure 1A

Description

[Technical field]

[0001] The present embodiment relates to continuous analyte monitoring, and more particularly, to fault determination and response processing within a continuous analyte monitoring system. [Background technology]

[0002] Diabetes is a disorder in which the pancreas cannot make enough insulin (type I or insulin-dependent) and / or in which insulin is ineffective (type II or insulin-independent). In the diabetic state, the patient or user suffers from hyperglycemia, which can lead to a number of physiological abnormalities associated with microvascular deterioration, such as kidney failure, skin ulcers, or bleeding into the vitreous of the eye. A hypoglycemic reaction (hypoglycemia) can be triggered by an accidental overdose of insulin or after normal administration of insulin or glucose-lowering drugs accompanied by excessive exercise or inadequate food intake.

[0003] Traditionally, a person with diabetes carries a self-monitoring blood glucose (SMBG) monitor, which typically requires an uncomfortable finger-pricking method. Due to lack of comfort and convenience, a person with diabetes typically measures their glucose levels only two to four times a day. Unfortunately, such time intervals are too far apart, making it highly likely that a person with diabetes will not detect a hyperglycemic or hypoglycemic situation until it is too late, sometimes resulting in dangerous side effects. Not only is a person with diabetes unlikely to notice and counter a dangerous situation in time, but based on traditional methods, they may not know if their blood glucose levels are rising (higher) or falling (lower). Thus, a diabetic patient may be prohibited from making educated insulin therapy decisions.

[0004] Another device that some diabetics use to monitor their blood glucose is a continuous analyte sensor, e.g., a continuous glucose monitor (CGM). A CGM typically includes a sensor that is placed invasively, minimally invasively, or non-invasively. The sensor measures the concentration of a given analyte, e.g., glucose, in the body and generates a raw signal that is generated by electronics associated with the sensor. The raw signal is converted into an output value that is rendered on a display. The output value resulting from the conversion of the raw signal is typically expressed in a format that provides meaningful information to a user, such as blood glucose expressed in mg / dL, and that allows the user to become familiar with the analysis.

[0005] The above discussion assumes that a reliable and true raw signal is received by the electronics. Occasionally, a fault or error may occur and the signal is no longer reliable and true. Prior art approaches to detecting such have generally been "one size fits all" approaches, as has the system's response thereto.

[0006] Failures or errors can be caused in a number of ways. For example, they may be associated with physiological activity in the host, such as metabolic responses, or with the in vivo portion of the sensor as it settles into the host environment. They may also be associated with primary events within the patient's control, or with the external environment surrounding the device. Other such occurrences are possible as well.

[0007] Furthermore, in the case of glucose monitoring, glucose levels and patterns vary from patient to patient, or even within a patient from day to day, so it can be difficult to distinguish noise from large fluctuations in glucose. Similarly, the best solution for a patient with stable glucose at a particular time may not be the best solution for the same or a different patient who is, for example, at or near hypoglycemia or hyperglycemia. Summary of the Invention

[0008] The present embodiments have several features, none of which is solely responsible for its desirable attributes. Without limiting the scope of the present embodiments as expressed by the claims which follow, their more prominent features will now be discussed briefly. In view of this discussion, and in particular upon reading the section entitled "Detailed Description of the Invention," one will appreciate how the features of the present embodiments provide the advantages described herein.

[0009] Systems and methods according to the present principles are situation It is recognized that the above-mentioned features are important in detecting and / or responding to errors or failures associated with an analyte sensor system. Systems and methods according to the present principles further provide a method for detecting and / or responding to errors or failures associated with an analyte sensor system, particularly when failure dynamics may appear similar to blood glucose dynamics. situation It will be appreciated that the present invention relates to determining the type of fault and its root cause, highlighting the importance of determining the fault and providing an appropriate response action. Accordingly, the disclosed systems and methods further take into account the condition or state of the patient in determining how to respond to a fault. Status In this way, situation adds an element of knowledge of clinical risk (eg, severity of illness) in the interpretation of sensor data and thus in the processing and display of sensor data.

[0010] In a first aspect, a method for determining a fault type in an in vivo continuous analyte monitoring system includes receiving a signal from an analyte monitor and situation receiving the clinical data; situation Clinical data situation Evaluate against criteria and clinical situation determining information based on the received signal from the analyte monitor and the clinical situation determining the fault type based on both the information and performing a response action based on at least the determined fault type.

[0011] In a second aspect, a method for determining a fault type in an in vivo continuous analyte monitoring system includes receiving a signal from an analyte monitor and situation receiving the clinical data; situation Clinical data situation Evaluate against criteria and clinical situation determining information about the malfunction based solely on the received signal; determining the malfunction type based solely on the received signal; and situation and performing a response process based on the information.

[0012] In a third aspect, a method for performing a response process in response to a fault in an in vivo continuous analyte monitoring system includes receiving a signal from an analyte monitor and situation receiving clinical data; and situation Clinical data situation Evaluate against criteria and clinical situation determining clinical information based on at least the received signal and the determined clinical information; situation and performing a response process based on the information.

[0013] In a fourth aspect, a method for determining a fault type in an in vivo continuous analyte monitoring system includes receiving a signal from an analyte monitor and situation receiving the clinical data; situation Clinical data situation converting the received signal from the analyte monitor and the clinical situation determining the fault type based on both the information and performing a response action based on at least the determined fault type.

[0014] In a fifth aspect, a method for determining a fault type in an in-vivo continuous analyte monitoring system includes receiving a signal from an analyte monitor, evaluating the received signal against fault criteria to determine fault information, and situationdetermining information, and situation determining the fault type based on both the information and performing a response action based on at least the determined fault type.

[0015] Implementations of the above aspects may include one or more of the following: determining the received signal, a clinical situation The determining may include classifying the fault based on the received signal, the clinical information, or both. situation The method may include classifying the fault based on the information, the signal strength, or both, where classifying the fault includes classifying the fault as a sensor environment fault or as a system error / artifact fault. The discriminating may include classifying the fault as a sensor environment fault and further includes subclassifying the fault as a compression fault or an early wound response fault. The discriminating may include determining whether the received signal or the received data matches or meets a predetermined criterion. The discriminating may include analyzing the signal using a time-based technique, a frequency-based technique, or a wavelet-based technique. The discriminating may include raw signal analysis, residualized signal analysis, pattern analysis, and / or slow versus fast sampling. The discriminating may include projecting the received signal onto a plurality of templates, each template corresponding to a failure mode. The discriminating may include variability analysis or fuzzy logic analysis. The received clinical situationThe data may be selected from the group consisting of age, anthropometric data, medications currently acting on the patient, temperature compared to a baseline, patient malfunction history, patient activity level, patient exercise level, patient interaction level with the glucose monitor, pattern of glucose signal values, clinical glucose value and its derivative, range of patient glucose levels over a period of time, duration that patient glucose levels are maintained in a range, patient glucose state, glycemic urgency index, time, or pressure. The method may further include processing the signal, where, for example, the processing removes or filters noise from the signal. The method may further include receiving a further signal, such as, for example, a sensor temperature signal, an impedance signal, an oxygen signal, a pressure signal, or a background signal. Clinical situation The information may correspond to data regarding the patient, excluding a signal value measured by a sensor associated with the analyte monitor.

[0016] clinical situation The criteria may include predefined values ​​or ranges of parameters selected from the group consisting of medications currently acting on the patient, temperature, the patient's malfunction history, the patient's activity level, the patient's exercise level, the patient's level of interaction with the glucose monitor, a pattern of glucose signal values, clinical glucose values ​​and their derivatives, a range of the patient's glucose levels over a period of time, the duration that the patient's glucose level has been maintained within a range, the patient's glucose state, a glycemic urgency index, time of day, or pressure. situation The data may include temperature, clinical situation The criteria may include a pattern of temperature, and evaluating may include determining whether the user is in clinical contact with water at the sensor site. situation The information may be determined and determining the fault type may include determining the fault type as water intrusion. situation The data may include the patient's activity level or time of day, situation The criteria may include a pattern of the patient's activity level, and evaluating may include determining whether the user is applying pressure to the sensor site. situationThe information may be determined and determining the fault type may include determining the fault type as compression. situation The data may include time since implantation, situation The criteria may include a range of implantation times during which dip and recover failure may occur, and evaluating may include determining whether the sensor was recently implanted or clinically appropriate. situation The information may be determined and determining the fault type may include determining the fault type as a descent recovery fault. situation The data may include clinical glucose values ​​and data selected from the group consisting of age, anthropometric data, activity, exercise, clinical use of the data, or patient interaction with the monitor.

[0017] The response action may include providing an indication to a user, the indication including a warning, alert, alarm, confidence index, a range of values, a predicted value, or a blank screen. Performing the response action may include adjusting a level of filtering of the received signal. Performing the response action may include performing a prediction of a future signal value based on the received signal. Performing the response action may include performing a self-diagnostic routine. Performing the response action may include performing a compensation step. Performing the response action may include switching from a first therapy mode to a second therapy mode.

[0018] In a sixth aspect, a system for carrying out any of the above methods is provided.

[0019] In a seventh aspect, there is provided a device substantially as shown and / or described in the present specification and / or drawings.

[0020] In an eighth aspect, there is provided a method substantially as shown and / or described in the present specification and / or drawings.

[0021] In a ninth aspect, there is provided an electronic device for monitoring data associated with a physiological condition, the electronic device comprising: a continuous analyte sensor configured to substantially continuously measure a concentration of an analyte in a host and provide continuous sensor data associated with the analyte concentration in the host; and a processor module configured to perform a method substantially as shown and / or described herein and / or in the drawings. The analyte may be glucose.

[0022] In a tenth aspect, there is provided an electronic device for delivering a drug to a host, the electronic device comprising: a drug delivery device configured to deliver a drug to the host, the drug delivery device operably connected to a continuous analyte sensor configured to substantially continuously measure a concentration of an analyte in the host and provide continuous sensor data associated with the analyte concentration in the host; and a processor module configured to perform a method substantially as shown and / or described herein and / or in the drawings. The analyte may be glucose and the drug may be insulin.

[0023] To facilitate understanding of the features described, continuous glucose monitoring is used as part of the following description. It will be appreciated that the described systems and methods are applicable to other continuous monitoring systems, such as analytes. For example, the described features may be used for continuous monitoring of lactate, free fatty acids, heart rate during exercise, IgG-antigliadin, insulin, glucagon, exercise tracking, fertility, caloric intake, hydration, salt, sweat / perspiration (stress), ketones, adipanectin, troponin, sweat, and / or body temperature. Glucose monitoring is used as an example and may be substituted with one or more of the alternative examples of these monitoring conditions.

[0024] Any of the features of the embodiments of the various aspects disclosed are applicable to all aspects and embodiments specified. Moreover, any of the features of the embodiments may be independently combined in any way with other embodiments described herein, either in part or in whole, e.g., one, two, or more than two embodiments may be combined in whole or in part. Moreover, any of the features of the embodiments of the various aspects may be optionally additional to other aspects or embodiments. Any aspect or embodiment of the method may be implemented by a system or device of another aspect or embodiment, and any aspect or embodiment of the system may be configured to implement a method of another aspect or embodiment. [Brief description of the drawings]

[0025] The present embodiments will now be described in detail with an emphasis on emphasizing advantageous features. These embodiments depict novel and unobvious fault determination and response processing systems and methods that are shown in the accompanying drawings, which are for illustrative purposes only, and in which like numerals indicate like parts, including the following figures:

[0026] [Figure 1A] FIG. 2 is an exploded perspective view of a glucose sensor according to one embodiment. [Figure 1B] FIG. 2 is a schematic perspective view illustrating the layers forming the in vivo portion of the analyte sensor in one embodiment. [Figure 1C] 1 is a schematic side view illustrating a formed in vivo portion of an analyte sensor; [Diagram 2] FIG. 2 is a block diagram illustrating sensor electronics in one embodiment. [Figure 3A] 1A-1C are schematic diagrams of receivers in the first, second, third and fourth embodiments, respectively; [Figure 3B] 1A-1C are schematic diagrams of receivers in the first, second, third and fourth embodiments, respectively; [Figure 3C] 1A-1C are schematic diagrams of receivers in the first, second, third and fourth embodiments, respectively; [Figure 3D] 1A-1C are schematic diagrams of receivers in the first, second, third and fourth embodiments, respectively; [Figure 4] FIG. 2 is a block diagram of receiver electronics in one embodiment. [Diagram 5] 1 is a flow diagram of a method according to the present principles; [Figure 6A] 2 is a more detailed flow diagram of a method according to the present principles showing how particular types of signals and response signal processing are performed; [Figure 6B] 1 is a plot illustrating types of noise filtering. [Figure 6C] 1 is a plot illustrating types of noise filtering. [Figure 6D] 1 is a plot illustrating types of noise filtering. [Figure 7] 1 is a plot showing the effect of temperature on noise. [Figure 8A] 1A-1D are plots showing various fault types, such as compression (A), reference electrode depletion (B), presence of noise (C), and sensor failure as determined by non-physiological behavior (D). [Figure 8B] 1A-1D are plots showing various fault types, such as compression (A), reference electrode depletion (B), presence of noise (C), and sensor failure as determined by non-physiological behavior (D). [Figure 8C] 1A-1D are plots showing various fault types, such as compression (A), reference electrode depletion (B), presence of noise (C), and sensor failure as determined by non-physiological behavior (D). [Figure 8D] 1A-1D are plots showing various fault types, such as compression (A), reference electrode depletion (B), presence of noise (C), and sensor failure as determined by non-physiological behavior (D). [Figure 9A] Illustrates slow vs. fast sampling. [Figure 9B] Illustrates slow vs. fast sampling. [Figure 10A] 1 illustrates a flow diagram and plots for applying fuzzy logic in noise determination. [Figure 10B] 1 illustrates a flow diagram and plots for applying fuzzy logic in noise determination. [Figure 10C] 1 illustrates a flow diagram and plots for applying fuzzy logic in noise determination. [Figure 10D] 1 illustrates a flow diagram and plots for applying fuzzy logic in noise determination. [Figure 11] Illustrates various types of clinical context information. [Figure 12A] 1 illustrates an aspect of a first regime of fault determination and response processing. [Figure 12B] 1 illustrates an aspect of a first regime of fault determination and response processing. [Figure 12C] 1 illustrates an aspect of a first regime of fault determination and response processing. [Figure 13A] 4 illustrates an aspect of a second regime of fault determination and response processing. [Figure 13B] 4 illustrates an aspect of a second regime of fault determination and response processing. [Figure 13C] 4 illustrates an aspect of a second regime of fault determination and response processing. [Figure 14A] 13 illustrates an embodiment of a third regime of fault determination and response processing. [Figure 14B] 13 illustrates an embodiment of a third regime of fault determination and response processing. [Figure 15] 4 is a flow diagram of another exemplary method according to present principles. [Figure 16] One classification scheme for fault discrimination is illustrated. [Figure 17] 1 illustrates another classification scheme for fault discrimination. [Figure 18] 1 illustrates a further classification scheme for fault discrimination. [Figure 19] 4 is a flow diagram of another exemplary method according to present principles. [Figure 20] 1 is a look-up table for use in response processing. [Figure 21] 11 is another table for use in response processing. [Figure 22] Illustrates the type of response signal processing. [Figure 23A] 1 illustrates selective filtering based on the signal and clinical context. [Figure 23B] 1 illustrates selective filtering based on the signal and clinical context. [Figure 23C] Illustrates fuzzy membership functions for use in fuzzy logic. [Figure 24] 1 illustrates a signal indicative of a compression failure. [Diagram 25] 4 illustrates a signal indicating a "drop recovery" fault. [Figure 26] 4 is a flow diagram of another exemplary method according to present principles. [Figure 27] 1 illustrates a signal showing a "shower spike." [Figure 28] 13 illustrates another signal indicative of a compression failure. [Figure 29A] 4 illustrates a signal indicative of a water ingress fault. [Figure 29B] 4 illustrates a signal indicative of a water ingress fault. [Diagram 30] 1 illustrates a signal showing end of life noise. [Diagram 31] 4 illustrates another signal indicative of a drop recovery fault. [Diagram 32] 4 illustrates a signal in which a lag error is present. [Figure 33A] 13 illustrates another signal indicative of a compression failure. [Figure 33B] 13 illustrates another signal indicative of a compression failure. [Figure 33C] 13 illustrates another signal indicative of a compression failure. [Figure 33D] 13 illustrates another signal indicative of a compression failure. [Figure 34A] 1 illustrates a flow diagram and signals for use in fault determination by template generation and matching. [Figure 34B] 1 illustrates a flow diagram and signals for use in fault determination by template generation and matching. [Figure 34C]1 illustrates a flow diagram and signals for use in fault determination by template generation and matching. [Figure 35A] 10 illustrates some examples of evaluation of a signal against a template. [Figure 35B] 10 illustrates some examples of evaluation of a signal against a template. [Figure 35C] 10 illustrates some examples of evaluation of a signal against a template. [Diagram 36] 1 illustrates a transmitter having an integrated force sensor. [Figure 37] Illustrates the use of linear regression in prediction or forecasting. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0027] definition To facilitate understanding of the preferred embodiments, several terms are defined below.

[0028] The term "analyte," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to one of skill in the art (and is not to be limited to any particular dedicated meaning), and further refers to, but is not limited to, a substance or chemical constituent in a biological fluid (e.g., blood, interstitial fluid, cerebrospinal fluid, lymphatic fluid, or urine) that may be analyzed. Analytes may include naturally occurring substances, man-made substances, metabolites, and / or reaction products. In some embodiments, the analyte for measurement by the sensor head, devices, and methods is an analyte. However, other analytes are contemplated as well, including but not limited to acarboxyprothrombin, acylcarnitine, adenine phosphoribosyltransferase, adenine deaminase, albumin, alpha-fetoprotein, amino acid profile (arginine (Krebs cycle), histidine / urocanic acid, homocysteine, phenylalanine / tyrosine, tryptophan), andrenostenedione, antipyrine, arabinitol enantiomers, arginase, benzoylecgonine (cocaine), biotinidase, biopterin, c-reactive protein, carnitine, carnosinase, CD4, ceruloplasmin, chenodeoxycholic acid, chloroquine, cholesterol, cholinesterase, conjugated 1-beta hydroxy-cholic acid, cortisol, creatine kinase, creatine MM isoenzyme, cyclosporine A, d-penicillamine, de-ethylchloroquine, dehydroepiandrosterone sulfate, DNA (acetylase polymorphism), alcohol dehydrogenase, α1-antitrypsin, cystic fibrosis, Duchenne / Becker muscular dystrophy, analyte-6-phosphate dehydrogenase, hemoglobin A, hemoglobin S, hemoglobin C, hemoglobin D, hemoglobin E, Hemoglobin F, D-Punjab, β-thalassemia, Hepatitis B virus, HCMV, HIV-1, HTLV-1, Leber's hereditary optic neuropathy, MCAD, RNA, PKU, Plasmodium vivax, Sex differentiation, 21-deoxycortisol), Desbutylhalofantrine, Dihydropteridine reductase, Diphtheria / tetanus antitoxin, Erythrocyte arginase, Erythrocyte protoporphyrin, Esterase D, Fatty acids / acylglycines,Free β-human chorionic gonadotropin, free erythrocyte porphyrin, free thyroxine (FT4), free tri-iodothyronine (FT3), fumarylacetoacetase, galactose / gal-1-phosphate, galactose-1-phosphate uridyltransferase, gentamicin, analyte-6-phosphate dehydrogenase, glutathione, glutathione peroxidase, glycocholate, glycosylated hemoglobin, halofantrine, hemoglobin variants, hexosaminidase A, human erythrocyte carbonic anhydrase type I, 17-α-hydroxyprogesterone, hypoxanthine phosphoribosyltransferase Spherase, immunoreactive trypsin, lactate, lead, lipoproteins ((a), B / Al, β), lysozyme, mefloquine, netilmicin, phenobarbitone, phenytoin, phytanic / pristanic acid, progesterone, prolactin, prolidase, purine nucleoside phosphorylase, quinine, inverted tri-iodothyronine (rT3), selenium, serum pancreatic lipase, sisomicin, somatomedin C, specific antibodies (adenovirus, antinuclear antibody, anti-zeta antibody, arbovirus, Aujeszky's disease virus, dengue virus, Dracaena spp., Echinococcus granulosus, Entamoeba histolytica, enterovirus, Giardiasis duodenalisa, Helicobacter pylori, Hepatitis B virus, Herpes virus, HIV-1, IgE (atopic diseases), Influenza virus, Leishmania donovani, Leptospira, Measles / mumps / rubella, Mycobacterium leprae, Mycoplasma pneumoniae, Myoglobin, Onchocerca volvulus, Parainfluenza virus, Plasmodium falciparum, Poliovirus, Pseudomonas aeruginosa, Respiratory syncytial virus, Rickettsia (scrub typhus), Schistosoma mansoni, Toxoplasma gondii, Treponema pallidium, Trypanosoma cruzi / rangeli, Vesicular stomatitis stomatis virus, bancrofti's heartworm, yellow fever virus), specific antigens (hepatitis B virus, HIV-1), succinylacetone, sulfadoxine, theophylline, thyrotropin (TSH), thyroxine (T4), thyroxine-binding globulin, trace elements, transferrin, UDP-galactose-4-epimerase, urea,Examples of analytes include uroporphyrinogen I synthase, vitamin A, white blood cells, and zinc protoporphyrin. Salts, sugars, proteins, fats, vitamins, and hormones naturally occurring in blood or interstitial fluids may also be analytes in certain embodiments. Analytes may be naturally occurring in biological fluids, such as metabolites, hormones, antigens, and antibodies. Alternatively, analytes may be introduced into the body, such as imaging contrast agents, radioisotopes, chemical agents, fluorocarbon-based artificial blood, or drugs or pharmaceutical compositions, such as insulin, ethanol, cannabis (marijuana, tetrahydrocannabinol, hashish), inhalants (nitrous oxide, amyl nitrite, butyl nitrite, chlorinated hydrocarbons, hydrocarbons), cocaine (crack cocaine), stimulants (amphetamines, methamphetamines, Ritalin, Cylert, Preludin, Didrex, PreState, Voranil, Sandrex, Plegine), depressants (barbiturates, methaqualone, tranquilizers, For example, analytes include, but are not limited to, Valium, Librium, Miltown, Serax, Equanil, Tranxene, etc.), hallucinogens (phencyclidine, lysergic acid, mescaline, peyote, psilocybin), narcotics (heroin, codeine, morphine, opium, meperidine, Percocet, Percodan, Tussionex, Fentanyl, Darvon, Talwin, Lomotil), synthetic narcotics (fentanyl, meperidine, amphetamine, methamphetamine, and analogs of phencyclidine, e.g., Ecstasy, etc.), anabolic hormones, and nicotine. Metabolites of drugs and pharmaceutical compositions are also contemplated analytes. Analytes such as neurochemicals and other chemicals produced in the body can also be analyzed, such as ascorbic acid, uric acid, dopamine, noradrenaline, 3-methoxytyramine (3MT), 3,4-dihydroxyphenylacetic acid (DOPAC), homovanillic acid (HVA), 5-hydroxytryptamine (5HT), and 5-hydroxyindoleacetic acid (FHIAA).

[0029] The term "ROM," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those skilled in the art (and is not to be limited to any particular dedicated meaning), and further refers, without limitation, to read-only memory, a type of data storage device manufactured in a fixed capacity. ROM is broad enough to include, for example, EEPROM, which is an electrically erasable programmable read-only memory (ROM).

[0030] The term "RAM," as used herein, is a broad term and is given its ordinary and accustomed meaning to those skilled in the art (and is not limited to any particular dedicated meaning), and further refers, without limitation, to a data storage device in which the order of access to different locations does not affect the speed of access. RAM is broad enough to include, for example, SRAM, which is a static random access memory that retains data bits in its memory as long as power is applied.

[0031] The term "A / D converter," as used herein, is a broad term that is to be given its ordinary and accustomed meaning to those skilled in the art (and is not intended to be limited to any particular specialized meaning), and further refers, without limitation, to hardware and / or software that converts an analog electrical signal into a corresponding digital signal.

[0032] The terms "microprocessor" and "processor," as used herein, are broad terms that are to be given their ordinary and accustomed meaning to those skilled in the art (and are not to be limited to any particular specialized meaning), and further refer, without limitation, to a computer system that performs arithmetic and logical operations using logic circuitry that responds to and processes the basic instructions that run a computer, state machines, and the like.

[0033] The term "RF transceiver," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those skilled in the art (and is not intended to be limited to any particular specialized meaning), and further refers, without limitation, to a radio frequency transmitter and / or receiver for transmitting and / or receiving signals.

[0034] The term "jitter," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those skilled in the art (and is not to be limited to any particular specialized meaning), and further refers to, but is not limited to, noise above and below the average caused by general noise caused by circuits and / or environmental effects, where jitter is seen in the amplitude, phase timing, or width of a signal pulse.

[0035] The terms "raw data stream" and "data stream" as used herein are broad terms and are given their ordinary and customary meaning to those skilled in the art (and are not limited to any particular dedicated meaning), and further refer to, but are not limited to, analog or digital signals directly related to the measured glucose from a glucose sensor. In one example, the raw data stream is digital data of "counts" converted from an analog signal (e.g., voltage or amperes) by an A / D converter, and includes one or more data points representing glucose concentration. The terms broadly encompass multiple time-spaced data points from a substantially continuous glucose sensor, including individual measurements taken at time intervals ranging from as little as a few seconds up to, for example, 1, 2, or 5 minutes or more. In another example, the raw data stream includes integrated digital values, where the data includes one or more data points representing a glucose sensor signal averaged over a period of time.

[0036] The term "calibration," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those skilled in the art (and is not to be limited to any particular dedicated meaning), and further refers, without limitation, to the process of determining a relationship between sensor data and corresponding reference data, which may be used to convert the sensor data into a meaningful value substantially equivalent to the reference data, with or without the immediate availability of the reference data. In some embodiments, i.e., for continuous analyte sensors, the calibration may be updated or recalibrated over time (at the factory, immediately and / or retroactively) as changes in the relationship between the sensor data and the reference data occur, for example, due to changes in sensitivity, baseline, transport, metabolism, etc.

[0037] The terms "calibration data" and "calibration data stream," as used herein, are broad terms and are to be given their ordinary and accustomed meaning to those skilled in the art (and are not to be limited to any particular specialized meaning), and further refer to data that has been transformed from its raw state to another state, for example, using a function such as a conversion function, to provide a meaningful value to a user.

[0038] The terms "smoothed data" and "filtered data," as used herein, are broad terms and are to be given their ordinary and accustomed meaning to one of ordinary skill in the art (and are not to be limited to any particular specialized meaning), and further refer to data that has been modified, for example, but not limited to, by performing a moving average of the raw data stream to make it smoother and more continuous, and / or to remove or reduce outliers. Examples of data filters include FIR (finite impulse response), IIR (infinite impulse response), moving average filters, etc.

[0039] The terms "smoothing" and "filtering," as used herein, are broad terms and are to be given their ordinary and accustomed meaning to those of skill in the art (and are not to be limited to any particular specialized meaning), and further refer to the modification of a data set, such as, but not limited to, making it smoother and more continuous by performing a moving average of a raw data stream, or removing or reducing outlier points.

[0040] The term "algorithm," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those of skill in the art (and is not to be limited to any particular specialized meaning), and further refers to a computational process (e.g., a program) that involves, for example, but not limited to, transforming information from one state to another by using computer processing.

[0041] The term "matched data pair," as used herein, is a broad term and is intended to be given its ordinary and accustomed meaning to those of skill in the art (and is not intended to be limited to any particular specialized meaning), and further refers, without limitation, to reference data (e.g., one or more reference analyte data points) that are matched to substantially time-corresponding sensor data (e.g., one or more sensor data points).

[0042] The term "count" as used herein is a broad term and is given its ordinary and customary meaning to one of ordinary skill in the art (and is not limited to any particular dedicated meaning), and further refers to, but is not limited to, a unit of measurement of a digital signal. In one example, the raw data stream measured by the count is directly related to a voltage (e.g., converted by an A / D converter) that is directly related to the electrode from the working electrode. In another example, the counter electrode voltage measured by the count is directly related to the voltage.

[0043] The term "sensor," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those of skill in the art (and is not intended to be limited to any particular specialized meaning), and further refers, without limitation, to a component or region of a device by which an analyte may be quantified.

[0044] The term "needle," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those of skill in the art (and is not intended to be limited to any particular specialized meaning), and further refers, without limitation, to an elongated instrument for introducing material into or removing material from the body.

[0045] The terms "glucose sensor" and "device for determining the amount of glucose in a biological sample" as used herein are broad terms and are to be given their ordinary and accustomed meaning to one of skill in the art (and are not to be limited to any particular dedicated meaning), and further refer, without limitation, to any mechanism (e.g., enzymatic or non-enzymatic) by which glucose may be quantified. For example, some embodiments utilize a membrane containing glucose oxidase, which catalyzes the conversion of oxygen and glucose to hydrogen peroxide and gluconate, as exemplified by the following chemical reaction: Glucose + O2 → Gluconate + H2O2

[0046] As each glucose molecule is metabolized, there will be a proportional change in the co-reactant O2 and product H2O2, and by monitoring the current change of either the co-reactant or product using electrodes, the glucose concentration can be determined.

[0047] The terms "operably connected" and "operably linked" as used herein are broad terms and are to be given their ordinary and customary meaning to one of ordinary skill in the art (and are not to be limited to any particular dedicated meaning), and further refer, without limitation, to one or more components that are connected to another component(s) in a manner that allows for the transmission of a signal between the components. For example, one or more electrodes can be used to detect the amount of glucose in a sample and convert that information into a signal, e.g., an electrical or electromagnetic signal, which can then be transmitted to an electronic circuit. In this case, the electrodes are "operably linked" to the electronic circuit. These terms are broad enough to include wireless connectivity.

[0048] The term "determining" encompasses a variety of actions. For example, "determining" may include calculating, computing, processing, deriving, investigating, referencing (e.g., referencing a table, database, or another data structure), ascertaining, and the like. Similarly, "determining" may include receiving (e.g., receiving information), accessing (e.g., accessing data in a memory), and the like. Similarly, "determining" may include resolving, selecting, choosing, calculating, deriving, establishing, and the like.

[0049] The term "message" encompasses a variety of formats for transmitting information. A message may include, for example, a machine-readable aggregation of information such as an XML document, a fixed-field message, a comma-separated message, etc. A message may, in some implementations, include a signal utilized to transmit one or more representations of information. Although described in the singular, it will be understood that a message may be composed / transmitted / stored / received, etc., in multiple parts.

[0050] The term "substantially," as used herein, is a broad term that is to be given its ordinary and accustomed meaning to those of skill in the art (and is not intended to be limited to any particular, specialized meaning), and further, but is not limited to, referring to being largely, but not necessarily completely, as specified.

[0051] The term "proximal," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to one of ordinary skill in the art (and is not to be limited to any particular dedicated meaning), and further refers to, but is not limited to, near a reference point such as the origin, attachment point, or midline of the body. For example, in some glucose sensor embodiments where the glucose sensor is the reference point, an oxygen sensor located proximal to the glucose sensor will be in contact with or near the glucose sensor such that their respective local environments are shared (e.g., similar levels of glucose, oxygen, pH, temperature, etc.).

[0052] The term "distal," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to one of ordinary skill in the art (and is not to be limited to any particular dedicated meaning), and further refers to being relatively distant from a reference point, such as, but not limited to, an origin or attachment point, or the midline of the body. For example, in some glucose sensor embodiments where the glucose sensor is the reference point, an oxygen sensor located distal to the glucose sensor will be sufficiently distant from the glucose sensor such that their respective local environments are not shared (e.g., glucose levels, oxygen, pH, temperature, etc. may not be similar).

[0053] The term "domain," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to one of skill in the art (and is not to be limited to any particular specialized meaning), and further refers to a region of a membrane system that may be, but is not limited to, a layer, a uniform or non-uniform gradient (e.g., an anisotropic region of a membrane), or a portion of a membrane.

[0054] The terms "in vivo portion" and "distal portion," as used herein, are broad terms and are to be given their ordinary and accustomed meaning to one of skill in the art (and are not intended to be limited to any particular specialized meaning), and further refer, without limitation, to a portion of a device (e.g., a sensor) that is adapted for insertion into and / or presence within the body of a host.

[0055] The terms "ex vivo portion" and "proximal portion," as used herein, are broad terms and are to be given their ordinary and accustomed meaning to one of skill in the art (and are not intended to be limited to any particular specialized meaning), and further refer, without limitation, to the portion of a device (e.g., a sensor) that is adapted to reside and / or exist outside the host organism.

[0056] The term "electrochemical cell", as used herein, is a broad term and is to be given its ordinary and customary meaning to those skilled in the art (and is not to be limited to any particular dedicated meaning), and further refers, without limitation, to a device in which chemical energy is converted to electrical energy. Such cells are typically composed of two or more electrodes held spaced apart from one another and in contact with an electrolyte solution. Connection of the electrodes to a direct current source causes one of them to be negatively charged and the other to be positively charged. Positive ions in the electrolyte migrate to the negative electrode (cathode) where they combine with one or more electrons, lose some or all of their charge, and become new ions or neutral atoms or molecules with a lower charge, and at the same time, negative ions migrate to the positive electrode (anode) and transfer one or more electrons to it, also becoming new ions or neutral particles. The overall effect of the two processes is the transfer of electrons from the negative ions to the positive ions, a chemical reaction.

[0057] The term "electric potential," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those skilled in the art (and is not to be limited to any particular specialized meaning), and further refers, without limitation, to the potential difference between two points in a circuit that causes an electric current to flow.

[0058] The term "host," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those of skill in the art (and is not intended to be limited to any particular specialized meaning), and further refers to, but is not limited to, mammals, particularly humans.

[0059] The term "continuous analyte (or glucose) sensor," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to one of skill in the art (and is not to be limited to any particular dedicated meaning), and further refers, without limitation, to a device that measures the concentration of an analyte continuously or continuously, for example, over time intervals ranging from just a few seconds up to, for example, 1, 2, or 5 minutes or more. In one exemplary embodiment, the continuous analyte sensor is a glucose sensor, such as that described in U.S. Patent No. 6,001,067, the entirety of which is incorporated herein by reference.

[0060] The term "continuous analyte (or glucose) sensing," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those of skill in the art (and is not intended to be limited to any particular specialized meaning), and further refers, without limitation, to periods during which analyte monitoring is performed continuously or continuously, for example, at time intervals ranging from just a few seconds up to, for example, 1, 2, or 5 minutes or more.

[0061] The terms "reference analyte monitor," "reference analyte meter," and "reference analyte sensor," as used herein, are broad terms and are to be given their ordinary and accustomed meaning to one of ordinary skill in the art (and are not to be limited to any particular specialized meaning), and further refer, without limitation, to a device that measures the concentration of an analyte and may be used as a reference for a continuous analyte sensor, e.g., a self-monitoring blood glucose meter (SMBG) may be used as a reference for a continuous glucose sensor for comparison, calibration, etc.

[0062] The term "sensing membrane," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to one of skill in the art (and is not to be limited to any particular dedicated meaning), and further refers, without limitation, to a permeable or semi-permeable membrane that can consist of two or more domains and is constructed of materials that are typically a few microns thick or more, that are permeable to oxygen and may or may not be permeable to glucose. In one example, the sensing membrane contains immobilized glucose oxidase enzyme, which allows for an electrochemical reaction to occur to measure glucose concentration.

[0063] The term "physiologically feasible" as used herein is a broad term and is to be given its ordinary and customary meaning to those skilled in the art (and is not to be limited to any particular dedicated meaning), and further refers to, but is not limited to, physiological parameters derived from continuous studies of glucose data in humans and / or animals. For example, a maximum sustained rate of glucose change in humans of about 4-5 mg / dL / min and a maximum acceleration of the rate of change of about 0.1-0.2 mg / dL / min / min are considered physiologically feasible limits. Values ​​outside these limits would be considered non-physiological and are likely to be the result of, for example, signal errors. As another example, the rate of glucose change is lowest at the maximum and minimum values ​​of daily glucose change, which are the areas of greatest risk in the treatment of a patient, and thus the physiologically feasible rate of change may be set to the maximum and minimum values ​​based on continuous studies of glucose data. As a further example, it has been observed that the best solution for the curve shape at any point along the glucose signal data stream over a particular period of time (e.g., about 20-30 minutes) is a straight line, which can be used to set physiological limits.

[0064] The term "ischemia," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those skilled in the art (and is not intended to be limited to any particular specialized meaning), and further, without limitation, refers to a localized and temporary lack of blood supply due to an interruption in circulation to a component (e.g., a sensor). Ischemia can be caused, for example, by a mechanical interruption of the blood supply (eg, arterial stenosis or disruption).

[0065] The term "system noise," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those skilled in the art (and is not to be limited to any particular specialized meaning), and further refers to, but is not limited to, unwanted electronic or diffusion-related noise, which may include, for example, Gaussian noise, motion-related noise, flicker noise, dynamic noise, or other white noise.

[0066] The terms "noise", "noise event(s)", "noise episode(s)", "signal artifact(s)", "signal artifact event(s)", and "signal artifact episode(s)", as used herein, are broad terms and are given their ordinary and customary meaning to those skilled in the art (and are not limited to any particular, dedicated meaning), and further refer to signal noise caused by substantially non-glucose related, such as, but not limited to, interfering species, hyperactivity or microactivity, ischemia, pH changes, temperature changes, pressure, stress, or even unknown mechanical, electrical, and / or biochemical noise sources. In some embodiments, the signal artifacts are transient and characterized by a higher amplitude than the system noise, and are described as "transient non-glucose related signal artifact(s) having a higher amplitude than the system noise". In some embodiments, the noise is caused by a rate limiting (or rate increasing) phenomenon. In some situations, the source of the noise is unknown.

[0067] The terms "stationary noise" and "stationary background", as used herein, are broad terms and are to be given their ordinary and customary meaning to those skilled in the art (and are not to be limited to any particular dedicated meaning), and refer, without limitation, to components of a noise signal that remain relatively constant over time. In some embodiments, stationary noise may also be referred to as "background" or "baseline". For example, certain electroactive compounds found in the human body are relatively constant factors (e.g., the baseline of the host's physiology). In some circumstances, stationary background noise may slowly drift (e.g., increase or decrease) over time, but this drift does not necessarily adversely affect the accuracy of the sensor, for example, because the sensor can be calibrated and recalibrated and / or the drift can be measured and compensated for.

[0068] The terms "non-stationary noise," "non-stationary background," "noise event(s), "noise episode(s), "signal artifact(s), "signal artifact event(s)," and "signal artifact episode(s)," as used herein, are broad terms and are given their ordinary and customary meaning to those of skill in the art (and are not limited to any particular dedicated meaning), and refer, without limitation, to components of a background signal that are relatively non-stationary, e.g., transient and / or intermittent. For example, certain electroactive compounds are relatively non-stationary due to host uptake, metabolism, wound healing, and other mechanical, chemical, and / or biochemical factors), which create intermittent (e.g., non-stationary) "noise" on the sensor signal that can be difficult to "calibrate" using standard calibration formulas (e.g., because the signal background does not remain constant).

[0069] The term "low noise," as used herein, is a broad term and is intended to be given its ordinary and accustomed meaning to those skilled in the art (and is not intended to be limited to any particular specialized meaning), and further refers to, but is not limited to, noise that substantially reduces signal amplitude.

[0070] The terms "high noise" and "high spikes", as used herein, are broad terms and are to be given their ordinary and accustomed meaning to those skilled in the art (and are not to be limited to any particular specialized meaning), and further refer to, but are not limited to, noise that substantially increases the signal amplitude.

[0071] The term "frequency content," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those skilled in the art (and is not to be limited to any particular specialized meaning), and further refers to the spectral density, including but not limited to the frequencies contained within a signal and its power.

[0072] The term "spectral density," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those skilled in the art (and is not to be limited to any particular specialized meaning), and further, without limitation, refers to the power spectral density of electromagnetic radiation of a given bandwidth being the total power within that bandwidth divided by the specified bandwidth. Spectral density is usually expressed in watts per Hertz (W / Hz).

[0073] The term "chronoamperometry," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those skilled in the art (and is not to be limited to any particular specialized meaning), and further refers, without limitation, to an electrochemical measurement technique used for electrochemical analysis or for determining the kinetics and mechanism of electrode reactions. A rising potential pulse is forced onto the working (or reference) electrode of an electrochemical cell, and the current flowing through this electrode is measured as a function of time.

[0074] The term "pulsed amperometric detection" as used herein is a broad term and is to be given its ordinary and customary meaning to those skilled in the art (and is not to be limited to any particular specialized meaning), and further refers, without limitation, to an electrochemical flow cell and control device that applies a potential and monitors the current generated by an electrochemical reaction. The cell may contain one or more working electrodes at different applied potentials. The multiple electrodes may be arranged so that they face the chromatographic flow independently (parallel configuration) or in series (series configuration).

[0075] The term "linear regression," as used herein, is a broad term and is given its ordinary and accustomed meaning to those skilled in the art (and is not limited to any particular specialized meaning), and further refers, without limitation, to finding the line where a data set has a minimum measure from that line. By-products of this algorithm include the slope, y-intercept, and R-squared value, which determines how well the measured data fits the line.

[0076] The term "nonlinear regression," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to one of skill in the art (and is not to be limited to any particular specialized meaning), and further, without limitation, refers to fitting a data set to explain the relationship between a response variable and one or more explanatory variables in a nonlinear manner.

[0077] The term "mean," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those of skill in the art (and is not to be limited to any particular specialized meaning), and further, without limitation, refers to the sum of the observations divided by the number of observations.

[0078] The term "trimmed mean" as used herein is a broad term and is to be given its ordinary and accustomed meaning to one of ordinary skill in the art (and is not to be limited to any particular specialized meaning), and further refers, without limitation, to an average taken after the extreme (e.g., high and low) values ​​of a variable have been eliminated or reduced (e.g., "trimmed"). A trimmed mean compensates for sensitivity to extreme values ​​by cutting off a certain percentage of the values ​​at the ends. For example, a 50% trimmed mean is the average of the values ​​between the top and bottom quartiles. A 90% trimmed mean is the average of the values ​​after truncating the lowest and highest 5% of the values. In one example, the top two and bottom two measurements are removed from a data set, and then the remaining measurements are averaged.

[0079] The term "non-recursive filter," as used herein, is a broad term that is to be given its ordinary and accustomed meaning to those skilled in the art (and is not intended to be limited to any particular specialized meaning), and further, without limitation, refers to an equation that uses moving averages as inputs and outputs.

[0080] The terms "recursive filter" and "autoregressive algorithm" as used herein are broad terms and are to be given their ordinary and accustomed meaning to those skilled in the art (and are not to be limited to any particular specialized meaning), and further refer to equations including, but not limited to, where the previous average is part of the next filtered output. More specifically, the generation of a series of observations whereby the value of each observation depends in part on the value immediately preceding the observation. One example is a regression structure where lagged response values ​​assume the role of independent variables.

[0081] The term "signal estimation algorithm factors," as used herein, is a broad term and is given its ordinary and accustomed meaning to one of ordinary skill in the art (and is not limited to any particular specialized meaning), and further refers, without limitation, to one or more algorithms that estimate unknown signal data stream values ​​using historical and / or current signal data stream values. For example, the signal estimation algorithm factors may include one or more algorithms, such as linear or nonlinear regression. As another example, the signal estimation algorithm factors may include one or more sets of coefficients that may be applied to an algorithm.

[0082] The term "variability," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to one of skill in the art (and is not to be limited to any particular specialized meaning), and further refers, without limitation, to the amount of divergence or change from a point, line, or set of data. In one embodiment, the estimated analyte value may have a variability that includes a range of values ​​outside the estimated analyte value that represents a range of possibilities, for example, based on known physiological patterns.

[0083] The terms "physiological parameters" and "physiological boundaries", as used herein, are broad terms and are to be given their ordinary and accustomed meaning to one of skill in the art (and are not to be limited to any particular specialized meaning), and further refer to, but are not limited to, parameters derived from serial studies of physiological data in humans and / or animals. For example, a maximum sustained rate of glucose change in humans of about 4-5 mg / dL / min, and a maximum sustained rate of glucose change in humans of about 0.1-0.2 mg / dL / min. 2The maximum acceleration of the rate of change of glucose would be considered the physiologically feasible limits, and values ​​outside these limits would be considered non-physiological. As another example, the rate of change of glucose is lowest at the maximum and minimum values ​​of daily glucose change, which are the areas of greatest risk in the treatment of the patient, and thus the physiologically feasible rate of change may be set to the maximum and minimum values ​​based on continuous study of glucose data. As a further example, it has been observed that the best solution for the curve shape at any point along the glucose signal data stream over a certain period of time (e.g., about 20-30 minutes) is a straight line, which may be used to set the physiological limits. These terms are broad enough to include physiological parameters for any analyte.

[0084] The term "measured analyte value," as used herein, is a broad term and is to be given its ordinary and accustomed meaning to those of skill in the art (and is not to be limited to any particular specialized meaning), and further, without limitation, refers to an analyte value or set of analyte values ​​over a period of time during which analyte data is being measured by an analyte sensor. The term is broad enough to include data from an analyte sensor before or after data processing (e.g., data smoothing, calibration, etc.) at the sensor and / or receiver.

[0085] The term "estimated analyte value," as used herein, is a broad term and is intended to be given its ordinary and accustomed meaning to one of skill in the art (and is not intended to be limited to any particular specialized meaning), and further refers, without limitation, to an analyte value or set of analyte values ​​that are algorithmically extrapolated from a measured analyte value.

[0086] The terms "interferant" and "interfering species," as used herein, are broad terms and are to be given their ordinary and accustomed meaning to one of skill in the art (and are not to be limited to any particular specialized meaning), and further refer, without limitation, to effects and / or species that interfere with the measurement of an analyte of interest in a sensor, producing a signal that does not accurately represent the analyte concentration. In one example of an electrochemical sensor, an interfering species is a compound that has an oxidation potential that overlaps with the oxidation potential of the analyte being measured, thereby producing a false positive signal.

[0087] As used herein, the following abbreviations apply: Eq and Eqs (equivalents), mEq (milliequivalents), M (molar), mM (millimolar), μM (micromolar), N (normal), mol (mole), mmol (millimole), μmol (micromole), nmol (nanomole), g (gram), mg (milligram), μg (microgram), Kg (kilogram), L (liter), mL (milliliter), dL (deciliter), μL (microliter), cm (centimeter), mm (millimeter), μm (micrometer), nm (nanometer), h and hr (hours), min. (minutes), s and sec. (seconds), °C (degrees Celsius). Overview / System Overview

[0088] The glucose sensor may use any system or method for providing a data stream indicative of the concentration of glucose in a host. The data stream is typically a raw data signal that is converted to provide a useful glucose value to a user, such as a patient or physician, who may be using the sensor. However, failures may occur that may affect the analysis of the signal, clinical situation , or both. Such a fault requires discrimination to distinguish it from the actual measured signal behavior, and for response signal processing that may vary according to the fault. Therefore, appropriate fault discrimination and response processing techniques are used. Glucose Sensor

[0089] The glucose sensor can be any device capable of measuring the concentration of glucose. One exemplary embodiment is described below, which utilizes an implantable glucose sensor. However, it should be understood that the devices and methods described herein can be applied to any device capable of detecting the concentration of glucose and providing an output signal representative of the concentration of glucose.

[0090] Exemplary embodiments disclosed herein relate to the use of glucose sensors to measure concentrations of glucose or substances indicative of the concentration or presence of another analyte. In some embodiments, the glucose sensor is a continuous device, such as a subcutaneous, transdermal, transdermal, non-invasive, intraocular, and / or intravascular (e.g., intravenous) device. In some embodiments, the device may analyze multiple intermittent blood samples. The glucose sensor may use any method of glucose measurement, including enzymatic, chemical, physical, electrochemical, optical, photochemical, fluorescence-based, spectroscopic, spectroscopic (e.g., optical absorption spectroscopy, Raman spectroscopy, etc.), polarimetric, calorimetric, iontophoretic, radiometric, etc.

[0091] The glucose sensor may use any known detection method to provide a data stream indicative of the concentration of an analyte in a host, including invasive, minimally invasive, and non-invasive sensing techniques. The data stream is typically a raw data signal that is used to provide a useful analyte value to a user, such as a patient or a health care professional (e.g., a physician), who may be using the sensor.

[0092] While most of the description and examples are drawn to a glucose sensor capable of measuring the concentration of glucose in a host, the systems and methods of the embodiments may be applied to any measurable analyte, the list of suitable analytes above. Some exemplary embodiments described below utilize an implantable glucose sensor. However, it should be understood that the devices and methods described herein may be applied to any device capable of detecting the concentration of an analyte and providing an output signal representative of the concentration of the analyte.

[0093] In one preferred embodiment, the analyte sensor is an implantable glucose sensor such as those described with reference to U.S. Patent No. 6,001,067 and U.S. Patent Publication No. US-2005-0027463-A1. In another preferred embodiment, the analyte sensor is a transcutaneous glucose sensor such as those described with reference to U.S. Patent Publication No. US-2006-0020187-A1. In yet other embodiments, the sensor is configured to be implanted within a host's blood vessel or externally as described in U.S. Patent Publication Nos. US-2007-0027385-A1, US-2008-0119703-A1, US-2008-0108942-A1, and US-2007-0197890-A1. In one alternative embodiment, the continuous glucose sensor comprises a transcutaneous sensor, such as, for example, that described in U.S. Patent No. 6,565,509 to Say et al. In another alternative embodiment, the continuous glucose sensor comprises a subcutaneous sensor, such as, for example, that described with reference to U.S. Patent No. 6,579,690 to Bonnecaze et al. or U.S. Patent No. 6,484,046 to Say et al. In another alternative embodiment, the continuous glucose sensor comprises a refillable subcutaneous sensor, such as, for example, that described with reference to U.S. Patent No. 6,512,939 to Colvin et al. In another alternative embodiment, the continuous glucose sensor comprises an intravascular sensor, such as, for example, that described with reference to U.S. Patent No. 6,477,395 to Schulman et al. In another alternative embodiment, the continuous glucose sensor comprises an intravascular sensor, such as, for example, that described with reference to U.S. Patent No. 6,424,847 to Mastrototaro et al.

[0094] The following description and examples describe the present embodiments with reference to the drawings, in which reference numerals indicate elements of the present embodiments, and which are reproduced below in conjunction with a discussion of the corresponding drawing features. System Specific Description

[0095] Figure 1A is an exploded perspective view of an exemplary embodiment including an implantable glucose sensor 10 that utilizes amperometric electrochemical sensor technology to measure glucose concentrations. In this exemplary embodiment, a body 12 and a head 14 house electrodes 16 and sensor electronics, which are described in more detail below with reference to Figure 2. The three electrodes 16 are operably connected to the sensor electronics (Figure 2) and are covered by a sensing membrane 17 and a biointerface membrane 18 that are attached by clips 19.

[0096] In one embodiment, three electrodes 16 protrude from head 14 and include a platinum working electrode, a platinum counter electrode, and a silver / silver chloride reference electrode. The top ends of the electrodes are in contact with an electrolyte phase (not shown), which is a free-flowing fluid phase disposed between sensing membrane 17 and electrode 16. Sensing membrane 17 includes an enzyme, such as, for example, glucose oxidase, which covers the electrolyte phase. A biointerface membrane 18 covers sensing membrane 17 and also serves to, at least in part, protect sensor 10 from external forces that may result in environmental stress cracking of sensing membrane 17.

[0097] In the illustrated embodiment, a counter electrode is provided to balance the current generated by the species being measured at the working electrode. In the case of a glucose oxidase-based glucose sensor, the species being measured at the working electrode is H2O2. Glucose oxidase catalyzes the conversion of oxygen and glucose to hydrogen peroxide and gluconate according to the following reaction: Glucose + O2 → Gluconate + H2O2

[0098] Because there is a proportional change in the product H2O2 for each glucose molecule metabolized, the change in H2O2 can be monitored to determine the glucose concentration. The oxidation of H2O2 by the working electrode is balanced by the reduction of ambient oxygen, enzymatically generated H2O2, or other reducible species at the counter electrode. H2O2 produced from the glucose oxidase reaction reacts further at the surface of the working electrode, producing two protons (2H + ), 2 electrons (2e -), and one oxygen molecule (O2).

[0099] 1B-1C illustrate an exemplary embodiment of an in vivo portion of a continuous analyte sensor 100 including an elongated conductive body 102. The elongated conductive body 102 includes a core 110 (see FIG. 1B) and a first layer 112 at least partially surrounding the core. The first layer includes a working electrode (e.g., located within the window 106) and a membrane 108 located over the working electrode configured and arranged for multi-axis bending. In some embodiments, the core and the first layer may be of a single material (e.g., platinum). In some embodiments, the elongated conductive body is a composite of at least two materials, such as, for example, a composite of two conductive materials, or a composite of at least one conductive material and at least one non-conductive material. In some embodiments, the elongated conductive body comprises multiple layers. In certain embodiments, there are at least two concentric (e.g., annular) layers, such as, for example, a core formed of a first material and a first layer formed of a second material. However, in some embodiments, additional layers may be included. In some embodiments, the layers are coaxial.

[0100] The elongated conductive body may be long and thin, yet flexible and strong. For example, in some embodiments, the smallest dimension of the elongated conductive body is less than about 0.1 inches, 0.075 inches, 0.05 inches, 0.025 inches, 0.01 inches, 0.004 inches, or 0.002 inches. Although the elongated conductive body is illustrated in FIGS. 1B-1C as having a circular cross section, in other embodiments, the cross section of the elongated conductive body may be oval, rectangular, triangular, polyhedral, star-shaped, C-shaped, T-shaped, X-shaped, Y-shaped, irregular, etc. In one embodiment, a conductive wire electrode is used as the core. To such a coated electrode, two additional conductive layers may be added (e.g., with an intervening insulating layer that provides for electrical shielding). The conductive layer may be made of any suitable material. In certain embodiments, it may be desirable to use a conductive layer that includes conductive particles (i.e., particles of a conductive material) in a polymer or other binder.

[0101] In certain embodiments, the materials used to form the elongated conductive body (e.g., stainless steel, titanium, tantalum, platinum, platinum-iridium, iridium, certain polymers, etc.) can be tough and rigid, and therefore resistant to breakage. For example, in some embodiments, the ultimate tensile strength of the elongated conductive body is about 80 kPsi to about 500 kPsi. In another example, in some embodiments, the Young's modulus of the elongated conductive body is about 160 GPa to about 220 GPa. In yet another example, in some embodiments, the yield strength of the elongated conductive body is about 60 kPsi to about 2200 MPa. In some embodiments, the small diameter of the sensor provides (e.g., imparts, allows) flexibility to these materials, and thus to the entire sensor. Thus, the sensor can withstand repeated forces applied to it by the surrounding tissue. In some embodiments, the fatigue life of the sensor is at least 1,000 flex cycles of about 28° to about 110° at a bend radius of about 0.125 inches.

[0102] In addition to providing structural support, resilience, and flexibility, in some embodiments, the core 110 (or components thereof) provides electrical conduction of an electrical signal from the working electrode to the sensor electronics (FIG. 2), which are described elsewhere herein. In some embodiments, the core 110 comprises a conductive material, such as stainless steel, titanium, tantalum, a conductive polymer, etc. However, in other embodiments, the core is formed of a non-conductive material, such as a non-conductive polymer. In still other embodiments, the core comprises multiple layers of material. For example, in one embodiment, the core includes an inner core and an outer core. In further embodiments, the inner core is formed of a first conductive material and the outer core is formed of a second conductive material. For example, in some embodiments, the first conductive material is stainless steel, titanium, tantalum, a conductive polymer, an alloy, etc., and the second conductive material is a conductive material selected to provide electrical conduction between the core and the first layer and / or to attach the first layer to the core (e.g., when the first layer is formed of a material that does not adhere well to the core material). In another embodiment, the core is formed of a non-conductive material (e.g., a non-conductive metal and / or a non-conductive polymer) and the first layer is a conductive material, such as stainless steel, titanium, tantalum, a conductive polymer, etc. The core and first layer can be a single (or the same) material, such as, for example, platinum. One skilled in the art will recognize that additional configurations are possible.

[0103] 1B-1C, in some embodiments, the first layer 112 is formed of an electrically conductive material. The working electrode is the exposed portion of the surface of the first layer. Thus, the first layer is formed of a material configured to provide a suitable electroactive surface for the working electrode, such as, but not limited to, platinum, platinum-iridium, gold, palladium, iridium, graphite, carbon, conductive polymers, alloys, and the like.

[0104] As illustrated in FIGS. 1B-1C, the second layer 104 surrounds at least a portion of the first layer 112, thereby defining the boundary of the working electrode. In some embodiments, the second layer 104 serves as an insulator and is formed of an insulating material, such as polyimide, polyurethane, parylene, or any other known insulating material. For example, in one embodiment, the second layer is disposed on the first layer and configured such that the working electrode is exposed through the window 106. In another embodiment, an elongated conductive body is provided that includes a core, a first layer, and a second layer, and the working electrode is exposed (i.e., formed) by removing a portion of the second layer, thereby forming a window 106 through which the electroactive surface of the working electrode (e.g., the exposed surface of the first layer) is exposed. In some embodiments, the working electrode is exposed (e.g., the window 106 is formed) by removing a portion of the second and (optionally) third layers. Removal of coating material from one or more layers of the elongated conductive body (e.g., to expose the electroactive surface of the working electrode) may be performed by hand, excimer laser treatment, chemical etching, laser ablation, grit blasting, or the like.

[0105] In some embodiments, the sensor further comprises a third layer 114 comprising a conductive material. In further embodiments, the third layer may comprise a reference electrode, which may be formed of a silver-containing material applied onto the second layer (e.g., an insulator). The silver-containing material may comprise any of a variety of materials and various forms, such as, for example, commercially available Ag / AgCl polymer pastes, paints, polymer-based conductive mixtures, and / or inks. The third layer may be processed using a pasting / dipping / coating step, for example, using a die-measured dip coating process. In one exemplary embodiment, the Ag / AgCl polymer paste is applied to the elongated body by dip coating the body (e.g., using a meniscus coating technique) and then drawing the body through a die to gauge the coating to a precise thickness. In some embodiments, multiple coating steps are used to build up the coating to a predetermined thickness. Such drawing methods may be utilized to form one or more of the electrodes in the device depicted in FIG. 1B.

[0106] In some embodiments, the silver grains in the Ag / AgCl solution or paste may have an average particle size corresponding to a maximum particle dimension that is less than about 100 microns, or less than about 50 microns, or less than about 30 microns, or less than about 20 microns, or less than about 10 microns, or less than about 5 microns. The silver chloride grains in the Ag / AgCl solution or paste may have an average particle size corresponding to a maximum particle dimension that is less than about 100 microns, or less than about 80 microns, or less than about 60 microns, or less than about 50 microns, or less than about 20 microns, or less than about 10 microns. The silver grains and silver chloride grains may be incorporated in a weight ratio of silver chloride grains:silver grains of about 0.01:1 to 2:1, or about 0.1:1 to 1:1. The silver grains and silver chloride grains are then mixed with a carrier (e.g., polyurethane) to form a solution or paste. In certain embodiments, the Ag / AgCl component forms about 10% to about 65% by weight, or about 20% to about 50%, or about 23% to about 37% of the total Ag / AgCl solution or paste. In some embodiments, the Ag / AgCl solution or paste has a viscosity (under ambient conditions) that is about 1 to about 500 centipoise, or about 10 to about 300 centipoise, or about 50 to about 150 centipoise.

[0107] In some embodiments, Ag / AgCl particles are mixed into a polymer, such as polyurethane, polyimide, etc., to form a silver-containing material for the reference electrode. In some embodiments, the third layer is cured, for example, using an oven or other curing process. In some embodiments, a cover of a fluid-permeable polymer having conductive particles (e.g., carbon particles) therein is applied over the reference electrode and / or the third layer. In some embodiments, a layer of insulating material is located over a portion of the silver-containing material.

[0108] In some embodiments, the elongated conductive body further comprises one or more intermediate layers positioned between the core and the first layer, e.g., in some embodiments, the intermediate layer is an insulator, a conductor, a polymer, and / or an adhesive.

[0109] It is contemplated that the ratio of the thickness of the Ag / AgCl layer to the thickness of the insulator (e.g., polyurethane or polyimide) layer may be controlled to allow for a certain margin of error (e.g., margin of error resulting from the etching process) that would not result in a defective sensor (e.g., due to defects resulting from the etching process cutting deeper than intended, thereby unintentionally exposing the electroactive surface). This ratio may vary depending on the type of etching process used, whether it is laser ablation, grit blasting, chemical etching, or some other etching method. In an embodiment in which laser ablation is performed to remove the Ag / AgCl and polyurethane layers, the ratio of the thickness of the Ag / AgCl layer to the thickness of the polyurethane layer may be from about 1:5 to about 1:1, or from about 1:3 to about 1:2.

[0110] In certain embodiments, the core comprises a non-conductive polymer and the first layer comprises a conductive material. Such a sensor configuration may provide reduced material costs in that typically expensive materials are replaced with less expensive materials. For example, in some embodiments, the core is formed of a non-conductive polymer, such as, for example, nylon or polyester fibers, strings, or cords, which may be coated and / or plated with a conductive material, such as, for example, platinum, platinum-iridium, gold, palladium, iridium, graphite, carbon, conductive polymers, and alloys or combinations thereof.

[0111] As illustrated in FIG. 1C, the sensor also includes a membrane 108 that covers at least a portion of the working electrode.

[0112] In embodiments where outer insulation is disposed, a portion of the coated assembly structure may be stripped or otherwise removed to expose the electroactive surfaces, e.g., by hand, excimer laser treatment, chemical etching, laser ablation, grit blasting, etc. Alternatively, a portion of the electrode may be masked prior to depositing the insulation to maintain exposed electroactive surface area.

[0113] In some embodiments, a radial window is formed through the insulating material to expose the circumferential electroactive surface of the working electrode. Additionally, a cross section of the electroactive surface of the reference electrode is exposed. For example, the cross section of the electroactive surface can be masked during deposition of the outer insulating layer or etched after deposition of the outer insulating layer. In some applications, cellular attack or migration on the sensor can cause reduced device sensitivity or function, especially after the first day of implantation. However, if the exposed electroactive surface is distributed circumferentially around the sensor (e.g., as a radial window), the surface area available for reaction can be sufficiently distributed to minimize the effect of local cellular invasion of the sensor on the sensor signal. Alternatively, a tangential exposed electroactive window can be formed, for example, by stripping only one side of the coated assembly structure. In other alternative embodiments, a window can be provided at the tip of the coated assembly structure such that the electroactive surface is exposed at the tip of the sensor. Other methods and configurations for exposing the electroactive surface can be used as well.

[0114] In some alternative embodiments, additional electrodes may be included in the assembly, such as, for example, a three electrode system (working, reference, and counter electrodes) and an additional working electrode (e.g., an electrode that may be used to generate oxygen configured as a baseline subtraction electrode or configured to measure an additional analyte). U.S. Pat. No. 7,081,195, U.S. Patent Publication No. US-2005-0143635-A1, and U.S. Patent Publication No. US-2007-0027385-A1, each of which is incorporated herein by reference, describe several systems and methods for implementing and using additional working, counter, and reference electrodes. In one implementation in which the sensor comprises two working electrodes, the two working electrodes are juxtaposed and the reference electrode is disposed (e.g., spirally wound) around them. In some embodiments in which more than one working electrode is provided, the working electrodes may be formed in a double, triple, forty-fold spiral configuration, etc. along the length of the sensor (e.g., surrounding a reference electrode, an insulating rod, or other support structure). The resulting electrode system may be configured with a suitable membrane system, where a first working electrode is configured to measure a first signal including glucose and a baseline signal, and a further working electrode is configured to measure a baseline signal consisting of only the baseline signal. In these embodiments, the second working electrode may be configured substantially similar to the first working electrode, but without an enzyme disposed thereon. In this manner, a baseline signal may be determined and subtracted from the first signal to generate a difference signal, i.e., a glucose-only signal that is substantially not subject to fluctuations in the baseline or interfering species on the signal, as described in U.S. Patent Publication Nos. US-2005-0143635-A1, US-2007-0027385-A1, and US-2007-0213611-A1, and US-2008-0083617-A1, which are incorporated herein by reference in their entireties.

[0115] It has been found that in some electrode systems including two working electrodes, i.e., some dual electrode systems, the working electrodes may sometimes differ slightly from each other. For example, the two working electrodes may differ slightly in thickness or permeability even when manufactured from a single facility due to the high sensitivity of the electrodes to environmental conditions (e.g., temperature, humidity) during fabrication. Thus, the working electrodes of a dual electrode system may sometimes have diverse diffusion, membrane thickness, and diffusion characteristics. As a result, the difference signal described above (i.e., the glucose-only signal generated by subtracting the baseline signal from the first signal) may not be completely accurate. To mitigate this, it is contemplated that in some dual electrode systems, both working electrodes may be fabricated with one or more membranes each including a bioprotective layer, which is described in more detail elsewhere herein.

[0116] It is contemplated that the sensing area may include any of a variety of electrode configurations. For example, in some embodiments, in addition to one or more glucose-measuring working electrodes, the sensing area may also include a reference electrode, or other electrode associated with the working electrode. In these particular embodiments, the sensing area may also include a separate reference or counter electrode associated with one or more optional auxiliary working electrodes. In other embodiments, the sensing area may include a glucose-measuring working electrode, an auxiliary working electrode, two counter electrodes (one for each working electrode), and a shared reference electrode. In still other embodiments, the sensing area may include a glucose-measuring working electrode, an auxiliary working electrode, two reference electrodes, and a shared counter electrode.

[0117] US Patent Publication Nos. US-2008-0119703-A1 and US-2005-0245799-A1 describe further configurations for using continuous sensors at different body locations. In some embodiments, the sensor is configured for percutaneous implantation into the host. In alternative embodiments, the sensor is configured for insertion into the circulatory system, such as a peripheral vein or artery. However, in other embodiments, the sensor is configured for insertion into the central circulatory system, such as, but not limited to, the aorta. In still other embodiments, the sensor may be placed in an extracorporeal circulatory system, such as, but not limited to, an intravascular access device that provides extracorporeal access to blood vessels, an intravenous fluid infusion system, an extracorporeal blood chemistry analysis device, a dialysis machine, a heart-lung machine (i.e., a device used to provide blood circulation and oxygenation while the heart is stopped during cardiac surgery). In still other embodiments, the sensor may be configured to be fully implantable, as described in US Patent No. 6,001,067.

[0118] FIG. 2 is a block diagram illustrating one possible configuration of the sensor electronics in one embodiment. In this embodiment, a potentiostat 20 is shown that is operatively connected to the electrode system and provides a voltage to the electrodes that biases the sensor and allows for the measurement of a current value indicative of the analyte concentration in the host (also referred to as the analog portion). In some embodiments, the potentiostat includes a resistor (not shown) that translates the current into a voltage. In some alternative embodiments, a current-to-frequency converter is provided that is configured to continuously integrate the measured current, for example, using a charge counting device. In the illustrated embodiment, an A / D converter 21 digitizes the analog signal into "counts" for processing. Thus, the resulting raw data stream in counts is directly related to the current measured by the potentiostat 20.

[0119] The processor module 22 is a central control unit that controls the processing of the sensor electronics. In some embodiments, the processor module includes a microprocessor, although as described herein, computer systems other than a microprocessor may be used, for example, an ASIC may be used for some or all of the sensor's central processing. The processor typically provides semi-permanent data storage, for example, storing data such as a sensor identifier (ID) and programming it to process the data stream (e.g., for data smoothing and / or signal artifact replacement, such as described in more detail elsewhere herein). The processor may additionally be used for system cache memory, for example, to temporarily store recent sensor data. In some embodiments, the processor module comprises memory storage components such as various types of ROM, RAM, flash memory, etc. In one exemplary embodiment, the ROM 23 provides semi-permanent data storage, for example, storing data such as a sensor identifier (ID) and programming it to process the data stream (e.g., for signal artifact detection and / or replacement, such as described elsewhere herein). In one exemplary embodiment, RAM 24 may be used for cache memory for the system, for example, to temporarily store recent sensor data.

[0120] In some embodiments, the processor module comprises a digital filter, such as, for example, an IIR or FIR filter, configured to smooth the raw data stream from the A / D converter. Generally, the digital filter is programmed to filter the sampled data at predefined time intervals (also referred to as the sample rate). In some embodiments where the potentiostat is configured to measure the analyte at discrete time intervals, these time intervals determine the sample rate of the digital filter. In some alternative embodiments where the potentiostat is configured to measure the analyte continuously, for example using a current-to-frequency converter, the processor module may be programmed to request digital values ​​from the A / D converter at predefined time intervals, also referred to as acquisition times. In these alternative embodiments, the values ​​obtained by the processor are advantageously averaged over the acquisition time due to the continuity of the current measurements. Thus, the acquisition time determines the sample rate of the digital filter. In a preferred embodiment, the processor module is configured with a programmable acquisition time, i.e., the predefined time interval for requesting digital values ​​from the A / D converter is user programmable into the digital circuitry of the processor module. Acquisition times of about 2 seconds to about 512 seconds are preferred, but any acquisition time can be programmed into the processor module. Programmable acquisition times are advantageous in optimizing noise filtering, time lag, and processing / battery power.

[0121] Preferably, the processor module is configured to construct data packets for transmission to an external source, such as, for example, RF transmission to a receiver as described in more detail below. Generally, the data packets include a number of bits that may include a sensor ID code, raw data, filtered data, and / or error detection or correction. The processor module may be configured to transmit any combination of raw and / or filtered data.

[0122] A battery 25 is operatively connected to the processor 22 and provides the necessary power for the sensor (e.g., 100). In one embodiment, the battery is a lithium manganese dioxide battery, but any suitable size powered battery may be used (e.g., AAA, nickel-cadmium, zinc-carbon, alkaline, lithium, nickel-metal hydride, lithium-ion, air-zinc, zinc-mercury oxide, silver-zinc, or sealed). In some embodiments, the battery is rechargeable. In some embodiments, multiple batteries may be used to power the system. In still other embodiments, the receiver may be powered transcutaneously, for example, via inductive coupling. A quartz crystal 26 is operatively connected to the processor 22 and maintains system time for the entire computer system.

[0123] An RF module, (e.g., RF transceiver) 27 is operatively connected to the processor 22 and transmits sensor data from the sensor (e.g., 100) to a receiver (see FIGS. 3 and 4). Although an RF transceiver is shown here, some other embodiments may include a wired rather than wireless connection to the receiver. A second quartz crystal 28 provides system time for synchronizing data transmission from the RF transceiver. Note that the transceiver 27 may be replaced by a transmitter in other embodiments. However, in some alternative embodiments, other mechanisms may be used to transmit and / or receive data, such as optical, infrared radiation (IR), ultrasonic, etc.

[0124] In some embodiments, a signal artifact detector 29 is provided, which includes one or more of an oxygen detector 29a, a pH detector 29b, a temperature detector 29c, and a pressure / stress detector 29d, and is described in more detail with respect to signal artifact and fault / error detection and discrimination. It should be noted that in some embodiments, the signal artifact detector 29 is a separate entity (e.g., a temperature detector) operatively connected to the processor, while in other embodiments, the signal artifact detector is part of the processor and utilizes readings from the electrodes to detect, for example, signal faults and artifacts. While the above description includes some embodiments in which all discrimination occurs within the sensor, other embodiments provide systems and methods for detecting signal faults within the sensor and / or receiver electronics (e.g., a processor module), as described in more detail elsewhere herein. Receiver

[0125] 3A-3D are schematic diagrams of a receiver 30 including a representation of an estimated glucose value on its user interface in a first, second, third, and fourth embodiment, respectively. The receiver 30 comprises a system for receiving, processing, and displaying sensor data from a glucose sensor (e.g., 100), such as those described herein. In particular, the receiver 30 can be, for example, a mobile phone-type device, and comprises a user interface having physical buttons 32 and a display screen 34, as well as one or more input / output (I / O) devices, such as one or more buttons 55 and / or switches 57, that perform one or more functions when activated or clicked. In the illustrated embodiment, the electronic device is a smartphone, and the display 34 comprises a touch screen, which also functions as an I / O device. In some embodiments, the user interface may also include a keyboard, a speaker, and a vibrator. The functionality of the receiver or smartphone may also be implemented as functionality within an application running on a bullet computer, or similar device. In other embodiments, the receiver may comprise a device or devices other than a smartphone, such as a smart watch, a tablet computer, a mini tablet computer, a handheld personal digital assistant (PDA), a games console, a multimedia player, a wearable device such as those described above, a screen in an automobile or other vehicle, a dedicated receiver device, etc.

[0126] FIG. 3A illustrates a first embodiment in which the receiver 30 shows a numerical representation of an estimated glucose value on its user interface. FIG. 3B illustrates a second embodiment in which the receiver 30 shows an estimated glucose value and approximately one hour of historical trend data on its user interface. FIG. 3C illustrates a third embodiment in which the receiver 30 shows an estimated glucose value and approximately three hours of historical trend data on its user interface. FIG. 3D illustrates a fourth embodiment in which the receiver 30 shows an estimated glucose value and approximately nine hours of historical trend data on its user interface. In some embodiments, a user may toggle through some or all of the screens shown in FIGS. 3A-3D using physical buttons or buttons implemented on a touch screen interface. In some embodiments, a user may be able to interactively select the type of output displayed on their user interface. In other embodiments, the sensor output may have alternative configurations.

[0127] FIG. 4 is a block diagram illustrating one possible configuration of a receiver, such as, for example, a smartphone, electronic device. Note that the receiver comprises a configuration such as that described with reference to FIGS. 3A-3D above. Alternatively, the receiver may comprise other configurations, including a desktop computer, a laptop computer, a personal digital assistant (PDA), a server (local or remote to the receiver), and the like. In some embodiments, the receiver may be adapted to connect (via a wired or wireless connection) to a desktop computer, a laptop computer, a PDA, (local or remote to the receiver), and the like, to download data from the receiver. In some alternative embodiments, the receiver and / or receiver electronics may be housed within a sensor (e.g., 100) or directly connected to the sensor in a manner that allows the sensor and receiver electronics to operate directly together and / or share data processing resources. Thus, the receiver electronics (or any combination of sensor and / or receiver electronics) may be generally referred to as a "computer system."

[0128] The quartz crystal 40 is operatively connected to an RF transceiver 41 which functions together to receive and synchronize a data stream (e.g., a raw data stream transmitted from the RF transceiver). Once received, a processor 42 processes the signal as described below.

[0129] The processor 42, also referred to as a processor module, is a central control unit that performs processes such as, for example, storing data, analyzing data streams, calibrating analyte sensor data, predicting analyte values, comparing predicted analyte values ​​with corresponding measured analyte values, analyzing variability in predicted analyte values, downloading data, controlling the user interface by providing analyte values, prompts, messages, warnings, alarms, etc. The processor includes hardware and software that performs the processes described herein, such as flash memory that provides permanent or semi-permanent data storage, stores data such as sensor ID, receiver ID, and programs to process the data stream (e.g., to implement predictive and other algorithms described elsewhere herein), and random access memory (RAM) that stores cache memory for the system and aids in data processing.

[0130] In one exemplary embodiment, the processor is a microprocessor that provides processing such as calibration algorithms that are stored in ROM 43. ROM 43 is operatively connected to processor 42 and provides semi-permanent data storage, stores data such as receiver ID, and is programmed to process the data stream (e.g., to implement the calibration and other algorithms described elsewhere herein). In this exemplary embodiment, RAM 44 is used for cache memory for the system and aids in data processing.

[0131] A battery 45 is operatively connected to the processor 42 and provides power for the receiver. In one embodiment, the battery is a standard AAA alkaline battery, although any suitable size powered battery may be used. In some embodiments, multiple batteries may be used to power the system. A quartz crystal 46 is operatively connected to the processor 42 and maintains system time for the entire computer system.

[0132] The user interface 47 comprises a keyboard 2, a speaker 3, a transducer 4, a backlight 5, a liquid crystal display (LCD 6), and one or more buttons 7, which may be implemented as physical buttons or buttons on a touch screen interface. The components that comprise the user interface 47 provide controls to interact with the user. The keyboard 2 may allow the user to enter information about himself / herself, such as, for example, meal times, exercise, insulin doses, and baseline glucose values. The speaker 3 may provide auditory signals or alerts, for example, regarding conditions, such as current and / or predicted hyper- and hypoglycemic conditions. The transducer 4 may provide tactile signals or alerts, for reasons such as those described with respect to the speaker above. The backlight 5 may be provided, for example, to help the user read the LCD in low light conditions. The LCD 6 may be provided, for example, to provide the user with visual data output, such as those illustrated in FIGS. 3A-3D. The buttons 7 may provide, for example, toggles, menu selections, option selections, mode selections, and resets.

[0133] In some embodiments, prompts or messages may be displayed on the user interface to convey information to the user, such as requests for reference analyte values, therapy recommendations, deviations of measured analyte values ​​from expected analyte values, etc. Additionally, prompts may be displayed to guide the user through calibration or troubleshooting of the calibration.

[0134] In some implementations, the continuous analyte sensor system includes a Dexcom G4® platinum glucose sensor and transmitter commercially available from Dexcom, Inc. for continuously monitoring a host's glucose levels.

[0135] In some embodiments, the system may execute various applications, such as, for example, a CGM application, which may be downloaded to a receiver or other electronic device via the Internet and / or a cellular network, etc. Data relating to the various applications may be shared between the device and one or more other devices / systems, and may be stored by cloud or network storage and / or on one or more other devices / systems. This CGM application may include a fault determination and response processing module, and / or may include sufficient processing to operate the fault determination and correction functions and methods described below. Data and and situation Overview of fault discrimination and response processing based on

[0136] Referring to Figure 5, a flow diagram 50 illustrates a general method in accordance with the present principles. The method generally involves receiving a signal from a monitoring device, such as, for example, an analyte concentration monitor such as a CGM (step 52). The signal may be "raw," in the sense that the signal represents a count and little or no significant processing has occurred. The method also includes receiving clinical data, which is generally information about the patient environment and clinical context. situation The process includes receiving information (step 54), such as appropriate clinical information for the management of diabetes. situation The information may include food ingested, insulin delivered, patient exercise, patient temperature, clinical glucose values ​​(as distinct from raw signal values), etc. situation , or both, one or both of which may further play a role in the response processing.

[0137] Proper fault discrimination is important in preventing inaccurate clinical glucose values, especially as displayed to the user, which may cause the user to take inappropriate action, may impair the performance of predictive or closed-loop algorithms, and may undermine the user's confidence in their GCM sensor.

[0138] In a method according to the present principles, the fault is then detected, determined, or discriminated (step 56), collectively "discriminated." The fault may be determined based solely on the received signal, or based on the received signal and the clinical situation A response process may then occur (step 58), which may include determining the fault and clinical condition as separate variables or parameters. situation It may be based on the number of failures or only on the number of failures identified (in this case, clinical situation In the particular case of the method, the received signal, or the received signal and the clinical situation The data may be used to determine a classification of the fault, and response action may occur based on the classification of the fault. Other special cases will be understood as well. These general principles will now be explained in more detail with examples. Received signals and clinical situation

[0139] As noted above, systems and methods according to the present principles generally involve one or more received signals, one of which generally relates to a raw sensor signal, e.g., an analyte concentration, e.g., glucose concentration, and other physiological data relating to the patient, clinical data relating to the patient's environment (activity level, pattern, time of day, etc.), and / or other physiological data relating to the patient. situation and performing fault determination and response processing methods based on data relating to each of these aspects. Each of these aspects is described in more detail below. Sensor signal analysis / other signals

[0140] 6 illustrates an embodiment of a received signal 62 and a method of discrimination of that signal. Initially, the fault discrimination and response processing method may be based on a raw signal 64, which is measured by the sensor electrodes and is in the form of an uncalibrated count with respect to time.

[0141] The method may then be based on the processed raw signal 66, but in this case the processing is independent of or precedes the determination of an analyte concentration value as used in clinical value determination. In other words, the processing is independent of or precedes the translation of the raw signal count into units meaningful for patient management, such as diabetes management, e.g., as a value expressed in mg / dL or mmol / L. In other words, the processed raw signal is not calibrated and is not useful per se for clinical value determination.

[0142] The processing performed on signal 66 is performed because non-glucose related artifacts can cause aberrations in the signal, and simple averaging or other processing cannot always grid the signal of such artifacts without losing important glucose concentration data within the signal itself.

[0143] One example of signal processing unrelated to converting raw signals to clinical values ​​includes processing related to noise filtering. Such processing results in a signal 68 that is more or less noisy filtered. Various aspects of noise filtering are described in more detail below. Details of specific processing steps for noise filtering are provided in U.S. Patent No. 8,260,393, issued September 4, 2012, owned by the assignee of the present application and incorporated herein by reference in its entirety.

[0144] For example, in the particular implementation of noise filtering illustrated in Figures 6B-6D, filtering is performed differently with different rates of change to achieve different levels of smoothness. More specifically, for a given received signal, the magnitude of the noise in the signal is measured over a time window, and then the rate of change for that signal is measured over a similar time window. While it is possible to completely remove the noise by filtering, errors are still created in the meantime, the error being equal to the rate of change multiplied by the delay being filtered. Since it is preferable to present smooth data to the user, the longest possible delay may be chosen for maximum smoothness. In other words, the filter delay is chosen to be equal to the magnitude of the noise divided by the rate of change. In this way, minimum error is achieved with maximum smoothness. Figures 6B and 6C illustrate two examples with different rates of change but the same noise magnitude. The resulting filter lengths are also illustrated. Thus, in one type of response processing, the filter delay is altered based on at least one signal characteristic, such as the noise magnitude and possibly the rate of change of the signal.

[0145] In another implementation, and referring to FIG. 6D, a threshold may be set for the level of smoothing not based on the accuracy of the signal, but rather on the user's perception of data quality. A characteristic of the signal may be measured. In one case, a third derivative characteristic (i.e., "jerk") is shown to indicate the signal quality as perceived by the user, which is a useful characteristic, especially when sampling occurs frequently, e.g., every 30 seconds, since it is instantaneous and easily measurable. A set of signals with varying jerk levels may be displayed to the user, and the user may select which signals they would like to see, e.g., the signals that deliver the most useful data to the user. By monitoring the selection, a decision may be made as to what jerk levels are acceptable or unacceptable to the user. Even in such a system, some level of maximum filtering may be implemented to meet the expectations of the user regarding signal smoothness.

[0146] In addition to filtering, other types of signal processing unrelated to conversion of the raw signal to a clinical value will be appreciated.

[0147] The received signal 62, raw signal 64, or processed signal 66 is then situation The signal data may be analyzed, with or without the use of other signals and / or information, to determine faults therein, and results 72 including data regarding the faults may be obtained, upon which response processing may be based. Details of the analysis and determination are described herein. In general, an example implementation may receive signal data and compare the signal data against fault determination criteria to determine or determine fault information.

[0148] Other signals 74 may be used in the discriminant analysis in addition to those of the received bioanalyte (e.g., glucose) electrode signal. Such other signals include those related to the temperature of the sensor and associated electronics, the impedance of the sensor and components, background noise encountered by the sensor, and the like. For example, referring to FIG. 7, a graph 101 is illustrated having a raw signal axis 105 and a temperature axis 107 plotted against time 109. As the temperature increases, one possible effect is to cause a gradual increase or decrease in the signal, which increase or decrease is unrelated to the actual glucose level. A raw data trace 113 is illustrated, which represents the actual glucose value without temperature effects, for example in mg / dL. Trace 115 is also illustrated, which represents the raw data for an elevated internal sensor temperature, where a higher temperature causes a gradual increase in the signal (and thus causes separation of the traces). By establishing a correlation between the elevated sensor temperature and the increased signal, the elevated sensor temperature may be used as an input in the discriminant analysis.

[0149] As another example, signals from other constituent sensors, such as an oxygen sensor, may be used for fault discrimination. For example, if the glucose sensor signal is blocked due to a compression fault (described in more detail below), the compression fault should also block the oxygen sensor. An example is shown by graph 122 in FIG. 8A, where raw signal values ​​are plotted on axis 124 against time on axis 126. Raw signal value 132 is illustrated along with oxygen sensor value 128. Raw signal value 132 falls at or near the same time as oxygen sensor value 128. Thus, detection of blocked signals on both sensors leads to a higher likelihood of a fault caused by compression.

[0150] Referring again to FIG. 6, as another example of the use of other signals 74, a signal from a reference electrode may be used in a fault determination method. For example, if the reference electrode signal drifts or shifts, it may indicate that the reference analyte is depleted. In these cases, the value determined by the reference electrode will be particularly oxygen sensitive. Thus, if the reference electrode value drifts and / or becomes highly oxygen sensitive, a fault of reference analyte depletion may be determined. An example is shown by graph 134 in FIG. 8B, where the ordinate represents the reference electrode signal, the abscissa is time, and curve 138 illustrates a gradual drift downwards of the reference signal electrode potential. In such cases, the response process may include performing a potential sweep to detect a shift in the reference bias. Such a potential sweep may be part of a set of self-diagnostic routines and is described in more detail below.

[0151] As yet another example of the use of other signals 74, if an implantable pump for a drug is used, data may be obtained from the pump insertion set, including data on pressure. Such data may be advantageously used in fault determination. In this regard, it is noted that when a pump is used, scar tissue may grow, which may impede delivery of the drug. Using high speed sampling or other such quantification of the pressure required to operate or initiate operation of a stepper motor inside the pump, a profile may be constructed of the pressure required for the amount of scar tissue, and the profile may be set for the user. In this way, a fault profile of scar tissue formation may be developed, and the fault profile may be used as a signal criterion for fault determination, as may further signals, such as temperature signals. In other words, a signal characteristic or template of scar formation may be determined, and if the signal characteristic or template is found in the signal being evaluated, a fault of scar tissue is determined. Once a fault is determined, it may be used to adjust the delivery and / or bolus delivery, and may be applied to future deliveries. Furthermore, the failure may be used to forestall blockages.

[0152] As yet another example of the use of other signals 74, signals related to impedance measurements may be used between a signal or working electrode (e.g., in the host) and an external electrode (e.g., on the skin), which may or may not be the same as the reference electrode. In this manner, electrochemical impedance may be measured between the physiological environment and the signal electrode. Even more importantly, changes such as increases or decreases in such electrochemical impedance may be used in fault determination. Further details of such impedance measurements are described below.

[0153] Next, a classification 76 of various techniques is considered along with their own set 78 of various techniques. Depending on the implementation, a particular technique or group of techniques may be used from the classification 76 or from the set 78. Most of these techniques generally involve considering normal variance in the signal and detecting if the signal (or signal transformation) deviates from what is expected or predicted by more than a predetermined amount, more specifically, where such deviation is determined using a predetermined confidence level. Normal variance in signals, and their confidence levels, and aspects of their calculation are described in U.S. Patent Publication Nos. US-2009-0192366-A1 and US-2014-0278189-A1, both of which are assigned to the assignee of the present application and are incorporated by reference in their entirety.

[0154] The classification 76 includes a time-based technique 82, a frequency-based technique 84, and a time-frequency ("wavelet") based technique 86. The time-based technique 82 is often considered to be the fastest. It will be appreciated that the analysis may be performed using multiple classifications of techniques. Various types of techniques are described herein.

[0155] For example, a step of raw signal analysis 88 may be performed, which may be performed in the time domain or the frequency domain. This step may be considered as a precursor to or is general to other steps generally performed. For example, such raw signal analysis 88 may include an analysis of the frequency of the noise, since certain faults lead to certain respective prevalent noise frequency values, which may then be used to discriminate between them. In the same manner, the noise may be divided into binary states, such as high amplitude / low amplitude, high frequency / low frequency, etc., and such binary states may be used for fault discrimination. The smoothness of the data, or the lack thereof, may be used for fault discrimination using raw signal analysis. For example, the lack of smoothness may indicate the presence of a fault, and vice versa. Referring to graph 142 of FIG. 8C, raw signal values ​​144 are illustrated with a noise region 146. All other factors being equal, it is more likely that a fault has occurred in the noise region 146 than in the rest of the curve. This same determination would also result from frequency analysis.

[0156] Signal changes that are not physiological may be detected by raw signal analysis. For example, there may be maximum and minimum values ​​for the physiological rate of change of glucose, and if a rate of change greater than the maximum or less than the minimum is measured, it may indicate a fault. For example, referring to graph 148 of FIG. 8D, a raw signal value 152 is illustrated with a sudden drop 154. The sudden drop 154 ​​may be of a magnitude greater than that which would be potentially or routinely encountered in a physiological system, for example, a user's raw signal value would not be expected to show such a drop (or conversely, an increase above a normal physiological threshold). Physiological criteria may be determined based on a priori dates from a particular patient or set or patient, and may be further individualized, for example, to the person's normal glucose profile. Thus, a non-physiological apparent glucose change may be determined as a fault.

[0157] The direction of the signal artifact may also be considered as part of the raw signal analysis. For example, as explained in more detail below, if the analyte concentration has a sharp downward trend accompanied by noise, it may be associated with a compression or "drop recovery" failure. An example is shown in the portion of the curve indicated by area 154 in FIG. 8D. Similarly, if the raw signal has a sharp upward trend accompanied by noise, it may be associated with an electrical short failure, for example, due to water ingress into the contact area of ​​the transmitter.

[0158] Referring again to FIG. 6, other signal processing involving raw signal analysis 88 is similarly understood, including complex frequency based analysis such as high pass filters, low pass filters, and matched filters.

[0159] Similarly, a step of residual signal analysis 92 may be performed in which raw signal data is analyzed relative to filtered signal data. More specifically, in yet another method for fault determination involving inspection or evaluation of signal information content, filtered (e.g., smoothed) data is compared to raw data (e.g., in the sensor electronics or in the receiver electronics). In one such embodiment, a signal "residual" is calculated as the difference between the filtered and raw data. For example, at a point in time (or a period of time represented by a single raw value and a single filtered value), the filtered data may be measured at 50,000 counts and the raw data may be measured at 55,500 counts, which would result in a signal residual of 5,500 counts. In some embodiments, a threshold may be set that represents a first level of noise (e.g., signal artifacts) in the data signal when the residual exceeds that level (e.g., 5000 counts). Similarly, a second threshold may be set that represents a second level of noise in the data signal (e.g., 8,000 counts). Further thresholds and / or noise classifications may be defined as will be appreciated by those skilled in the art, so that signal filtering, processing, and / or display decisions may be performed based on these criteria (e.g., a given level of noise).

[0160] While the example described above illustrates one method of determining the level of noise or signal artifact(s) based on a comparison of raw data for a time point (or a single value representing a time period) to filtered data, a variety of alternative methods are contemplated. In an alternative exemplary embodiment for determining noise, the signal artifact is evaluated for noise episodes that last for a certain period of time. For example, a processor (in the sensor or receiver) may be configured to look for a certain number of signal remnants above a certain threshold (representing a noise time point or period of time) over a predetermined period of time (e.g., minutes to hours or more).

[0161] In an exemplary embodiment, the processor is configured to determine the signal residual by subtracting the filtered signal from the raw signal over a predetermined period of time. The filtered signal may be filtered by any known smoothing algorithm, such as those described herein, such as a three-point moving average type filter. The raw signal may include an average value, e.g., values ​​are integrated over a predetermined period of time (e.g., over 5 minutes). It is further noted that the predetermined period of time may be representative data for a time point or a period of time (e.g., 5 minutes). In some embodiments, if a noise episode over a predetermined period of time is being evaluated, the difference may be obtained by comparing the signal residual to a previous signal residual (e.g., comparing the residual at time (t)=0 to the residual at (t)=5 minutes). Similar to the thresholds described above for the signal residual, one or more thresholds may be set for the difference, whereby one or more differences above one of the predefined difference thresholds define a particular noise level. It has been shown in certain circumstances that difference measurements may be more sensitive to noise episodes when compared to the residual measurements described herein, without amplifying the noise and increasing false positives due to fast but physiological rates of change. Thus, a noise episode or noise episode level may be defined by one or more points (e.g., residual or difference) above a predefined threshold, and in some embodiments, over a predefined period of time. Similarly, the noise level determination may be reduced or altered when a different (e.g., reduced) number of points above a predefined threshold may be calculated within a predefined period of time.

[0162] In some embodiments, one or more signal residuals may be obtained by comparing the received data to the filtered data, thereby determining a signal artifact. In some embodiments, it is determined that a signal artifact event has occurred if the residual is greater than a threshold. In some exemplary embodiments, it is determined that another signal artifact event has occurred if the residual is greater than a second threshold. In some exemplary embodiments, it is determined that a signal artifact event has occurred if the residual is greater than a threshold over a period of time or amount of data. In some exemplary embodiments, it may be determined that a signal artifact event has occurred if a predetermined number of signal residuals that exceed a predetermined threshold occur within a predetermined period of time (or amount of data). In some exemplary embodiments, an average of multiple residuals is evaluated over a period of time or amount of data to determine if a signal artifact has occurred. The use of residuals for noise detection may be preferred in situations where data gaps (non-contiguous) data are present.

[0163] In some exemplary embodiments, the difference, also referred to as the derivative of the residual, is determined by comparing a first residual (e.g., at a first time point) with a second residual (e.g., at a second time point), where a signal artifact event is determined to have occurred when the difference exceeds a predetermined threshold. In some exemplary embodiments, a signal artifact event is determined to have occurred if the difference is greater than the threshold over a period of time (or amount of data). In some exemplary embodiments, an average of multiple differences is calculated over a period of time or amount of data to determine if a signal artifact has occurred. Other details of residual analysis are described in U.S. Patent No. 8,260,393, incorporated by reference above.

[0164] Referring again to FIG. 6, a pattern analysis 94 may also be performed, which may lead to a particular expected or predicted change in signal value that would be measured in the absence of a fault, so that if a change in signal that fits that pattern is measured, a fault need not be determined. Without pattern analysis, a similar change in signal value may be sufficient to lead to a fault being misidentified. Conversely, if a signal is received that does not fit the pattern, a fault may be detected, and a signal characteristic and / or clinical situation Depending on the pattern analysis, a fault may be determined.

[0165] More specifically, certain signal characteristics or patterns may be indicative of or be signatures of various faults, and when such signal characteristics or patterns are seen in a subsequent signal, it may provide evidence that the corresponding fault has reoccurred. An example of the use of signal templates is provided below. A template is determined for a given fault, and the signal is projected onto the template to determine how much of the signal can be attributed to the template waveform, and thus to the fault associated with that template waveform. This is described in more detail below.

[0166] Further details of such pattern analysis techniques are provided in U.S. Patent Publication Nos. US-2013-0035575-A1 and US-2014-0129151-A1, both of which are assigned to the assignee of the present application and are hereby incorporated by reference in their entireties.

[0167] Another step that may be performed for signal discrimination is that of "slow vs. fast" sampling (step 96). In these techniques, data is sampled simultaneously or sequentially at two or more different sampling rates. Such techniques may be performed continuously or at specific times, such as during a "self-diagnosis" mode. For example, data may be sampled at both 30 second and 5 minute intervals. Data sampled at 30 second intervals is more granular and may reveal noise components and fault-related features, especially high frequency noise components, that are not evident from data sampled at 5 minute intervals.

[0168] For example, referring to FIG. 9A, data sampled at 30 second intervals is illustrated by solid line 162, and data sampled at 5 minute intervals is illustrated by dotted line 164. Along line 162, a sudden drop is seen at time 166, with a corresponding sudden rise at time 168. This drop and rise are characteristic of a "squeezing" fault, for example, where the user's weight or part of it crashes onto the sensors and associated electronics (drop) and then the bump is removed (rise). FIG. 9B illustrates another example of the use of slow vs. fast sampling, where not only can fast sampling provide better curve definition for fault discrimination, but additional features can also be gleaned from the data. For example, data sampled every 5 minutes is sufficient to know the glucose concentration and show certain spikes that are likely due to noise. However, inspection of the same data sampled every 30 seconds clearly shows the presence of high frequency noise portions 167. Thus, the use of both slow and fast sampling provides better noise discrimination and reduced time lag before noise is announced and responsive processing occurs. Analysis of such high frequency noise components using fast sampling further enables accurate end of life detection methods, extends wear duration, and creates more efficient replacement claim procedures. Another example of end of life detection is provided below.

[0169] As another example of signal processing, a step using fuzzy logic 103 may be used. This can be conceptually applied to any noise detection scheme, where the noise detection measures the level of noise, rather than a binary or broadly classified noise level scheme. Specifically, using a fuzzy filter, filtering may be applied more gradually or smoothly by adaptively weighting the raw signal against the filtered signal to achieve a gradually more or less aggressively filtered signal. Fuzzy filtering may also be applied to slow vs. fast sampling signal techniques, or indeed with any technique using two different signal resolutions. The fuzziness of the filter may be influenced by the level of noise and / or the clinical situation may be applied based on

[0170] More specifically, as mentioned above, residual analysis can be used in noise management algorithms that include the residual (difference) between raw and unfiltered values ​​or the δ residual, i.e., the change from one residual to another. These algorithms are useful for estimating noise levels. In one implementation, the residual may be passed through three different filters, e.g., one slow moving, one medium moving, and one fast moving, and based on the ratio of the output of the three different filters to a very slow moving average, the algorithm can determine whether the noise condition is clean, mild, moderate, or severe.

[0171] One problem with such techniques is that they are binary: in some cases the signal is "clean" and the delay or time lag in the signal is related only to the sampling periodicity, e.g., 5 minutes. In other cases, filtering is applied and the time lag is related to the sampling window of the filter, e.g., 10 minutes. For a noisy signal, this long time lag can be problematic, especially when the user's glucose level is dropping fast, e.g., -5 mg / dL / min. The use of fuzzy logic, and specifically, a fuzzy filter, can reduce this delay as follows:

[0172] Specifically, the estimated glucose value may be determined by the following equation: EGV = (count-baseline) / slope Where, count = 2 x α x filter(N) + (1-α) x raw(N) where Filter(N) represents the filtered signal, Raw(N) represents the raw signal, α is a weighting factor that is close to zero when the residual or δresidual is small and close to one when the residual / δresidual is large, and α may be described by any of a variety of continuous functions, but is often linear and monotonically increasing.

[0173] The calculation of α may vary based on the underlying model used. At each point, the absolute residual / δ residual may be calculated and thus a new weight may be calculated at each point. In addition to the absolute residual, other metrics may be utilized, such as lightly filtered residual, moderately filtered residual, heavily filtered residual, ratio of lightly filtered residual to slow moving filtered residual, ratio of moderately filtered residual to slow moving filtered residual, ratio of heavily filtered residual to slow moving filtered residual, etc. Signed residuals (negative / positive) may be utilized to manipulate time lags, such as sensor lag in glucose rise and sensor lead in glucose fall. For example, if the underlying trend is falling and the sensor is leading, further filtering may be provided. On the other hand, if the trend is rising, the raw signal may be averaged if the current residual is small, and projected values ​​may be calculated over 5 minutes from the current time, and further weights may be given to the projected values ​​on the raw values. In this manner, the fuzzy filter may be applied gradually. In this way, when little noise is present, filtering may occur, but only lightly and not aggressively, whereas when significant noise is present, filtering may be applied more aggressively.

[0174] In an even more sophisticated implementation, the concept of a fuzzy unit (FU) may be defined as shown below:

number

[0175] When the first fuzzy unit is calculated, the filter may start as follows: Current noise percent = FU, Prior noise percent = Prediction prior to FU, RawError[n] = Error between current noise percent and previous noise percent, smoothederror[n] = filtered error between current noise percent and previous noise percent, e.g., 0, α = smoothing factor between previous and current noise percentage, e.g., 0.65; β = smoothing factor between smoothed error and raw error, e.g., 0.65, and

number

[0176] The noise weights are used to create new filtered coefficients as shown below:

number

[0177] The relationship between the fuzzy unit FU and the noise weight takes the form illustrated in Figure 10 A. As can be seen, as the noise increases, the filter is applied to an increasing extent.

[0178] Referring to flow diagram 75 of FIG. 10B, after the first fuzzy unit is calculated, the following steps may be performed for each subsequent calculation of the fuzzy unit (current noise percentage) (step 77).

[0179] First step to calculate and store the raw error (step 79):

number

[0180] If n≧3, then β is updated (step 81) as shown below: 2 point error = |raw error [n-2] - raw error [n-1]|, 1 point error = |raw error [n-1] - raw error [n]|,

number

number

[0181] Also now, for example, if μ=0.75 and σ=0.29, then β takes the shape shown in FIG. 10C, which shows that as the variance of the error increases, β increases and thus more emphasis is placed on the current smoothed error.

[0182] The next step is to update the smoothed error (step 83):

number

[0183] The next step is to update α (step 85):

number

[0184] Also, α takes the shape illustrated in FIG. 10D.

[0185] The next step is to make a prediction of the current noise percentage (step 87): Previous noise percent = α × current noise percent + (1-α) × previous noise percent.

[0186] After this step, the noise weights are updated (step 89):

number

[0187] Next, new filter coefficients are calculated (step 91):

number

[0188] The filter counts and raw counts are then fed into an appropriate noise management algorithm (step 93 (FIG. 10B)), which then updates the points and noise conditions, for example using probabilistic thresholds. Exemplary probabilistic thresholds are shown in the table below: [Table 1]

[0189] It will be appreciated that other thresholds may also be used.

[0190] In tests, the fuzzy filter has provided significantly faster response to noise spikes and quicker recovery from noise episodes than demonstrated by previous attempts involving simple filtering. The fuzzy filter has also demonstrated greater accuracy than such previous attempts.

[0191] In yet another variation in the signal analysis method, metrics may have associated weights, which may be standard or may vary depending on the metric. In this variation, adjustments are made for the observation that some metrics are more indicative of certain faults than others. For example, skewness and variance are more indicative for oxygen noise (indicated by high frequency noise and a downward trend) than for shower spikes (a smooth upward rise).

[0192] It will be appreciated from this disclosure that other types of signal processing may also be used according to implementation, and other factors, parameters, and variables may be used in fault determination. For example, a timestamp of the data may be used in certain analyses, for example, to detect certain time-based patterns or to determine the implantation age that is strongly related to the end of life determination. In this regard, it is noted that in some cases, the raw signal data is correlated with established patterns of the patient. For example, raw sensor data indicating a potentially faulty condition due to an abnormally high signal value may initially appear to indicate a fault, but may be caused by the user eating a normal meal. The determination that the user has eaten a normal meal may be by timestamp data and machine learning (or other techniques) where a pattern may be established. Similarly, a spike in the data at a certain time of day may indicate a water-related error, for example, related to daily showers. Similarly, other types of faults may be more likely to occur at night, for example, compression artifacts.

[0193] Other types of signal processing may include analysis of the duration of time since implantation of the analyte monitor, which may be particularly important, for example, to check to account for malfunctions or errors that may occur over time or for older implants.

[0194] Finally, it is noted that where particular factors, parameters, or variables are set forth above, the corresponding durations over which they occupy a range of values ​​may be used in fault discrimination, the range of values ​​being narrow or broad. Combinations of the above factors, parameters, and variables may also be used. clinical situation information

[0195] Here, clinical situation Types of data corresponding to information are described. These types of data relate to the observation and treatment of actual patients, rather than simply looking at internal signals that are unknown to the patient. However, clinical situationIt should be understood that information may be derived from the internal signal using processing that converts the internal signal into data relevant to actual patient observation and treatment. For example, the raw sensor signal is unknown to the patient, while the actual glucose concentration, calibrated and converted from the raw signal, may be known to the patient based on clinical practice. situation Similarly, while temperature and time information may be considered internal information, the specific clinical information of the temperature and time may be subject to change without notice. situation Comparison with criteria (e.g., sleep / wake, shower, sedentary / active information) was performed situation Generally, such clinical situation The information includes anything that affects a person with a condition, such as diabetes, including social, emotional, physical and environmental influences, and indeed anything that may relate to or affect the physiological health or environment surrounding the patient. situation Identifying information provides additional intelligence for purposes of fault determination and response processing.

[0196] Figure 11 shows the clinical situation Illustrates various types of data or other information that may constitute or be involved in the determination of information 186. situation The primary trigger for information 186 is glucose concentration data 184. This data may include aspects 188, such as the actual clinical value, its rate of change, acceleration, higher order derivatives, etc. Glucose concentration data 184 may also include aspects 192, such as the range of glucose concentration, e.g., the range maintained by the patient's glucose concentration, and the duration the user has a clinical glucose value within a particular range. Similarly, conditions may be defined and used as ranges, e.g., hypoglycemia, hyperglycemia, or euglycemia. Such condition data may also include impending, predicted, or expected conditions. Other possible triggers for glucose concentration data 184 may include whether the patient is in a stable state or experiencing a change in their glucose concentration.

[0197] The concentration of the specimen, such as glucose concentration, situationWhen used as information, for example, mg / dL or mmol / L, or mg / dL / min or mg / dL / min 2 These constitute information that has been translated from raw signals, such as time derivatives including , into values ​​that are meaningful for diabetes management. These are distinct from "sensor data" because they have been calibrated for clinical relevance. Quantities derived in part from glucose concentration information may also be used, such as the Glucose Urgency Index ("GUI"), Dynamic Risk ("DR"), and Static Risk ("SR"), etc.

[0198] In many implementations, clinical situation The information will include or be determined from at least some aspect of the analyte concentration, e.g., glucose concentration, or a quantity derived in part from the analyte concentration as described above, e.g., glucose rate of change, glycemic state, GUI, etc. In other words, the clinical situation When the information is used in fault discrimination and response processing, the clinical glucose value or a condition associated therewith will be used. For example, given a particular fault, the response processing may often depend on whether the user is hypoglycemic, hyperglycemic, or euglycemic. Such conditions may relate to whether glucose display is suspended and / or a warning is given to the user. Similarly, when a fault causes the user's glucose value to be unknown or uncertain, the response processing may often use the rate of change, as this will be strongly related to whether the glucose level was rising or falling prior to the fault, and the rate of rise or fall. Higher order derivatives may be used to determine whether the user's glucose level is likely to return to normal or whether a further excursion is expected. Thus, clinical glucose values ​​and related parameters are often used in fault discrimination and response processing.

[0199] With respect to non-glucose concentration related information, it may include information and data relating to age 194, because such data may be used for clinical purposes. situationIn other words, sensor data may be considered faulty for one patient and not for another, and age may be an indicator of "which bucket the data falls into." For example, for very young and elderly patients, fault data may be more imminent than for a more robust young adult patient, so each data point may be given significant weighting. Thus, the system may be configured to be particularly sensitive in such situations and for such patients, and therefore to discriminate more faults. In general, age is a predictor of clinical outcomes. situation Numerous clinical factors may be considered in determining information 186. situation It will be appreciated that the data variables are one and the same and therefore the actual resulting fault discrimination and response processing behavior will depend on a number of factors.

[0200] Similarly, anthropometric data 196, generally relating to physical information such as BMI, may also be relevant to determining whether a malfunction has occurred. One method of measuring anthropometric information for such uses is by measuring impedance from the tip of the sensor to the base patch, as described above and in more detail below. While the above discussion relates to measuring impedance to determine internal aspects of the sensor or sensor electronics, it will be appreciated that impedance measurements may be used in conjunction with certain clinical indications. situation It should be noted that the impedance measurements may be used to determine external aspects of the information, including clinical, tissue type, BMI, etc. situation This may result in information decisions.

[0201] A further embodiment includes data 185 regarding whether medications such as insulin have been administered. In this case, raw sensor data indicating a potential fault condition due to an abnormally low signal value may indicate a fault, but may also be caused by a recent injection or bolus of medication such as insulin. situationTaking into account the above, what might otherwise be considered a fault may be determined to be actual physiological data, i.e., not a fault. Conversely, if a recent injection or bolus has been administered but the signal is abnormally high, this may increase the likelihood of a fault being determined. Data regarding potentially interfering drugs may also be considered in the fault determination.

[0202] Another type of clinical situation The information is clinical situation To determine the information, clinical situation The data includes data on external temperature compared to a standard 187. In general, temperature data and its comparison to a standard may be useful in assessing the status of a particular patient in other clinical settings. situation Such clinical information is combined with situation The information may indicate that the patient has had a hot tub, taken a shower, exercised, has a fever, etc.

[0203] Yet another embodiment includes data 191 regarding the patient's activity level, and data 193 regarding user-related movement. Data 191 or 193 may be inferred from another wearable sensor, as well as from a "fit band," a gyroscope, an on-skin electronic accelerometer, a smartphone or smartwatch accelerometer or GPS, or via patient input and other means. Data 191 may be quantitative or qualitative, and in the latter case may be measured, for example, as "sedentary" or "active." Other stages will be understood as well. Data 193 may also be quantitative or qualitative, and in the former case may provide an indication of the amount of movement the sensor underwent, the duration the movement occurred, and derived quantities such as calories burned.

[0204] clinical situationYet another source of information 186 may be data 189 regarding a patient's failure history. In particular, certain patients may have a particularly active failure discrimination history. Such patients may have a high wound response, experience early wound effects more frequently, etc., which increases the overall likelihood of such failures being experienced in future sessions for that patient. Other patients may be more prone to pressure failures, and this tendency may be factored into the analysis.

[0205] clinical situation A further incentive for information 186 is data 195 regarding clinical use of the data. This incentive relates to how the data is used, for example, in a closed loop system, an open loop system, an artificial pancreas system, or with an integrated pump. More specifically, if the data is used within a closed loop system, it may provide a driver for a pump that administers a drug, such as insulin. In a closed loop system, the determination of whether a given signal has failed may be more conservative, for example, since drug delivery is signal dependent. In other words, the system may provide a signal that is consistent with this clinical use. situation In an alternative implementation, the pump drive algorithm may have its own fault discrimination routine and the system may be configured to be less discriminating or less sensitive to faults.

[0206] The above assumes a potential fault signal to drive the pump. Pump information may also be used in the reverse fashion to supplement, inform and drive fault detection. For example, if a large bolus of insulin was recently injected, a negative rate of change in glucose would be expected, but a positive spike in the signal would not be expected. Thus, if a positive spike in the signal occurs, it is more likely to be a fault than an actual glucose excursion. It will be appreciated that variations on the use of "clinical use" data are also provided in the present teachings.

[0207] clinical situation Another source of information 186 is data 197 regarding patient interaction with their analyte monitor, such as, for example, a CGM. The level to which a patient or user interacts with their analyte monitor may be a clinically relevant factor. situation Information may be a factor in determining whether or not a patient's CGM is being used. For example, if a user does not consult or interact with their CGM very often, each interaction, i.e., each data point received by the user, may be considered to receive significant weight in the user's management of their disease, based solely on the relative rarity of the data point the user encounters. Conversely, for a highly interactive user, each data point is important, but a failure may be less of a risk, since the user is more likely to receive another data point relatively soon. A patient's interactions with their CGM may be tracked by button presses, menu selections, or screens viewed, calibrations, etc. on the receiver.

[0208] Other clinical situation The information will be understood in a similar manner. For example, correlation with normal glucose behavior may be of clinical interest. situation The information may be constructed based on the patient's glucose levels. In this example, a pattern of glucose values ​​may be established for the patient. Such a pattern may be time-based or event-based, but generally is indicative of the patient's normal glucose behavior. Time-based patterns may be based on the hour of the day, week, month, etc. The current glucose level may then be compared to normal or expected glucose patterns and profiles for the same hour of the day, week, or month, respectively, and deviations from the mode may then be indicative of a clinical event. Wavelet correlation may be used in this analysis.

[0209] Localized pressure around the sensor may be of clinical value as it may detect certain patient movements (or lack thereof) that may affect sensor function. situationA suitable pressure sensor may be incorporated onto or adjacent to the sensor and / or sensor electronics described above by including strain gauges or piezoelectric material in the framework or outer body of the transmitter body, or in a pressure plate, gauge, or material within the base (e.g., flexible portion) of the body. Such thin film sensors may be used for pressure detection and quantification. Generally, however, the use of such data is limited to clinical applications. situation As an input to the overall determination of, and not simply to determine sensor function itself. In one implementation, pressure may be used as an input in determining whether a patient is moving, sedentary, awake, asleep, etc. For example, pressure spikes detected by such sensors may be combined with pattern data and / or time of day data (e.g., the patient typically goes to bed at the same time that increases in pressure occur), as well as patient motion data (e.g., the patient is exhibiting little or no motion). These signals evaluated together may be combined to provide a clinically relevant indication of whether a patient is moving, sedentary, awake, asleep, etc. situation Further details of certain of these aspects are provided in US PGP2012 / 0078071, which is owned by the assignee of the present application and is incorporated herein by reference in its entirety. Additionally, exemplary thin film sensors are described below, which may be integrated into the sensor electronics or transmitter housing.

[0210] The specific types of clinical situation Information may be provided by the patient and / or entered into the monitor, particularly if the monitor is embodied by a smartphone or other device with a substantial user interface. For example, the user may be queried about meals eaten, exercise performed, etc. In some cases, questions to the user may be prompted by the fault to disambiguate the fault. For example, the user may be queried as to whether they have been lying on their sensor, for example, to determine a compression fault. Other questions will be understood as well.

[0211] Questioning of the patient may be prompted to determine whether a meal preceded the malfunction in an attempt to resolve an ambiguous rise in signal value. For example, if questioning determines that the patient recently ate a meal, it would be more likely that the rise in signal value could be attributed to a post-prandial rise rather than an error or malfunction. Such information may also be provided by a processor module in data communication with, for example, a food intake application, a camera for photographing the meal, etc.

[0212] Similarly, a patient question may be prompted to determine if insulin uptake preceded the failure. Insulin information may also be provided by the integrated pump. For example, a sudden drop in the raw signal may be determined in this manner to be a drop in the insulin bolus, rather than a failure.

[0213] Another type of clinical situation Information is behavior or The situation Such information may correspond to how the patient uses their device and therefore includes specific data determined by the device. The situation Behavior also The situationInformation may be acquired through the system and may include amount of interaction, glucose alert / alarm status, sensor data, number of screen hits, alarm analysis, events (e.g., characteristics associated with user response, time to response, glycemic control associated with response, user feedback associated with alarm, not acknowledging alerts / alarms within X minutes, time to acknowledgement of alert / alarm, time in alarm state, etc.), diabetes management data (e.g., CGM data, insulin pump data, insulin sensitivity, patterns, activity data, calorie data), data on fatty acids, heart rate during exercise, IgG-antigliadin, stress levels (sweat / perspiration) from skin patch sensors, free amino acids, troponin, ketones, adipanectin, sweat, temperature, etc. Input may be provided by a sensor in data communication with the monitoring device. In some implementations, information may be acquired through an intermediary such as a remote data store. In some situations, patients may use more than one device to track their diabetes (e.g., glucose displayed on a medical device receiver and smartphone).

[0214] clinical situation May be provided for information Situation Information may include personal biology, location, sensed surroundings (e.g., light, sound levels), and environmental data (e.g., weather, temperature, humidity, barometric pressure). Input may be received via peer-to-peer or mesh networks via machine-to-machine communications. .situation The information may include daily routine information from a calendar application, which may vary, particularly from weekdays to weekends. .situation The information may include the number of times the monitoring device is touched or squeezed, even when no interaction is occurring, based on detected device movement.

[0215] The photo is ,situationFor example, a picture of one or more of a glucose meter reading, an insulin pen or pump IOB, a place (e.g., a gym, a park, a house, an Italian restaurant), or a meal may be provided. ,situation The photograph may be used to provide information about the patient's condition, such as the type of insulin used, the amount of insulin consumed, and the type of insulin used. situation may be provided to the monitoring system as a useful contribution of information .situation may also be provided by a basal or bolus setting provided to or determined by the monitoring device.

[0216] situation / Clinical behavioral data situation Other inputs to the information include: Non-situation Data types referenced elsewhere in the behavior input may include, for example, exercise information from a fitness bike or the like, glucose sensor information from a blood glucose (BG) meter or CGM, insulin delivery from an insulin delivery device, insulin on board calculations for the device, and other devices that provided or calculated information. Status / Behavioral data inputs may include hydration level, heart rate, target heart rate, internal temperature, external temperature, external humidity, body analytes, hydration input, output (cycle), sweat rate, rhythm, and adrenaline levels, stress, illness / disease, metabolism / calorie burn rate, fat breakdown rate, current weight, BMI, desired weight, target calories (consumed) per day, target calories (extended) per day, location, preferred foods, and level of exertion.

[0217] The above behavior also The situation For any of the inputs, the system may be configured to receive and / or generate analytical metrics based on the inputs. For example, a composite value may be generated based on the user's glucose level, temperature, and the data time that generated the index value. The composite value may then be used to analyze behavior and and situation Clinical information situationThis may be taken into consideration in determining the contribution of the information.

[0218] This information may be collected from a variety of sensors internal or external to the device, such as accelerometers, GPS, camera data, and third-party tracking applications, including sleep cycle applications. For example, such tracking applications may use geolocation information to The situation and behavior may be determined. ,situation The behavior may also be determined through the use of available social networking information about the user, where a social networking feed associated with the user may be used to situation It is provided to provide a source of data in forming the information.

[0219] situation Further details regarding the behavioral information can be found in U.S. Patent Publication No. US-2015-0119655-A1, which is owned by the assignee of the present application and is incorporated by reference in its entirety, specifically FIG. 4 and the accompanying text.

[0220] Signals and Signal Analysis, and Clinical situation The information is a description of a particular method Status , and are further described below in some examples.

[0221] As shown in Figure 5, situation The information may be used for fault discrimination and response processing. The following figures detail these methods. Referring first to FIG. 12A, a flow diagram 222 illustrates how fault discrimination is performed in a clinical setting. situation The first regime (Regime I) is illustrated, which arises from signal analysis alone, regardless of information. In this regime, the first step is to determine the fault from the signal alone (step 228). Then, the clinical situation The information may be received or otherwise determined (step 232). The determined fault is then ,situation In other words, the response is based on two separate metrics: signal data and clinical situationwhere the former is used as the single metric for discriminating between faults and the latter determines the response signal processing (given that particular discriminated fault).

[0222] The table of FIG. 12B illustrates regime I in another way. First, a signal behavior SBi leads to a corresponding determined fault DFi. Similarly, ,situation The variables VCj are the corresponding clinical situation This leads to the data CDj. Then, using SBi and CDj, the response process RPji is determined.

[0223] For example, referring to graph 236 of FIG. 12C, where the abscissa axis 126 represents time and the ordinate axis 136 represents the raw signal as previously described, a sharp downward trend 237 (SBi) is seen in the raw signal 235 along with noise, which then plateaus and has less noise or variability 239 in the plateaued portion. These signal characteristics may by themselves indicate a compression failure (DFi). situation If one of the variables (VCj) known about indicates that it is a normal time for the user to go to bed (CDj), the response action (RPji) may be to do nothing. situation If one of the variables known about indicates that the user is unlikely to be asleep (CDk), the response action may be to perform a self-diagnosis (RPki) or to prompt the user to check their sensors.

[0224] As another example, the same detected fault may situation For example, variations in response processing may occur based on whether the measured glucose level is high or low, or whether the rate of change in glucose level is slow or fast.

[0225] Next, regime II is illustrated in FIG. 13. FIG. 13A shows the failure discrimination in clinical situationFIG. 2 shows a flow diagram 224 of the signal processing system that combines information resulting from signal analysis with the analyte monitor. In this regime, the first step is to receive a signal from the analyte monitor and, optionally, perform any of the various signal processing functions described above (step 242). Before, after, or simultaneously with receiving the signal, a clinical situation The signal and the received clinical information may then be received or determined (step 244). situation Based on the data, a fault is determined (step 246). In other words, two separate metrics, namely, signal data and clinical situation is used to determine the fault. The fault is responded to appropriately based on the fault itself (step 248).

[0226] The table in FIG. 13B illustrates regime II in another way. First, the signal behavior SBi is situation The faults CDj are used in combination with the corresponding discriminated faults DFji, which are based on clinical, rather than on signal data alone. situation The faults identified are also based on the data.The response process RPji is then based directly on the identified faults DFji.

[0227] For example, referring to graph 236 of FIG. 13C, a sudden upward trend 238 (SBi) in the raw signal could potentially indicate a water ingress fault, but the indication is ambiguous since other factors could cause such behavior as well. situation If one of the variables (VCj) known about indicates that it is a normal time for the user to take a shower (CDj), then CDj may be used in combination with SBi to disambiguate the fault and determine that it is a water intrusion fault (DFji). Then, the response action (RPji) may be to do nothing. Meanwhile, the patient's clinical situation If one of the variables known about indicates that the user is unlikely to be taking a shower (CDk), a fault DFki may be determined and the response action may be to perform a self-diagnosis step (RPki).

[0228] Finally, regime III is illustrated in Figure 14. Figure 14A shows a flow diagram 226 in which fault discrimination is based optionally on clinical outcomes from signal analysis. situation It occurs in combination with information, but clinical situation The information is also used to drive response processing. In this regime, the first step is to receive a signal from the analyte monitor and, optionally, perform any of the various signal processing functions described above (step 252). Before, after, or simultaneously with receiving the signal, the clinical situation Information may be received or determined (step 254). Then, based on the signal, and optionally, clinical situation Based on the data, a fault is determined (step 256). situation Based on both, an appropriate response is made (step 258).

[0229] The table of FIG. 14B illustrates regime III in another way. First, the signal behavior SBi is, optionally, situation are used in combination with the data CDj to derive a corresponding discriminated fault DFji, which is based on the signal data and, optionally, on clinical situation The faults DFji are then determined based on the clinical data. situation It is used in combination with the data CDk to determine the step of the response process RP (k-ji).

[0230] It will be appreciated that the above regimes are exemplary and other regimes are possible as well. For example, referring to flow diagram 261 of FIG. 15, in some cases, the signal data (step 262) and the clinical situation It may not matter how the fault determination is characterized and / or determined, so long as both (step 264) and (step 266) are considered when determining the response action, i.e., fault determination or classification may not be required as a separate step.

[0231] In another variation, a "no discrimination zone" may be defined with respect to one parameter, factor, or variable (collectively a "metric"), such as, for example, clinical glucose value, or with respect to many or all of these, e.g., clinical glucose value, rate of change, smoothness of the trace, etc. In such a no discrimination zone, faults may be prevented or inhibited because the effect or risk of faults is defined to be low. Conversely, "zones of danger" may also be defined where faults are always discriminated and responsive action always occurs.

[0232] In some cases, only a single input is required for clinical situation may be used to determine information, where a single input is situation The results of the comparison are compared against a standard. situation The single input may be based on a signal received from a sensor, such as a CGM sensor, or may be based on a different type of input, such as the time of day. In other cases, multiple inputs may be used to determine clinical information regarding one or more malfunctions. situation The inputs may be used to determine clinical information, where one or all of these inputs are situation They are compared against criteria and / or otherwise combined by mathematical formulas.

[0233] In any of these regimes, the fault being discriminated may be determined to fall into one of several predefined fault classifications, and the response to the fault may then be determined at least in part by the classification into which the fault is classified. An exemplary fault classification scheme is now described.

[0234] With reference to diagram 268 of FIG. 16, a fault classification scheme is illustrated. In this scheme, a fault discrimination classification scheme 272 may be broadly classified into two types: faults that are detectable and treatable without user intervention 274, and faults that are detectable but not treatable without user intervention 276, where user intervention corresponds to the user taking an action to correct the fault, such as, for example, "rolling" in the case of a compression fault, providing information to confirm the fault type, such as, for example, answering a prompted question or providing a reference glucose value, or treating their diabetes without using CGM data, such as, for example, treating their diabetes based on their meter readings. With respect to faults 274, various processing steps may be taken to provide services to the user until such time as the fault is mitigated or otherwise responded to. For example, an estimated or predicted signal 276 may be provided to the user. Alternatively, a processed signal 278 may be provided to a user, where processing may include filtering, smoothing, or other steps required to reduce the effect of the fault. For example, if the signal spikes quickly, which is typical of a water ingress fault, the signal may be replaced with a short-term prediction. Alternatively, if random noise is encountered, the signal may be filtered or smoothed.

[0235] In other classifications, the fault 276 may be detected but not fully responded to by the system. In this case, a warning 282 may be provided to the user that the displayed clinical glucose value or GUI may be inaccurate or should not be trusted. Two examples are given in FIG. 16. First, a drop recovery fault 284 may be faced, for example, corresponding to an early wound response. In such a fault, the user may be warned that their glucose level may not be correct. Another example is a compression fault 286. If this type of fault is detected, the response action may be to prompt the user to change position.

[0236] It is noted that the above relates to response processes or other actions taken after a fault has been determined. The actions or steps of fault determination itself may include steps of signal analysis and, optionally, clinical situation The response process generally follows fault discrimination.

[0237] Diagram 290 of Figure 17 illustrates another classification scheme 292 that may be used for fault discrimination. In classification scheme 292, faults are separated into transient faults 294 and permanent faults 296. Transient faults 294 are faults that tend to be self-mitigating or self-resolving, such as squeeze-related faults 298, shower spike-related faults 302, transient noise-related faults 304, etc. Permanent faults 296 are faults that do not resolve or correct over time, including oxygen noise 306 that is encountered at the end of the sensor's life.

[0238] Diagram 308 of FIG. 18 illustrates another classification scheme that may be used for fault discrimination. In classification scheme 312, exemplary fault classifications are shown, classified by specific technical classifications. For example, faults are divided into faults 314 related to the local environment around the sensor that cause erroneous measurements and faults 316 related to system errors that subsequently cause erroneous signal artifacts. Faults 314 tend to be more compartmentalized and localized. Examples of faults 314 include faults 318 related to compression and faults 322 related to early wound response. Examples of faults 316 include those related to water spikes, which are often caused by seal failures.

[0239] In each of these cases, namely compression, early wound response, and water spike or seal failure, as well as other signal behaviors, there will be predefined signal criteria that are used by a fault discrimination and response processing routine or algorithm. If the received signal meets the criteria, a fault classification will be assigned accordingly.

[0240] The predetermined signal criteria for a compression fault 318 may be based on the type of noise pattern and / or the rate of change of the raw signal, i.e., typically downward. A compression fault is generally not preceded by a post-prandial rise typically associated with the rise in signal with the ingestion of a meal. Another exemplary compression fault signal criterion is that it is rather more binary and not a smooth transition from one state to another. Status Other signal criteria that may be examined in are for signals that appear to follow patterns that are not associated with physiological changes. situation The predetermined criteria of information may include time of day, e.g., during the night when a sleeping user may roll over on their sensor, and accelerometer data, which may also be indicative of sleep, or heart rate information, which may be slower for a sleeping user, as well as impedance data. As an example of compression in an intravenous system where a glucose sensor is placed in a vein, increased impedance may result from the sensor resting against the vessel wall, generating non-glucose reaction rate limiting noise due to, for example, lack of oxygen. Clinical situation The use of impedance data in determining information is more fully described in U.S. Patent Publication No. US-2012-0265035-A1, which is owned by the assignee of the present application and is incorporated herein by reference in its entirety. Exemplary uses of impedance data and devices for calculating impedance are described below.

[0241] Prescribed clinical situation Other data that may be used as criteria include whether a meal has been recently eaten or insulin has been recently delivered, since pressure failure is often not preceded by a meal or drug delivery.

[0242] Another aforementioned failure classification relates to the initial wound response, one type of which is a temporary wound healing response called "drop recovery." Without wishing to be bound by theory, it is believed that drop recovery may be caused by trauma from the insertion of the implanted sensor, and possibly also from inflammation of nerve bundles near the implant, resulting in the nerve bundles reducing blood flow to the implant area. Alternatively, drop recovery may be related to injury to nearby blood vessels, resulting in a vasospastic event. Generally, interruption of localized blood flow in the implant area over a period of time leads to a reduced amount of glucose in the area of ​​the sensor. During this time, the sensor may have reduced sensitivity and may not be able to track glucose accurately. Thus, drop recovery typically manifests as a suppressed glucose signal. Drop recovery often manifests within one day of signal implantation, most commonly within 12 hours of implantation. Importantly, drop recovery resolves normally within 6-8 hours. Identification of drop recovery may provide the patient, physician, or user with information that the sensor is only temporarily affected by a short-term physiological response, and that there is no need to remove the implant, as normal function will likely return within a few hours. Further details may be found in U.S. Patent Publication No. US-2014-0005509-A1, which is owned by the assignee of the present application and is incorporated herein by reference in its entirety.

[0243] Exemplary signal criteria that may be used to detect a drop recovery failure include a decrease in signal severity indicative of the physiological conditions described above, data regarding implantation time, and internal sensitivity measurements. Patterns may also be used, such patterns having been previously identified along with such failures. Signals repeating such patterns may be inferred to be indicative of a drop recovery failure.

[0244] Exemplary Clinical Trials That May Be Used to Detect Descent Recovery Failures situationThe criteria include pattern analysis, where a base pattern is defined by the patient's clinical glucose profile. The current signal can be compared against such patterns to determine if the current signal is outside of the patient's normal glucose pattern. Patterns may also be established and used to determine if the patient is at higher risk for a wound response type breakdown, for example, if the patient has a pattern that leads to such a breakdown.

[0245] In addition to drop recovery, other wound responses may also be responsible for failures and therefore may be classified in their own classification or as part of a broader wound effect classification. Appropriate response processing may then be defined for such failures. Exemplary signal criteria for other wound responses include impedance measurements between the working electrode and an external electrode to measure an increase in electrochemical impedance between the physiological environment and the working electrode. Such impedance measurements are described in more detail elsewhere.

[0246] In another exemplary failure classification, local effects at the sensor may prevent an analyte such as glucose from being properly measured. An example of this type of failure or error includes a sensor membrane being adversely affected. Generally, however, such failures are characteristic of a sensor approaching its "end of life" period. For example, biofouling may cause such failures. In this case, exemplary signal criteria may include the amount of time since implantation of the sensor, and a particular characteristic noise pattern. Other criteria include increased noise at higher glucose levels compared to those at lower glucose levels. In yet another signal criterion that may be used to detect such local effect failures of the sensor, impedance measurements between the working electrode and the external electrode may be used to measure an increase in electrochemical impedance between the physiological environment and the working electrode. Comparative responses at different electrode potentials may also be used.

[0247] In addition to biofouling, oxygen noise may also be a fault caused as a local effect in the sensor or membrane. An exemplary oxygen noise signal criterion includes several episodes of similar signal characteristics. This can be contrasted with the biofouling fault described above, where, for example, several "small" episodes exist before larger episodes begin to appear. In other words, in a biofouling fault, there are several episodes of lower frequency and duration before larger episodes of higher frequency and longer duration appear. In general, an increase in signal frequency over a sensor session may be a criterion for an oxygen noise fault.

[0248] These "local effect" failures are clinically situation may also be used in determining how to respond. In one type of response processing, the monitor may only display or trust glucose values ​​at "low glucose" levels, such as those below 100 mg / dL, because noise is more likely at higher glucose levels.

[0249] Another type of failure includes the end of life ("EOL") of the sensor. Specifically, embodiments of the continuous glucose sensor described herein may have a useful life during which the sensor may provide reliable sensor data. After the useful life, the sensor may no longer be reliable and provide inaccurate sensor data. Signs of EOL may be recognized and may prevent resulting user safety or inconvenience. To prevent use beyond the useful life, some embodiments notify the user to change the sensor after it is determined that the sensor should no longer be used. Various methods can be used to determine that a sensor should no longer be used, such as a predetermined amount of occurrences from the sensor being used for the first time (e.g., when first implanted in a user or first electrically connected to the sensor electronics), or a determination that the sensor is defective (e.g., due to a ruptured membrane, unstable sensitivity, etc.). After it is determined that a sensor should no longer be used, the sensor system may notify the user that a new sensor should be used, for example, by audibly and / or visually prompting the user to use a new sensor and / or by turning off the display or ceasing to show new (or immediate) sensor data on the display.

[0250] In some embodiments, multiple risk factors indicative of sensor EOL may be evaluated, for example, using risk factor command(s), algorithm(s), and / or function(s). Generally, EOL symptoms are progressive, e.g., not every symptom (or episode) is indicative of sensor failure. Each of the risk factors may be evaluated periodically or intermittently upon receipt of sensor data (e.g., every 5 minutes), or more intermittently (e.g., every few hours or daily). Risk factors may be repeatedly determined, averaged, or trended over time, and the results may be used for subsequent processing. In some embodiments, the evaluation of one or more risk factors may be triggered by another event, such as a trended error in the BG meeting one or more criteria (e.g., by outlier detection).

[0251] In some embodiments, detection of EOL may be achieved using a combination of methods that detect each EOL signature or risk factor individually. The combination of methods or signatures may result in improved specificity (e.g., lower false positives). It should be appreciated that an EOL determination method or algorithm may use a combination of risk factors in determining EOL.

[0252] In some embodiments, suitable risk factors may be selected from a list including, but not limited to, the following: number of days the sensor has been in use (e.g., implanted), sensor sensitivity or including magnitude and history, whether there has been a decrease in signal sensitivity (e.g., a change in the amplitude and / or variability of the sensor's sensitivity compared to one or more predefined criteria), noise analysis (e.g., EOL noise factors (skewness, spikeiness, and rotation)), duration, magnitude, and history, spectral content analysis, pattern recognition), oxygen (e.g., concentration, and / or whether there is a predefined oxygen concentration pattern), glucose pattern (e.g., average, variability, meal characteristics, expected behavior vs. unexpected behavior, e.g., glucose not rising as expected after a meal, etc.), error between baseline BG values ​​and EGV sensor values, including the direction of the error (whether BG or EGV is reading higher compared to the other), and a measurement of sensor linearity (or lack thereof). Sensor linearity refers to the consistency of a sensor's sensitivity over a particular measurement range (e.g., 40-400 mg / dL for a glucose sensor). For example, linearity may be assumed when the sensor signal is reading low with low BG and high with high BG, whereas nonlinearity (based on the error between the reference BG value and the EGV sensor value) may be assumed when the sensor signal is reading low with low BG but not high with high BG (does not change or increase beyond a particular BG value).

[0253] One risk factor that may be useful in determining EOL is the number of days the sensor has been in use (e.g., implanted). In some embodiments, the number of days the sensor has been in use is determined in part based on sensor initialization using initial calibration data, operative connection of the sensor with the sensor electronics, user-entered data, etc. In some embodiments, the system may detect sensor restart and use the restart information in determining the number of days since implantation.

[0254] In some embodiments, when a particular threshold, such as, for example, a particular number of days, is met, a particular variable associated with that threshold may be automatically used in the EOL function. For example, if the number of days the sensor has been used is determined to be at least 4 days, the number of days the sensor has been used is automatically used and / or is a simple yes / no indicator that the threshold has been met. In some embodiments, if the number of days the sensor has been used is at least 1 / 3 of the number of days the sensor is approved for use, the number of days the sensor has been used is automatically used. In other embodiments, if the number of days the sensor has been used is at least 1 / 2, 2 / 3, or 3 / 4, etc., of the number of days the sensor is approved for use, the number of days the sensor has been used is automatically used. In some embodiments, the actual number of days the sensor has been used is always used in the EOL function. In some embodiments, the EOL function is implemented after a predetermined number of days of sensor use.

[0255] Additionally or alternatively, time since insertion may be mapped to an EOL risk factor value (e.g., chance of recovery or probability of future sensor failure) because the longer a sensor is used since implantation, the more sensor-tissue interface changes are likely to affect sensor function. In one example, the EOL risk factor value is mapped to approximately 1.0 for days 1-5, and progressively decreases from day 5, reaching 0.5 at day 8, 0.2 at day 10, and approximately 0.1 at day 14. Other values ​​and thresholds may be used as can be appreciated by one of skill in the art.

[0256] Another risk factor that may be useful in determining EOL is sensor sensitivity or whether there has been a decrease in signal sensitivity, including magnitude and history (e.g., a change in the amplitude and / or variability of the sensor's sensitivity compared to one or more predefined criteria). In some embodiments, the processor module may be configured to determine whether there has been an effect on signal sensitivity. For example, for some sensors, their sensitivity drifts upward or remains relatively flat over a large portion of the sensor's life, e.g., 3, 5, or 7 days. Towards EOL, the sensor's sensitivity to changes in glucose may decrease. This reduction may be recognized as a drop in sensitivity that occurs monotonically over several hours (e.g., 12 hours) by determining either: (a) a change in sensitivity (e.g., m=m×glucose+baseline in the raw signal) or (b) a reduction in the sensor raw count signal. For example, the following equation may be used: If the median (raw counts over the last 12 hours) - median (raw counts over the last 12-24 hours) < 2 x standard deviation over the last 12 hours, then the sensor may be close to EOL.

[0257] In some embodiments, other forms of signals describing statistics related to signal sensitivity (e.g., median, percentage, interquartile range, etc.) may be used to detect EOL. In some embodiments, determining whether there has been a decrease in signal sensitivity includes comparing the measured signal sensitivity against a predefined signal sensitivity threshold or profile to determine whether the measured signal sensitivity is within an acceptable range. The acceptable range may be based on a priori information, such as prior to in vitro and / or in vivo testing of the sensor. In some embodiments, the measured signal sensitivity is outside the acceptable range and the signal sensitivity may be automatically used in the EOL function. In some embodiments, the measured signal sensitivity, the change in sensitivity, and / or a predefined indicator of sensitivity decrease may be used as an input or variable in the EOL function.

[0258] In some embodiments, the sensitivity variable in the EOL function is based on a trend in sensitivity during a particular sensor session (e.g., during the life of the sensor in the host). For example, determining whether there has been a decrease in signal sensitivity includes comparing a first measured signal sensitivity at a first time point to a second measured signal sensitivity at a second time point to determine whether the rate of change in the measured signal sensitivity is within an acceptable range. The acceptable range may be determined by a priori information, such as prior to in vitro and / or in vivo testing of the sensor. In one example, a change of more than 20% over a day may be indicative of EOL and may be useful as an input in the EOL detection function. In one example, a rate of acceleration (e.g., rate of sensitivity drop) of more than 20% over a 12-hour period may be indicative of EOL and may be useful as an input in the EOL detection algorithm.

[0259] In some embodiments, the rate of change in signal sensitivity may be determined based in part on a slow moving average of the raw sensor data (e.g., counts). This embodiment takes advantage of the fact that for most patients, average glucose over time (e.g., over several days) remains relatively constant, and thus a change in the average of (uncalibrated (raw or filtered)) sensor data (e.g., over 2, 3, 4, 5, 6, 7 days) may be interpreted as a change in the sensitivity of the sensor over time. The result of the slow moving average may be a quantifiable amount of sensitivity degradation, which may be useful as an input or variable into an EOL function, and / or a simple yes / no indicator.

[0260] For example, the processor module may calculate a slow moving average using an average of the last x hours (e.g., 24 hours), a rectangular window average, or an alpha filter with a named functional forgetting factor to evaluate the sensor sensitivity over time. In one example of an alpha filter with named functional forgetting, "alpha" may be used as follows: Parameter(n) = Parameter(n-1) x (1-α) + New information x α In the equation, α defines how much history we want to forget (how quickly we forget). In the above equation, α is the "forgetting factor". α may vary between 0 and 1, and its value describes how quickly old measurements are forgotten by the model. For values ​​of α closer to 1, the model adapts to recent measurements more quickly. For values ​​of α closer to 0, the model adapts to recent measurements more slowly. The value of α may depend on the time since the sensor was implanted. If α is 0.01, then in 1 / 0.01 (i.e., 100 time constants) samples, 63% of the previous information is forgotten. Thus, if the sampling rate is 12 samples / hour, 63% of the signal will be forgotten by 100 samples, e.g., about 8 hours. In such an example, with three time parameters or constants, which is about one day, it would mean that only 5% of the signal from the previous day would remain (i.e., 0.37 x 0.37 x 0.37 = 0.05). It is further noted that the calculation may be recursive or non-recursive.

[0261] In some embodiments, a loss of sensitivity may be indicative of the EOL. Sensitivity loss may occur toward the sensor EOL due to physiological wound healing and foreign body mechanisms surrounding the sensor or other mechanisms, including reference electrode capacitance, enzyme depletion, membrane changes, etc.

[0262] In some embodiments, sensor sensitivity may be calculated using analysis of uncalibrated sensor data (e.g., raw or filtered). In one example, a slow moving average or median of raw counts may begin to show a negative trend and the sensor may be losing sensitivity. Loss of sensitivity may be calculated by calculating the short-term (e.g., about 6-8 hours) average (or median) of the sensor output and normalizing it by the expected longer-term (48 hours) average sensor sensitivity. If the ratio of short-term sensitivity to long-term sensitivity is less than 70%, there may be a risk of the sensor losing sensitivity. Loss of sensitivity may be translated into an EOL risk factor value, e.g., a value of about 1 until the ratio is about 70%, 0.5 at 50%, <0.1 at 25%, etc.

[0263] An alternative calculation of EOL risk relative to sensitivity may use an external measure such as a glucose finger prick reading. In either case, the specific estimated sensitivity loss may be converted to an EOL risk factor value using functions described elsewhere herein.

[0264] In some embodiments, the sensor sensitivity may be calculated by comparing the sensor data (e.g., uncalibrated sensor data) to a reference blood glucose (BG). For example, the calibration algorithm adjusts the glucose estimate based on the systematic bias between the sensor and the reference BG. The EOL algorithm may use this bias, called the error in calibration or downward drift, to quantify or qualify the EOL condition. The error in calibration may be normalized to account for irregular calibration times and smoothed (e.g., moving average or nominal smoothing) to give more weight to recent data. In some embodiments, an EOL risk factor value is determined based on the resulting smoothed error in calibration. In such embodiments, the EOL risk factor value is 1 for all values ​​of error at calibration > -0.3, reduces to 0.5 for error at calibration = -0.4, and reduces to < 0.1 for error at calibration = -0.6.

[0265] Another risk factor that may be useful in determining EOL is noise based on noise analysis, such as, for example, EOL noise factors (skewness, spikyness, and rotation), duration, magnitude, and history spectral content analysis, pattern recognition, etc. In some embodiments, the processor module may be configured to evaluate the noise (e.g., amplitude, duration, and / or pattern) to determine if a predefined noise pattern indicative of EOL is present. For example, a typical sensor EOL signature may include an increase in spiking activity, which may be detected using various methods of spike detection (e.g., by calculating the average rate of negative change).

[0266] In some embodiments, the duration of the noise may be indicative of EOL. Some noise detection algorithms that may be useful are described in more detail in U.S. Pat. No. 8,260,393, which is incorporated herein by reference in its entirety. In some embodiments, the input to the calculation of the noise duration risk factor metric is the noise classification of the sensor data. For example, each raw sensor count may be classified as clean, mildly noisy, moderately noisy, or severely noisy based on the relative magnitude of the sensor and filtered sensor counts and their derivatives. This information may be used to translate the severe noise duration (e.g., the amount of sensor data in a severely noisy state) into a metric that reflects EOL risk. The assumption behind the calculation of this metric is that a sensor EOL appears as an episode when continuous noise is detected, rather than intermittent noise of several samples. Thus, the EOL algorithm may penalize longer durations more. Thus, at each sample time, the total duration of noise up to that point is used to calculate the EOL risk factor value at that point.

[0267] In some embodiments, determining whether a predefined EOL signature (noise pattern) is present includes evaluating the measured signal using a pattern recognition algorithm to determine and identify the predefined EOL signature in the sensor signal. For example, it may be determined whether a recorded noise pattern is similar to an expected noise pattern by comparing the measured sensor signal against noise pattern characteristics of the edges of the noise.

[0268] In other embodiments, determining whether a predefined noise pattern (EOL signature) is present includes comparing the measured signal against a predefined noise pattern to determine whether the recorded noise pattern resembles the predefined noise pattern. For example, the predefined noise pattern may include a series of specific negative spikes within a short time frame. The predefined noise pattern may also include an increase in spike activity over a given time frame.

[0269] In one embodiment, threshold detection on the rate of change may be used to detect up or down spikes. Spikes may be detected by a variety of methods, as would be appreciated by one of ordinary skill in the art, such as point-to-point difference and thresholding, sharpness filters, etc. For example, an algorithm or function may output a+1 for up spikes and a-1 for down spikes. Using this spike data time series, either an up spike detection algorithm, or a down spike detection algorithm, or a total spike detection (e.g., positive or negative spike time series) algorithm may be used.

[0270] In some embodiments, EOL detection using these spike detection functions may be accomplished using a negative threshold on the moving average of the spike time series (e.g., 2x negative spikes than positive), or a threshold on total spike activity that indicates a 3-4x increase in total spike activity (e.g., 3 or 4). Other forms of spike detection, such as a least squares accelerating filter, may also be used. In some embodiments, the EOL risk factor value may be determined to be 1 if the value of the spike metric is <1, may be reduced to 0.5 if the spike metric >2, may be reduced to <0.1 if the spike metric >5, and so on.

[0271] Additionally or alternatively, high frequency activity or patterns may be used for EOL detection. For example, an EOL signature pattern may show a significant increase in high frequency activity when a power spectral density (PSD) or fast Fourier transform (FFT) is performed on the sensor data. A normal glucose signal has very low frequencies (e.g., 0 and 1.8 mHz). As a result, a high pass or bandpass filter may be used to detect EOL patterns associated with high frequency activity.

[0272] In some embodiments, a slowly varying long-scale average signal may be used to normalize the data to enhance the reliability of the detection method, e.g., signal sensitivity or noise pattern, for example, using the following definitions: Long Scale = Long-term (1-2 days) running average or filtered raw glucose data Signatures = short-term (approximately 4-6 hours) filtered data (including spike detection) Normalized signal = signature / long time scale

[0273] A normalized signal threshold and duration constraint may be applied to detect the EOL signature, such that EOL may be detected when: Threshold for normalized signals > greater than a certain duration

[0274] In some embodiments, the threshold and duration may be optimized to achieve a particular sensitivity and specificity, or having a short duration constraint may be used to detect oxygen noise instead of EOL.

[0275] In some examples, EOL noise may be determined to be sensor EOL specific based on various algorithms that evaluate known EOL failure modes identifiable on the signal. It may have a large (>30% point-to-point drop) downward spike that slopes negatively for the duration of the episode, with intermittent rapid rotations or oscillations, such as multiple peaks and valleys or some derivative symptom changes. Using these features, noise discrimination can identify if the sensor is exhibiting EOL symptoms and, depending on the magnitude and duration, can calculate an EOL risk factor value from the episode, which is also called the noise factor.

[0276] Another risk factor that may be useful in determining EOL is oxygen (e.g., concentration and / or whether a certain concentration pattern is present). For example, in some embodiments, the processor module may be configured to determine whether a certain oxygen concentration and / or trend or pattern associated with the oxygen concentration is present. Any oxygen sensor useful for quantifying oxygen concentration may be useful here, and may be separate or integral to the sensor. In electrochemical sensors including potentiostats, pulsed amperometric detection may be used to determine oxygen measurements. Pulsed amperometric detection involves switching, cycling, or pulsing the voltage of a working electrode (or reference electrode) in an electrochemical system, for example, between a positive voltage (e.g., +0.6 to detect glucose) and a negative voltage (e.g., -0.6 to detect oxygen). In some embodiments, oxygen deficiency is seen at the counter electrode when insufficient oxygen is available for reduction, which affects the counter electrode in that it cannot balance the current from the working electrode. When insufficient oxygen is available at the counter electrode, the counter electrode can be driven in an electrochemical search for electrons to its most negative value, which can be ground or 0.0 V, which shifts the reference and reduces the bias voltage as explained in more detail below. In other words, a typical result of ischemia would be seen as a drop in the sensor current (e.g., lower sensitivity) as a function of glucose concentration. This occurs because, due to the reduced bias, the working electrode no longer oxidizes all of the H2O2 that reaches its surface.

[0277] In some embodiments, a non-enzymatic electrode or sensor may be used as an oxygen sensor. In an exemplary dual working electrode sensor with enzymatic and non-enzymatic working electrodes, the non-enzymatic electrode may be used as an oxygen sensor by changing the bias potential from a positive value (e.g., 600 mV-800 mV) to a negative value (e.g., negative 600 mV-800 mV). At this potential, dissolved oxygen is reduced, producing a negative current through the non-enzymatic electrode. In some embodiments, a bifunctional electrode is created by switching the bias potential on the non-enzymatic electrode between a designated positive and negative bias. When a positive bias is applied, the current may be related to the baseline, and when a negative bias is applied, the current may be related to the local oxygen concentration.

[0278] Glucose oxidase-based sensors are known to be limited by the amount of oxygen present. When the oxygen level decreases below a threshold, the enzyme electrode current drops ("oxygen starvation"), while the glucose concentration remains constant. This oxygen starvation may result in reduced accuracy, since a lower than actual glucose level may be reported. Oxygen starvation may occur at the end of the sensor's life, such as when the sensor is encapsulated in a subcutaneous environment. As a result, being able to measure oxygen allows for detection of the EOL of this encapsulation and the sensor.

[0279] In some embodiments, the presence of a predetermined oxygen concentration pattern includes a determination that includes reviewing the oxygen concentration pattern to see if the oxygen concentration is adequate. For example, an oxygen concentration pattern that indicates a reduction in oxygen availability over time may indicate EOL of the sensor.

[0280] Another risk factor that may be useful in determining EOL is glucose pattern (e.g., average, variability, meal characteristics, e.g., peak-to-peak excursions, expected versus unexpected behavior, e.g., glucose not rising as expected after a meal, etc.).

[0281] Yet another risk factor that may be useful in determining EOL is the error between the reference BG value and the corresponding calibrated sensor data (estimated glucose value, or EGV), including the direction of the error (e.g., whether BG or EGV is reading higher compared to the others) and / or utilizing flagged outliers. In some embodiments, the system may identify a difference between the reference value (e.g., BG) and the sensor value (e.g., EGV). For example, if there is a large difference between the reference value and the sensor value, there is a high probability that something is not working properly. In certain embodiments, a large difference between the reference value and the sensor value may indicate the end of the sensor life. Without wishing to be bound by any particular theory, it is believed that this is because the sensor is reading either higher or lower than it should. In some embodiments, the direction of the error, e.g., whether BG is higher or lower than EGV, is used as an EOL indicator. Yet another risk factor that may be useful in determining EOL is a measurement of the linearity (or lack thereof) of the sensor. As explained above, sensor linearity refers to the consistency of a sensor's sensitivity over a particular measurement range (eg, 40-400 mg / dL for a glucose sensor).

[0282] In some embodiments, the processor module is configured to evaluate various risk factors to provide an EOL risk factor value, which may include a simple binary (yes / no) indicator, a likelihood or probability score (e.g., a scaled relative value or a percentage), and / or an actual number (e.g., the output of various tests). The risk factor values ​​may be scaled when the weights used in the algorithm are modified.

[0283] In some embodiments, the processor module is configured to execute a probability function to determine the probability of EOL and / or the likelihood of recovery for one or more of the EOL risk factors. In some embodiments, the risk factors are mapped to a score (e.g., 0-1) based on one or more parameters. The score may be mapped by a function that translates a particular risk factor or set of risk factors into an EOL risk factor value that indicates, for example, the likelihood that the sensor will recover from the particular risk factor from EOL. Other methods of translating the risk factor output into an EOL risk factor value may be used, for example, by using one or more criteria, algorithms, functions, or equations, as will be appreciated by those skilled in the art.

[0284] In some embodiments, risk factors are fuzzified using a predefined membership function to quantify their propensity to indicate EOL. As used herein, a membership function defines the degree to which a condition is satisfied or a value belongs to a fuzzy set defined by that function. In binary logic, a number either fully satisfies a condition or not at all, whereas in fuzzy logic, a number may satisfy a condition to a certain degree described by the membership function.

[0285] As an example of a binary indicator function, the noise level is compared to a hard threshold, e.g., "5", and any value below 5 (e.g., 4.9) is treated as not containing noise, and any value above 5 (e.g., 5.1) is treated as having an unacceptable level of noise. As an example of a fuzzy membership function, an S-shape may be used to define a smooth transition in the assessment of the noise level. The inflection point of the curve is set at 5, so there is no discontinuity at that point. Thus, the same noise values ​​as above (4.9 and 5.1) are treated here very similarly. Fuzzification is the determination of the degree to which a value belongs to a fuzzy set defined by a particular membership function.

[0286] In some embodiments, each of the plurality of risk factors is partially indicative of the EOL of the sensor if each variable is determined to meet a threshold value, in some embodiments, if at least two of the plurality of risk factors are determined to meet a threshold value, the combination of the at least two risk factors is indicative of the EOL of the sensor.

[0287] The system may be configured to determine the EOL status. In one embodiment, a probability or probability analysis may be used to determine the EOL status of the sensor. The output of the risk factors becomes an input to the EOL determination process. For example, the output of the risk factors may be mapped to an EOL risk factor value, such as, for example, a value between 0 and 1, a probability or probability score, an actual value (output from the risk factor assessment(s)), etc. The EOL risk factor value then becomes an input to an EOL determination function, whereby the risk factors may be weighted or otherwise processed using probability analysis, decision matrices, various subroutines, etc. to determine an actual EOL indicator, a probability (or likelihood) of EOL, a predicted time to EOL, etc. The probability function, decision function, various subroutines, etc. may be implemented as the EOL determination function, as will be appreciated by those skilled in the art.

[0288] In one embodiment, decision fusion may be used as a function through which various inputs are processed. Decision fusion may provide a fused Bayesian likelihood estimate based on the sensitivity and specificity of the individual detector algorithms associated with each input or variable. Suitable risk factors are measured and fused together to determine whether the sensor has reached EOL. A decision may be made as to a "yes" EOL or a "no" EOL based on each individual risk factor. For example, if the sensor sensitivity has decreased by more than Δm for a certain amount of time Δt, then it is a "yes" EOL, otherwise "no", or if the sensor has severe noise (above a predefined threshold level) for more than 12 hours out of the last 24 hours, then it is a "yes" EOL, otherwise "no".

[0289] The individual decisions may be combined into a single Bayesian likelihood value, which can be used to make the best final decision regarding EOL using the sensitivity and specificity of each variable in detecting EOL. First, each decision is converted to a likelihood value using the following equation:

number

[0290] In some embodiments, the individual likelihood values ​​are multiplied together for a final fusion likelihood value that takes into account the ability of each individual variable to separate EOL from non-EOL. Thus, more sensitive and specific tests will be given more weight in the final decision. A threshold value may be empirically determined for the final fusion likelihood value to achieve the best separation of EOL and non-EOL.

[0291] In some embodiments, Linear Discriminant Analysis (LDA) may be used as the EOL decision function by taking input variables and providing an output decision.

[0292] In some embodiments, when an EOL input or variable is fuzzified using a predefined membership function, the resulting degree of membership for all data quality metrics is scaled according to predefined weights and combined to generate an indication of the overall quality of the calculated glucose value. A weight may be applicable to each metric and may indicate how indicative the metric is of EOL. These embodiments may determine the degree of severity of each data quality metric using several fuzzy logic concepts, such as membership functions and fuzzification, for example, as described above. The result of EOL detection may be a confidence indicator that determines the likelihood of EOL beyond a simple pass / fail value.

[0293] In some embodiments, the EOL state may be determined based on the likelihood that the sensor has not recovered from an event, rather than the occurrence of the event, which may be defined as a state in which the sensor is likely no longer functional or has had a long episode of noise (e.g., based on risk factor assessment(s)). The EOL indicator may also indicate the likelihood of recovery (e.g., the episode may be transient rather than terminal). In some embodiments, the system may be configured to determine the likelihood of recovery and / or monitor the sensor or sensor data over the next x hours to determine whether the sensor may recover from the EOL condition (e.g., the likelihood that the sensor will provide accurate data to the user in the next 24 hours). In some embodiments, the sensor will only be determined to be EOL if a high probability is determined that the sensor is not tracking glucose in the future (e.g., 24 hours) or has not indicated any glucose for several hours (e.g., 12 hours) (e.g., inaccuracy may be determined by comparing the EGV to a reference BG using a standard (e.g., within 20% or 20 mg / dL)).

[0294] The system may optionally be configured to monitor the risk factors (e.g., more frequently after the EOL indicator has determined the likelihood of EOL) to determine if there is a high probability that the sensor will not recover from the EOL determination. Suitable functions or algorithms for determining if the sensor will recover from EOL may be selected from those known to those skilled in the art. For example, the determination of whether the sensor will recover may be scaled from 0 to 1 based on an evaluation of one or more risk factors.

[0295] In some embodiments, the system may be configured to determine the likelihood that the sensor will recover from an EOL determination based on recent history. For example, the EOL determination function may determine an EOL condition with a high probability if there is a high probability that the sensor will not track glucose in the future or that the sensor has not detected any glucose for an extended duration. The extended duration may include a period of more than 12 hours. In some embodiments, the processor module may be configured to suspend the display of the sensor data during the verification or determination of the likelihood of recovery, and thereafter, the processor module may also be configured to re-enable the display of the sensor data if it is determined that the sensor has recovered from the EOL symptom.

[0296] In some embodiments, intermittent indications of EOL may be used to initiate advanced signal filtering techniques, such as those described in more detail in U.S. Patent No. 8,260,393, the entirety of which is incorporated herein by reference.

[0297] In some embodiments, the monitoring application may start a countdown timer that upon expiration requests or suggests insertion of a new sensor.

[0298] Further details on the issue of "end of life" sensors can be found in U.S. Patent Publication No. US-2014-0182350-A1, which is owned by the assignee of the present application and is hereby incorporated by reference in its entirety.

[0299] Returning to FIG. 18, and specifically to fault 316, a fault may be classified as a system error, for example resulting in an erroneous signal artifact. In these faults, system errors such as those related to the sensor, connections within the sensor or sensor electronics, transmitter errors, etc., may cause a variety of detrimental signal artifacts resulting in unreliable analyte readings. One subclassification includes a seal failure that may lead to a water spike. Exemplary predetermined signal criteria that may be used to test the signal for such faults include, for example, a change in the raw signal over a short period of time, such as a quick positive rise. Other signals that may be used for this fault discrimination include the temperature of the sensor, which may indicate that the patient has entered a shower or hot tub. Other exemplary signal criteria include the time of day and / or a signal from a temperature sensor, where the temperature is indicative of a clinical condition. situation In determining information Status and in clinical practice as opposed to using temperature itself, e.g. for temperature compensation. situation It should be noted that the above is used for the purpose of determining situation For use as a clinical situation Compared to the standard, clinical situation As stated, the decision was based on further clinical information to avoid ambiguity. situation For example, the temperature of a sensor (the signal being measured) may rise to a certain value or drop to a certain value (clinical situation Criteria), for example, if it rises to a certain threshold above 5°, it may indicate taking a shower (clinical situation Information). Other clinical situationWhen considered with information, pattern data, e.g., regular shower times, and / or signal data, e.g., spikes, can lead to a clear assessment of water spikes. Another subclassification includes faults related to static noise, which may be caused by friction of clothing on the sensor or electronics patch, especially during repetitive activities and in dry weather. Signal information that may indicate such faults includes analysis of the frequency content of the signal and comparison with fault criteria. Clinical situation The data may include, for example, an indication of a user's movements or motion, collected from accelerometer data on the transmitter or on an associated mobile device in data communication with the monitoring application.

[0300] Another subclassification includes failures related to motion artifacts, e.g., those caused by movement or other motion around the sensor patch area. Such failures may be particularly common when patients wear sensors on the underside of their upper arms or other similar locations, as such locations are generally more susceptible to motion affecting the sensor site. Signal criteria that may be used to distinguish such failures include analysis of the signal shape itself, including morphology, time, frequency, and distribution aspects. Clinical evaluation of such failures may be performed. situation Data may include, for example, detection of movement or activity levels, which may be collected from an accelerometer, GPS, user input, and the like.

[0301] Further subclassifications include faults related to drift. For example, drift may be either quantity m or b, where y=mx+b, which is a regression equation, where the slope M represents the sensitivity of the sensor and the intercept b represents the background or offset. Signal criteria for discriminating such faults may include measuring the potential a first time and measuring the potential a second time at the same electrode and determining if a drift of m or b has occurred. Calibration errors may also indicate such drift faults.

[0302] Yet another subclassification includes faults related to bad connections and / or broken lines. Signal criteria that may be used to distinguish such faults include detecting high frequency noise that may be characteristic of a bad connection or broken line. Fuzzy logic may also be used to determine the type of noise, particularly as it may be distinguished from other kinds of noise, e.g., during rapid changes in rate of change.

[0303] Other classification schemes will be understood as well. For example, referring to flow diagram 299 of FIG. 19, if the fault discrimination routine determines that the signal can be corrected (step 301), e.g., by prediction or other types of signal processing, the fault may be classified as a "Class 1 fault" (step 303) and appropriate signal processing may be applied thereto. If not, the routine may determine whether the display should be paused (step 305), and if so, the fault may be classified as a Class 2 fault (step 307), and the display is paused. If the display is not paused, the routine may determine whether user input would aid in correcting or determining the fault (step 309), and if so, the fault may be classified as a Class 3 fault (step 311). In this case, such user input may be prompted. Finally, the routine may determine whether the sensor, sensor electronics, monitoring device, or a combination thereof should be replaced (step 313). If so, the fault may be classified as a Class 4 fault (step 315). As noted above, for all of these classifications, there are predefined signal criteria and predefined clinical outcomes that, if met, will associate a fault with one or more classifications. situation Criteria will be defined. It will be appreciated that the above steps may be performed in a variety of orders.

[0304] As another example of a classification scheme, specific signal behavior and clinical situation A look-up table that keys off information may be used by the routine. For example, referring to FIG. 20, the signal behavior and clinical situationA look-up table is shown that keys off the information. situation Although information is shown, the limited number shown is for clarity, and it will be understood that generally numerous others could be used as well.

[0305] Similarly, a hierarchical approach to fault discrimination or classification may be applied. Referring to FIG. 21, fault discrimination or classification may involve appropriate signal analysis and clinical evaluation of each fault. situation The criteria may be generated by listing the faults in order of occurrence, for example, from the most common fault to the least common fault. The algorithm or routine then performs a criterion analysis based on the current signal analysis and clinical findings, starting with the most common fault. situation It may be determined whether the data meets the criteria. If so, the analysis may end there and a fault or fault classification may be determined from the table. If not, the analysis may continue to the next most common fault, using the current signal analysis data and clinical situation The information is again applied to the predefined criteria for a given fault. By a process of elimination, a fault or fault classification may be determined. Because faults are listed in order of their prevalence, such a hierarchical approach may lead to a fast, and possibly optimized, fault discrimination or classification.

[0306] Yet another approach to fault discrimination or classification includes the use of "decision fusion" methods. In these methods, fault discrimination, classification, or decisions may be made from multiple inputs. Decision fusion uses statistical models to combine information from multiple inputs, e.g., clinical situation The data and signal analysis data, etc. are optimally combined to generate a likelihood value that the data will be associated with a particular fault or fault classification. Such methods are particularly useful in combining disparate inputs, such as glucose change rate and number of receiver button presses in the last 20 minutes, into a single likelihood scale. Prior information regarding the sensitivity and specificity of each input in predicting an undesirable event, e.g., hypoglycemia, etc., is used to determine how much weight should be given to each input in the final output.

[0307] Further details regarding decision fusion methods are provided in U.S. Patent Publication No. 2014-0182350-A1, which is owned by the assignee of the present application and is incorporated herein by reference in its entirety.

[0308] Although the above description describes example categorizations and fault types and their discrimination, it will be understood that any method for identifying a particular type or classification of fault, whether qualitative or quantitative, may be used. Response signal processing

[0309] The various types of steps in response signal processing 328 are illustrated by diagram 326 in Figure 22. The steps taken in a given situation depend on the faults and clinical situation Dependent on the information, specific examples are provided below: Below is an exemplary, but non-exhaustive, list of such steps.

[0310] First, step 332 may be used in which filtering is adjusted or a prediction of analyte concentration is made, as described above in connection with faults that are detectable and treatable without user intervention. These steps may be performed for a number of reasons, including to compensate for time lags due to faults or due to signal processing to compensate for faults.

[0311] In the case of noise filtering, noise filtering is used to reduce the clinical situation In this manner, when an analyte such as glucose has a rapidly changing concentration, reduced filtering will allow additional data points to be taken or received to obtain a more accurate picture of the user's glycemic state. In this manner, values ​​during rapid changes may be observed more closely and quickly, thereby allowing for a more rapid response, if a response is required. Conversely, filtering may be used to reduce the clinical impact of low rates of change in analyte concentration. situationDuring the filter operation, the filtering may be increased, especially in high noise conditions. In an alternative implementation, the filtering may be enhanced by the use of fuzzy filtering, as described above and below. Status Other techniques, including regression and the use of residuals, as described below, may also be used in other implementations.

[0312] More specifically, filtering may be used to detect clinical conditions involving low analyte concentrations, such as hypoglycemic conditions. situation In one embodiment, the filtering may be reduced during a period of time. In this manner, the reduced filtering allows the data points to be processed in a more timely manner, i.e., with relatively less time lag compared to more filtering. In FIG. 23A, a situation is seen where the clinical glucose value is seen approaching a hypoglycemic state. The noise value is relatively stable. However, as the glucose value approaches or becomes a hypoglycemic state, the level of filtering is reduced to provide a more responsive system for detecting and treating hypoglycemic events as needed.

[0313] Another situation in which step 332 may be applied is where the noise is of a particular predetermined type or severity. In these cases, situation Even when not taking into account , filtering may be adjusted, for example, to increase filtering of particularly noisy signals or decrease filtering of particularly smooth ones. This situation is illustrated in FIG. 23B, where again filtering is turned down during periods of hypoglycemia. Filtering is seen to return to normal values ​​when the user is no longer in a hypoglycemic state. In addition, heavier filtering is applied when the raw signal faces greater levels of noise, for example when the user is running or jogging. In general, filtering is based on clinical factors such as the rate of change of analyte concentration. situation In addition to the information, it may be balanced based on noise type and severity.

[0314] In a related step with regard to response processing using signal processing or manipulation, the sampling rate may be altered as described above with respect to Figure 9. In particular, the sampling rate may be adjusted faster or slower to accommodate various fault conditions. For example, if a fault is indicated by a sudden upward or downward direction in the raw signal, sampling may be automatically increased to allow more data to be received, allowing a better understanding of the underlying phenomenon.

[0315] In another step in the response process, the bias potential may be changed to one that is likely to be less susceptible to certain fault types, for example, less susceptible to noise.

[0316] Returning to FIG. 22, another type of response process is for the monitoring device to enter a self-diagnostic mode (step 334). In this mode, the monitoring device may execute several routines to test itself, thereby attempting to determine the source of the fault. In some cases, the fault may be automatically corrected, and in other cases, the fault may require user intervention. The level of user intervention may vary, for example, from relatively non-intrusive steps of performing a calibration step to more drastic steps calling for replacement of the monitoring device. In the self-diagnostic mode, a sweeping step may be performed across a variety of potentials to determine the proper behavior of the sensor, for example, to detect a reference bias shift indicative of reference electrode depletion or instability due to an ischemic condition. Self-diagnostic routines may also be performed with transient signals, pulsed signals, etc., which may be scanned across various frequencies. Such a mode may also be used to test fault behavior at different potentials, since faults may behave differently at different electrode potentials or when electrode potentials are switched, as evidenced, for example, from a transient response or decay curve. The self-diagnostic routine may also test the transmitter and may further perform a resolution comparison of slow versus fast sampling techniques as presented elsewhere.

[0317] A further type of response processing is to perform a step to compensate for the fault (step 336). For example, one type of compensation is to provide a predicted, expected, or anticipated analyte concentration value for the duration of the fault. For example, if an upward spike is seen in the raw signal values, but the clinical situation However, if the patient has high or normal glucose levels, and other clinical situation If the information indicates that the time is morning, a water ingress fault, such as may be caused during a morning shower, may be inferred and the actual glucose value may be compared to the value seen before the water ingress, e.g., based on other clinical situation The prediction may be replaced based on the information and the predicted one. Redundant and average signals may be used as well. The use of prediction may be used to determine, for example, whether the user is sleeping versus eating, etc. Status It is noted that this may depend on

[0318] The predictions may also be used to compensate for time lags based on glycemic status. Specifically, as noted above, time lags can lead to adverse consequences, especially during times of high analyte concentration change rates and low total analyte concentration values. Thus, the use and display of predicted or forecast values ​​may be advantageously used during these times.

[0319] However, some clinical situation In some cases, the use of the predictive value may be prevented. For example, fault compensation by the predictive value may be safe at high glucose values, but more dangerous at low glucose values. Rather than performing a compensation step in these situations, a clinical situation may indicate that the response process should prompt for a finger prick measurement to be taken.

[0320] In addition to providing predictions, failures may be compensated for through the use of certain algorithms. For example, to compensate for a determined compression failure, a particular clinical situationHowever, if the time is nighttime and indicates that the patient is above a certain clinical glucose value, then a maximum average algorithm may be used.

[0321] Other types of response processing may be seen as well. For example, in a multi-sensor system (e.g., a multi-electrode system) described in more detail below, redundant signals may be received and used as described above, and in such a system, effects due to the local sensor environment may be isolated and thereby compensated for. In yet another type of response processing, if a large amount of noise is present on the signal, the CGM value may be muted or a very heavy or aggressive filter may be applied. The CGM may still provide reports, especially trend reports, but they will only include what is seen through the aggressive filter. For actual blood glucose values, the user may be prompted to use a finger prick.

[0322] Steps 332, 334, and 336 above may be performed without significant user input, or even without the user's knowledge that the steps are occurring. In contrast, many implementations of steps 338 and 342 below require user knowledge and possibly user intervention.

[0323] For example, another type of response action is to change the display of the monitoring device (step 338). In this way, the user can be alerted to a situation where, for example, the current glucose value is unreliable, and the user may be further prompted to input further information, for example, diet or exercise information, that may be used by the algorithm in further processing. The user may also be alerted to perform certain steps to mitigate the fault. For example, further calibration may be requested by the system during periods of high noise, especially when the user is close to critical values ​​such as hypoglycemic and hyperglycemic thresholds. Similarly, if the fault is determined as a compression fault, the user may be instructed to ease the pressure from the sensor. The user may also be quizzed about various possible causes of the determined fault, for example, "Were you lying on the sensor?" In certain implementations, the user may be prompted to perform a finger prick to determine their actual blood glucose value, especially when the CGM is detecting low glucose concentration values. The results of such measurements and questions may be fed back into the user profile and later used for personalized fault determination routines. In other words, adaptation or machine learning may be triggered so the system is alerted to fault characteristics or becomes typical for a given patient, allowing for even faster actionable alerts.

[0324] In yet another implementation, the output may be provided along with an indication of the confidence with which the algorithm calculated the analyte concentration, such as a confidence level, a "fault index" indicating the type or severity of the current fault, etc. (step 339). Display or background color may be used to conservatively indicate data confidence to the user. The output may be delayed or a cautionary notice may be placed on the output. A range of possible analyte concentration values ​​may be provided to indicate the inherent uncertainty of the data. Alerts and alarm conditions may be modified or adjusted to account for the uncertainty of the data due to faults, e.g., to more conservative values. Alternatively, certain alarms may be suspended to mitigate false or nuisance alerts. Additionally, the output may be altered based on the confidence in the data quality, e.g., via a confidence metric.

[0325] More specifically, there are various sources of inaccuracy and imprecision that may be present in an analyte monitor. These include noise and / or imprecision in the raw sensor signal, reference and calibration errors, partitioning effects, and temporal mismatches of the analyte reference, physiological foreign body responses, and transient, electrical, chemical, and biological interferences. These error sources can be quantified by examining previous sensor data and / or using error budgeting modes based on Monte Carlo methods. A combination of errors results in an inaccurate CGM glucose reading, but the degree of inaccuracy varies with a variety of conditions. For example, it is well known that current subcutaneous sensors are less accurate on the first day of use compared to later days. Some sensors tend to perform more inaccurately during fast glucose change rates (compared to steady state glucose trends). A cumulative accuracy measure does not account for differences in accuracy between individual points. Errors can be visualized and inspected in detail during post-processing steps by utilizing Bland-Altman plots or Clarke error grids, but such tools can be difficult to use and are generally retrospective in nature.

[0326] One method of determining the reliability index is by the following technique: First, raw sensor data and diagnostic information are collected. Diagnostic information may include signal noise level, local trend information, data from auxiliary sensors, etc. This information is used to calculate normal glucose values, but can also be used to evaluate the reliability of the data. In a next step, all pieces of data are evaluated according to empirical and / or adaptive criteria to determine the quality of various aspects of the signal, such as noise level, agreement with previous measurements, etc. Intermediate calculations are performed regarding sensor sensitivity and baseline, sensor operating conditions, etc. Operational characteristics are also evaluated according to a separate set of criteria to provide further information. In this step, significant information is gathered and qualified to provide the reliability technique with enough to generate a good estimate of the reliability.

[0327] Next, data quality metrics of all applicable pieces of information are included using a predefined "membership function" to quantify their propensity to cause inaccuracy in the resulting glucose value. In the art of fuzzy logic, this is called "fuzzification" or "fuzzified". The resulting degrees of membership for all data quality metrics are scaled according to predefined weights and combined to generate an index of the overall quality of the glucose value calculation. Weights are applicable to each metric and indicate how indicative the metric is of an inaccurate glucose value. For example, the relative contributions of different metrics to the overall confidence index may be as follows: recent glucose values ​​may have a quality of 19% of their propensity to cause inaccuracy, while glucose trend consistency may indicate a confidence index contribution of 34%, glucose signal stability may contribute 28% and pressure signal may contribute 19%.

[0328] The technique may then determine the degree of severity of each data quality metric. In the above terminology, the membership function defines the degree to which a condition is satisfied or a value belongs to a fuzzy set defined by that function. In traditional binary logic, a number either only satisfies a condition or does not satisfy it at all; in fuzzy logic, a number may satisfy a condition to a certain degree described by the membership function. Fuzzy logic may then be used to determine if a level of noise in a signal is a cause for concern.

[0329] FIG. 23C shows a fuzzy membership function. The S-shape defines a smooth transition in the evaluation of the noise level. The inflection point of the curve is set at 5, so there is no discontinuity at that point. Thus, similar noise values, for example 4.9 and 5.1, are treated very similarly. Fuzzification is the determination of the degree to which a value belongs to a fuzzy set defined by a particular membership function, and this figure demonstrates the results of fuzzification of values ​​4.9 and 5.1 using the membership function provided. The resulting membership values ​​of 0.45 and 0.55 reflect both noise levels relative to the threshold in the similarity between the two noise levels. The use of such fuzzy logic includes the accuracy of the system and the selectivity of the algorithm in indicating points with the possibility of being unacceptably inaccurate. The present technique may increase the number of glucose values ​​that are accurate and displayed, and decrease the number of accurate but erroneously blanked values. This may also help reduce the occurrence of false alarms and provide users with actionable alarms to aid in resolving problems that arise with data within the system.

[0330] In addition to including such in the determination of noise, a confidence level indicator can be provided with the displayed glucose value (step 339), which can be used in the calculation. Advantages include that the confidence level indicator can determine whether a glucose value is acceptable or not beyond a simple pass / fail value. This can help eliminate single point failures and allow the analyte monitoring algorithm to be more intelligent in partitioning the data.

[0331] Finally, in a display modification step 338, the output display may be modified to display the clinical, even with a common value of analyte concentration. situation It is noted that this may vary based on the information.

[0332] Referring again to FIG. 22, as yet another example of responsive processing, the monitoring device may switch between therapeutic modes (step 342). In this implementation, a failure may cause a closed loop or connected drug pump to enter a mode that is closed loop or only half closed loop. Similarly, the system may change from a state having therapeutic data utility to a state having ancillary data utility. Rather than basing pump movement on a clinical glucose value, the clinical glucose value may be provided to a user who may then control pump movement or perform a positive step of verifying or validating pump movement, where the user generally has other known data in mind. For example, such a reassignment to closed loop processing may occur upon determination of a drop recovery failure or a biofouling failure.

[0333] Similarly, the analyte monitor may be placed in a calibration mode, such as one in which the analyte monitor is calibrated against a blood glucose meter calibration or a factory calibration. Calibration schemes may also be modified to affect the interpretation or weighting of values ​​measured by finger prick or other measurements. In the related art, the system may instruct the user to provide a blood glucose finger prick value, but that calibration may be tagged as a known error and it will be used only for calibration purposes until it is determined that the error or failure has been corrected, in which case the user is signaled to provide a further calibration point and the system will thereafter ignore the previously determined failed calibration point.

[0334] Even if mode switching is not involved, one type of responsive action is to manage or control, or have a user manage or control, or instruct a user to manage or control the interactions between devices involved in diabetes management, such as, for example, meters, pumps, CGMs, etc. In this way, a user may be instructed to more closely or manually control the interactions between devices so that a failure on a device does not propagate and cause errors on downstream devices. Such responsive actions may include reducing risk thresholds for insulin amounts, adjusting default basal modes, etc.

[0335] Other types of response processing will be understood as well. For example, a "flag" may be placed on the data to indicate that the data is less reliable. For certain data known to be faulty, the weight attributed to that data may be reduced even if the data is still used. Regardless of the type, the benefit and advantage to the described response processing step is that it tends to extend the life of the sensor by allowing the sensor to continue to operate until a permanent failure is detected. This provides significant cost and convenience advantages to the user. This advantage may be in contrast to previous sensors, which generally have a strict shutdown after a predetermined number of days.

[0336] The above types of response processes generally do not necessarily indicate a clinical situation In cases where the signal can be discriminated without taking into account the situation However, as mentioned above, both the fault discrimination and the response process are clinically situation It may be based on information, i.e., signals and clinical situation may both be taken into account in discriminating between faults and in deciding how to respond to a fault (regime III).

[0337] Specific examples would include combinations of the above. In one specific example, the failure may be due to a clinical event at the time of situation , i.e., on a night-to-night basis. situation , for example, based on glycemic status. For example, if a user has high glucose levels, the fault may be compensated for by prediction.

[0338] In another example, the failure may be determined as an early wound response, i.e., a drop recovery failure. The response action may be to ignore the calibration during this time and revert to the established factory value, i.e., a priori calibration. Calibration may be requested and used as soon as it is determined that the drop recovery artifact has ended.

[0339] In yet another example, a water spike or water ingress failure may be used to determine the effectiveness of the data and clinical situation The response process for them may be determined based on, for example, ,situation When presented with evidence, a step of compensation may be taken by subtracting out the "spike" profile and optionally further performing further calibration.

[0340] In yet another example, when a short duration noise event fault is encountered and the glucose change rate is known and low, the response processing may include using a predictive algorithm to estimate the glucose value during the noise event. Further response processing may include providing an indication of the confidence level of the signal, for example numerically, or using colors such as red, yellow, and green. EXAMPLES

[0341] Various specific examples are provided herein.

[0342] FIG. 24 illustrates an exemplary occurrence of a compression failure. Compressions typically occur over shorter periods of time, for example, approximately 20 minutes up to several hours. Often, compression failures also occur when glucose levels are relatively stable, as the user is typically asleep and has not consumed food or administered an insulin bolus. Thus, the onset of a compression episode may be detected based on a drop in glucose, compared to a predicted value, or by examining the rate of change. Signal analysis may thus indicate a drop in potential or counts, What is the clinical situation? It will show the difference from the normal glucose profile (for a particular host), the difference from the predicted value for that host (based on extrapolation from the instantaneous values), etc. Other clinical situation The data may include, for example, time of day, such as nighttime. situation The data may be that the glucose levels were stable before the sudden drop.

[0343] Response processing, e.g., compensation, for a failure may be by prediction, e.g., using known or reliable values, e.g., extending the value of the signal until a period of time, e.g., 40 minutes, from when glucose is still reliable, as long as the glucose level before the episode was higher than some threshold, e.g., 100 mg / dL, and the rate of prior change was small, e.g., less than 0.5 mg / dL / min.

[0344] If these conditions are met, alternative response actions may be implemented, for example the display may be blanked and the user may be woken by an alarm, since if the conditions are not met the user may be on the verge of becoming hypoglycemic.

[0345] Other responses to compression will be appreciated as well. For example, the user may be prompted to change body position to relieve the compression. However, clinical situation However, it is night time. situation When indicated, such prompting may be suspended and response processing may be limited to operating in a closed loop mode (unless user intervention or an alert is required). situation In some cases, the response action may be to do nothing, especially if the response action adds little value. For example, an alert to a user when a nighttime compression episode is detected may not provide further insight to the user if their current reading is 180 mg / dL (indicating true glucose is above that value). This reading indicates an elevated glucose value, and therefore the clinical response would be the same, e.g., compensation. In other words, just because a fault is detected does not mean that the response action is always a positive action with respect to that fault, and the response action may mean not performing an action or step.

[0346] As another example, the user's actual glucose level may be 87 mg / dL and the CGM may be 77 mg / dL. situationBased on this, a compression event may be appropriately detected. The fault discrimination algorithm may quantify the fault as a -25 mg / dL bias, but in this example, the true compression bias is actually 10 mg / dL. If the response action were to compensate for the detected fault, 25 mg / dL would be added to the CGM reading, resulting in a reading of 102 mg / dL. The alternative would be to leave the negative bias in the CGM. In this case, leaving the negative bias in the CGM is a safer or more conservative approach, and thus the fault discrimination algorithm may choose not to perform a compensation step in this situation based on the glycemic state of the user (e.g., below a predetermined glucose threshold).

[0347] With respect to compression, generally, multiple inputs feed into the definitive determination of a particular failure. As explained above with respect to sensor end of life ("EOL") determination, methodologies may be used to definitively determine such failure or to determine such failure with a desired degree of probability. These multiple inputs may constitute risk factors, which may be evaluated, for example, periodically or intermittently, or otherwise, along with receipt of sensor data. Risk factors may be determined, averaged, or trended repeatedly over time, and the results may be used for subsequent processing.

[0348] Suitable risk factors for compression may include sensor readings, sensor variability, time of day, pattern data, and various others. In some embodiments, the processor module is configured to evaluate the various risk factors to provide a compression risk factor value, which may include a simple binary (yes / no) indicator, a likelihood or probability score (e.g., relatively scaled or a percentage), and / or an actual number (e.g., output of various tests). As with the EOL risk factors, the processor module may be configured to perform a probability function to determine the probability of compression and / or the likelihood of recovery for one or more of the compression risk factors. In some embodiments, the risk factors may be mapped to a score (e.g., 0-1) based on one or more parameters, which may then be mapped by a function that translates a particular risk factor or set of risk factors, for example, into a compression risk factor value that indicates the likelihood that the sensor will recover from the particular risk factor from compression. Other methods of translating the risk factor output may be used, for example, by using one or more criteria, algorithms, functions, or equations, as will be appreciated by those skilled in the art. In other implementations, fuzzy logic may be used in determining the probability of compression failure, as well as decision fusion, both of which are described elsewhere. Look-up tables, expert rules, neural nets, and the like may also be used in the determination depending on the implementation.

[0349] In the above examples, multiple alternatives were seen for the response process. For a particular fault, there may be only one alternative. For example, if the fault is drop recovery or oxygen noise, the display may display the other alternatives. Status Regardless of the information or specific data characteristics, it may be left blank.

[0350] FIG. 25 depicts the early wound response, e.g., a case of "downward recovery" failure. Such failure can manifest itself in a number of ways, such as low glucose levels, and can have clinical consequences. situation This can be difficult to determine by simply reviewing the uncalibrated data without considering the

[0351] As discussed above, drop recovery failure is a characteristic of recently implanted sensors that is due to a physiological initial wound response. Thus, in one exemplary implementation, and with reference to flow diagram 344 of FIG. 26, implantation time may be used as a signal analysis criterion in fault determination. For example, if signal analysis indicates that implantation time is, for example, between 2 and 6 hours (step 346), and clinical situation If it is determined that a hypoglycemic state has occurred (step 348), a potential fault may be detected (step 351).

[0352] To determine a failure, one type of response action is to enter a frequent sampling mode, e.g., every 30 seconds, to ascertain the rate of change of the glucose level (step 352). If the rate of change is characteristic of a drop recovery failure (step 354), e.g., if the indication is negative and the magnitude is greater than a threshold, various types of response actions may be taken. For example, the user may be prompted to measure their blood glucose level manually, e.g., by finger prick (step 356). Alternatively, if a suitable sensor is provided, another different chemical species may be measured, e.g., NO, which may be released by inflammatory cells, which may be thought to be caused by a drop recovery failure, since inflammatory cells may consume glucose. In some cases, both steps may be performed. In this way, the failure may be detected as a data and / or clinical indication. situation will be judged on both.

[0353] If the failure is determined to be due to drop recovery, then a response action may occur. The response action may take several forms, including compensating for the failure by relying on the patient's most recent blood glucose input to predict or forecast (step 362). Optionally, if the basal signal is not representative of glucose concentration, and a predictive algorithm is not available, for example due to minimal data, the compensation step may be skipped and no action may be taken (step 364).

[0354] Other types of response processing will be appreciated as well. For example, if blood glucose was previously measured within the hyperglycemic range, the display screen of the monitoring device may be blanked (step 366) so as not to convey a reading that may be erroneous. Although the patient is starting out in a hyperglycemic state, the rate of change indicates a decrease in glucose values, which may not immediately present a dangerous situation. Of course, the length of time the display screen is blanked will depend on clinical factors, such as, for example, the level of hyperglycemia, the magnitude of the rate of negative change, etc. situation It may vary depending on.

[0355] On the other hand, if the patient begins to become euglycemic or hypoglycemic, the patient may be alerted (step 368). In this case, an alarm may be issued to warn the patient of a potential impending hypoglycemic condition only if the rate of change reflects the actual glucose value and is not caused by a fall recovery failure. As with a hyperglycemic condition, the actual steps taken are clinically situation may depend on other aspects of the.

[0356] The specific implementation is situation The patient may be prompted to provide further information regarding the amount of insulin consumed (step 372). For example, the patient may be asked whether they have eaten a meal within the last few hours and / or whether they have taken insulin recently. If no response is subsequently received from the patient, an alarm may be sounded.

[0357] Other variations include providing modifiers to the displayed glucose values, such as a range of glucose values, to indicate potential clinical values ​​due to uncertainties caused by faults. Historical data may also be employed to calculate ranges or inform other response processing, for example based on time of day or other clinical circumstances.

[0358] Like compression and EOL determinations, drop recovery failures also generally involve multiple inputs feeding into their determination. As noted above, methodologies may be used that involve consideration of multiple risk factors that are periodically or intermittently evaluated to determine such failures definitively or to determine such failures with a desired degree of probability.

[0359] Suitable risk factors for drop recovery may include sensor readings, implantation time, pattern data, and various others. In some embodiments, the processor module is configured to evaluate the various risk factors to provide a drop recovery risk factor value, which may include a simple binary (yes / no) indicator, a likelihood or probability score (e.g., relatively scaled or a percentage), and / or an actual number (e.g., output of various tests). As with the EOL risk factor, a probability function may be executed by the processor module to determine the probability of drop recovery and / or the likelihood of recovery for one or more of the compression risk factors. Other methods of interpreting the risk factor output may be used, for example, by using one or more criteria, algorithms, functions, or equations, as will be appreciated by those skilled in the art. In other implementations, fuzzy logic may be used in determining the probability of drop recovery failure, as well as decision fusion, both of which are described elsewhere. Look-up tables, expert rules, neural nets, and the like may also be used in the determination according to the implementation.

[0360] FIG. 27 illustrates the fault discrimination of a “shower spike” based on signal analysis including temperature data and time reference, which when combined, provide a clinical signal indicating that the patient is taking a shower. situation Specifically, in the signal analysis step, the sensor signal output 374 shows an increase in signal at point 375, which correlates with an increase in temperature at point 377 in the temperature plot 376. situationIn analyzing the information, the time of the spike is matched to the user's shower time, either in comparison to other users or based on pattern data for this particular user. situation Information may be seen, such as a drop in temperature at portions 382 and 384 before the spike at point 377, which may indicate that the patient has gotten out of bed.

[0361] To further identify this defect, signal analysis and clinical situation Based on this, a test may be performed to identify potential faults and look for "shorted" electrodes. For example, a self-diagnostic mode may be entered and bias potentials may be changed. The system may then look for an absence of response (a short may appear to be generally unresponsive to various given stimuli).

[0362] For example, in one implementation, systems and methods according to the present principles include signal analysis and clinical situation A method of determining faults may be provided that includes identifying potential faults based on data about the device. Other steps may include going into a self-diagnostic mode and performing various tasks, such as varying bias potentials and checking the response. For example, an absence of response may indicate a "short circuit" since it may appear to be generally unresponsive to various given stimuli.

[0363] Similar to EOL, compression, and drop recovery failures, suitable multiple risk factor inputs may be used in determining shower spike failure using statistical and probabilistic models including fuzzy logic and decision fusion analysis, and using look-up tables and the like in the determination.

[0364] FIG. 28 illustrates another example of fault discrimination for a compression fault. Trace 386 corresponds to temperature, and traces 388 and 392 correspond to unscaled paired sensor traces. Signal analysis characteristics based on the traces include low levels of high frequency noise 394, as well as abrupt shifts 396 in the sensor signal at the beginning and end of the compression event. These shifts are illustrated in sensor trace 388. Paired sensor trace 392 represents a similar type of sensor worn on the opposite side of the patient, and therefore not undergoing compression. Trace 392 therefore shows a reliable glucose signal compared to trace 388, which has significant artifacts. Clinical situation The information may indicate a user who is asleep, which may be determined by time of day compared to a particular standard of sleep. situation Information includes high temperature and lack of gauge readings (not shown) indicating the patient has not recently performed a finger prick.

[0365] In one implementation of a method for fault discrimination of compression failure, a signal is received and analyzed for various aspects. For example, the received signal may be analyzed for low levels of high frequency noise. As another example, the received signal may be analyzed for shifts in the sensor signal greater than a predetermined threshold at the beginning and end of a significant or sustained decrease in the sensor signal, such as one characterized by an abrupt drop in signal value, followed by a sustained reduced value for a period of time, followed by an abrupt increase in signal value. The method of fault discrimination may be performed in a clinical setting. situation Clinical data situation Analyze against the standard and clinical situation For example, determining the clinical situation The data may include the elevated temperature compared to what would be expected in the absence of a malfunction, the time of day, e.g., whether it is a time expected for sleep, and other such data. If two sensors are worn, the received signal may be analyzed for situations where one sensor experiences a decrease in signal value and the other does not. situationAccording to the information, the system and method may determine that a compression failure has occurred.

[0366] 29A and 29B illustrate another example of fault discrimination involving a fault for water intrusion. These figures represent the same patient on different days. Traces 402 and 402' show temperature readings, and traces 404 and 404' show sensor readings.

[0367] clinical situation The information may be, for example, a clinical indication of the temperature to be expected. situation By comparing with the norm, time (or another scale, e.g., timeofweek) can be used to assess clinical outcomes. situation This may be determined in a number of ways, such as by comparing to a baseline, and thus determining a behavioral pattern, such as, for example, showering. Such a pattern appears to be quite consistent during the weekdays, and thus the clinical significance in this example. situation The information indicates that the user is taking a shower. More specifically, situation The criteria show regular times when the signal undergoes a sudden increase (see failure area described) followed by a decay over a period of several hours. situation Clinical information to determine situation Other data that may be compared to the baseline include temperature, e.g., a decrease in temperature likely caused by an occupant getting out of bed, and noise, e.g., noise levels in the previous data, such as may be caused by water intrusion caused by a shower.

[0368] In this case, temperature compensation would be insufficient to compensate for shower spikes. Specifically, some previous attempts at implementing temperature compensation have used in vivo temperature measurements at the sensor site and ex vivo on the transmitter, but in the case of water ingress, the problem is not caused by an inaccurate or improper temperature reading at the sensor, but rather by a short circuit 29 to water ingress in the transmitter. In fact, the temperature sensor 27 in FIG. 30 is the opposite of that in FIG. 28, both of which are caused by water ingress. Thus, further signal analysis and clinical validation, for example by adding a constant value, may be required. situation Providing a level of temperature compensation without also providing a level of temperature compensation can lead to significant errors. Therefore, the temperatures used as inputs herein are used to approximate the temperature data to a given clinical situation Compared to the standard, clinical situation information (e.g., whether the user is sleeping or taking a shower) The situation It should be understood as applying in the

[0369] More specifically, water ingress failure causes moisture to enter the seal or into cracks in the insulator, in either case driving additional signals that are not related to glucose, but still emanate from an electrochemical mechanism (but different from that of glucose). The signal from the non-glucose-related mechanism is a function of various factors, including the additional exposed surface area of ​​the sensor, reactions with non-glucose analytes, and exposure of the different working electrodes to moisture. During amperometric detection, the signals from both sources appear similar and are difficult to distinguish. However, in some cases, a unique signature may be obtained from the actual electrochemical reaction with the analyte. The electrochemical signal resulting from the exposure of the additional surface area is also different from that of other interferents, such as acetaminophen.

[0370] Thus, in one implementation, electrochemical means are utilized using multiple potentials or AC or pulse voltammetry to obtain a quantitative measurement of surface area, which can therefore provide another indication of water intrusion.

[0371] For example, AC voltammetry may be performed intermittently to share the function of the working electrode, e.g., glucose detection may occur for 4 minutes of a 5 minute period, while for the last minute an oscillating potential may be applied to the electrode to see if any of the signals are from non-glucose related signals or from those related (or not) to hydrogen peroxide or other potentially interfering analytes. It is not essential to distinguish or separate the interfering analytes from each other, but distinguishing the moisture intrusion signal from other signals is generally all that is needed for this type of fault discrimination. The above signal distribution is merely an example. Generally, the method uses a portion of the measurement period for error checking to see if there is any unexpected electroactive surface area exposure. Other techniques that may be used for such include oscillating potential, impedance measurement, pulsed amperometric detection, etc.

[0372] For example, in one implementation, systems and methods according to the present principles may be used to measure or determine water ingress failure by using the following steps: In a first step, a signal is received, the signal being related to an electrochemical mechanism caused by the analyte and the sensor. In a next step, a quantitative measurement of the surface area is determined, e.g., from the signal or from alternative electrochemical means, e.g., multiple potentials, AC voltammetry, pulse voltammetry, etc. The quantitative measurement of the surface area is then used to determine if water ingress has occurred, e.g., to see if a portion of the sensor's surface area is being adversely taken up by moisture. Optionally, further steps may be performed, such as detecting a signature detected from the signal, a signature associated with a level of interferents and / or surface area of ​​the sensor.

[0373] FIG. 30 illustrates an example of a signal in which an "end of life noise" failure may be identified on sensors worn simultaneously on a patient. The units of sensor output when scaled are current [nanoamps] / clinical glucose value [mg / dL] or nanoA / mg / dL. Traces 408 and 412 are shown, where trace 408 is a sensor trace illustrating a failure. Trace 412 illustrates a reliable signal, while trace 408 is rendered unreliable due to the failure. Since noise may be site specific, for example affected by a particular microenvironmental defect, it is not surprising that a site specific failure occurs in one location on a patient but not another location on a patient worn for the same period of time.

[0374] Certain characteristic shapes 414 on the trace 408 are seen that may be used to discriminate between failures, including a steep downward spike at the beginning of a noise episode, high frequency noise present throughout the episode, and a positive overshoot at the end of the noise episode. Signal analysis may also indicate other potential signal criteria, including the tendency of noise episodes to be close in time to similar episodes, and a tendency for episodes to become more frequent as time passes and the sensor endures more wear. Another signal-related predetermined criterion that may be used to discriminate between this type of failure is that the failure generally coincides with a gradual decrease in sensitivity. One type of clinical indication for this failure is situation The information is that the breakdown occurs more frequently when glucose is elevated. situationThe information criterion is that there is an increased probability of failure occurring when there is a high average sensor current during the session or a high integrated current from the beginning of the session. Other exemplary parameters that may be used for end of life detection include the amplitude and / or variability of sensitivity, typically indicating, for example, a 5%, 10%, 20% decrease over the last 6, 8, 10, 12, 24 hours, as well as noise patterns, spectral content, days since implantation, oxygen concentration, glucose value, glucose value error in calibration, etc.

[0375] In certain implementations of systems and methods according to the present principles, particularly as applied to determining end of life noise faults, a step may include receiving a signal trace, analyzing the signal trace for certain characteristic shapes. For example, the signal trace may be analyzed to detect a steep downward spike at the beginning of a noise episode, high frequency noise present during the noise episode, a positive overshoot at the end of the noise episode. If such is seen, at least an initial determination or discrimination of end of life noise may be made. Other aspects may contribute to such a determination or discrimination. For example, if multiple such signal traces are seen, particularly over a predefined time window where each signal trace includes the above aspects, the likelihood or probability of end of life noise may be increased and the confidence level of such a determination or discrimination may be raised. If such episodes become more frequent over time, again the likelihood or probability of end of life noise may be increased and the confidence level of such a determination or discrimination may be raised. Similarly, if further data is detected regarding the sensitivity of the sensor, and if the sensitivity appears to be decreasing over time, particularly in a progressive manner, then again the likelihood or probability of end of life noise may be increased and the confidence level of such a determination or discrimination may be raised.

[0376] clinical situationThe information may also increase the likelihood or probability of end of life noise and therefore the confidence level of such a determination or discrimination. For example, if glucose levels have been elevated for an extended period of time, that may tend to increase the likelihood or confidence of a determination of end of life failure. Other types of clinical information that serve as inputs to the determination or discrimination of end of life failure may also be used. situation The information may include time since implantation, oxygen concentration, glucose value, error in calibration, etc.

[0377] 31 illustrates another example of a drop recovery failure. In this example, a user wore two sensors simultaneously, both of which exhibited some artifacts, but one of them exhibited a more severe failure caused by a drop recovery failure. Specifically, trace 416 is a sensor trace showing a drop recovery failure, while trace 418 shows only noise-related artifacts. Circle 422 represents the instrument value.

[0378] One signal characteristic indicative of a drop recovery failure includes a signal drop consistent with the instrument value 422, as seen at point 424. Another potential signal characteristic is an increase in noise in a particular frequency range (as seen in both traces 416 and 418), or an increase in noise that is uncorrelated with the paired redundant sensor (note the lack of correlation between 416 and 418 during drop recovery). The level of noise in a particular frequency range may be determined by appropriate frequency transformation. A further potential signal characteristic consistent with a drop recovery failure is a downward deviation of the signal from the redundant sensor data, which is also shown in FIG. 31 by the deviation between trace 418 and trace 416. One type of clinical feature that may be used for fault discrimination or response processing is situation Information includes time since implant, as the onset of this type of failure generally occurs several hours after sensor insertion.

[0379] Thus, in one implementation, the system and method according to the present principles relate to a method for determining a drop recovery failure. The first step in the exemplary method is receiving a signal and analyzing the received signal. Various characteristics may be used in the analysis to determine if the received signal is consistent with a drop recovery failure. For example, a determination or determination of a drop recovery failure may be made if the received signal decreases at the same time that the blood glucose meter value does not decrease. Alternatively, if there is an increase in noise within a particular frequency, that may also lead to a determination or determination of a drop recovery failure. If the sensors are paired, i.e., the user is wearing two sensors, noise present in one but not the other, or a signal decrease present in one but not the other, may further lead to a determination or determination of a drop recovery failure. The method for determining or determining a drop recovery failure may include a clinical situation Information may also be used, e.g., clinical situation The data may be received by a system or method according to current principles, in which case the clinical situation The data constitutes the time since implantation, which may be compared against a criteria, for example, the criteria including whether the time since implantation is before or after a predetermined threshold amount of time since implantation, such as, for example, 12 hours. If the signal information exhibits a decrease compared to the blood glucose meter value, or exhibits an increase in noise within a particular frequency range, or meets one or more of the other criteria described above, and the time since implantation is less than the predetermined threshold, then a determination or discrimination may indicate the occurrence of a drop recovery fault.

[0380] 32 illustrates another type of failure: lag. Results for two different sensors implanted simultaneously in a host are shown and illustrated by traces 428 and 432 along with corresponding peak detection curves 432 and 434, respectively. One sensor (traces 426 and 434) experienced a lag of several minutes compared to the others. Such lag errors can be attributed to the falling glucose after a meal. StatusIn this case, the time lag may be important, with the risk that a low glucose alarm may be delayed or missed. This phenomenon may be a permanent property of the sensor site, or it may be transient, depending on the local blood perfusion. In either case, a response process may be performed based on the glucose concentration and / or the rate of change, in particular the use of prediction or forecast values. In this way, the effect of the time lag may be mitigated, so that it does not delay the user in responding to the hypoglycemic event.

[0381] In certain implementations of systems and methods according to the present principles, it may be used for the discrimination of such lag failures. The first step is to receive a signal from a monitor, such as, for example, a CGM or other analyte monitor. The next step is to analyze the signal for the presence of lag. The analysis for lag may include analyzing the received data signal itself, or analyzing the received signal together with another received signal, such as, for example, one from a paired glucose sensor. After a lag failure is determined or discriminated, a response process may be implemented. For example, for lags greater than a predetermined threshold, or indeed for any lag, a forecast or predicted value may be displayed to the user instead of the delayed value to provide the user with a more accurate indication of their current situation.

[0382] Figures 33A-33D illustrate another example of compression failure, in this case manifested in a pediatric patient. The raw signal is shown and the counts are measured, with roughly 200,000 counts corresponding to a clinical glucose value of 100 mg / dL. Figures 33A and 33C show multiple days of data for two different patients, and Figures 33B and 33D illustrate a more detailed view of a particular compression episode in each of Figures 33A and 33C, respectively. As can be seen, the sensor signal drops completely to baseline values ​​during these compression failures. In these cases, the sensor was attached to the lower back / buttocks. Figures 33A and 33B further illustrate the rebound effect after compression, where after relaxation of compression, the signal has passed the equilibrium or previous raw signal value.

[0383] In another implementation, a fault may be detected by identifying a signal shape that has a known signal shape associated with the fault. Specifically, when the signal under evaluation consists of one or more predictable shapes, it is beneficial to establish expectations of normal or abnormal signal characteristics to detect artifacts and aberrations. Such expectations, or "templates," may be compared with each newly arriving signal to assess its correlation with or deviation from the template. This comparison may be performed to detect failure modes that have, for example, a shape or characteristic effect on the sensor signal, such as a damped or unstable response, to distinguish known failure modes and to assess severity.

[0384] In such systems, one way to achieve accurate blood glucose readings is to superimpose a blood draw with pressure and a glucose sensor signal consistent with the patient's blood access and typical enzyme sensor response. The goal of this process is to distinguish between fault conditions that the monitoring algorithm can reliably mitigate and those that will produce glucose measurements that do not meet the required accuracy. When the algorithm cannot display accurate results, the decision logic may partition failure modes with known mechanisms and characteristics into actionable alarms. These alarms may identify failures that require user action to correct or that indicate complete sensor failure.

[0385] Referring to flow diagram 431 of Figure 34A, two steps are seen in the process. The first step is the generation of a template for the signal in the measurement period (step 433). The second step is the matching of the template to the received data (step 435). The steps will now be described in more detail.

[0386] In generating templates, one approach is to use singular value decomposition or related factor analysis to determine the source of variation in the training set. To do this, the training set is compiled using a representative data set for reliable operation that meets accuracy requirements or targets known failure modes. Such a set is arranged in an m×n matrix M of sensor signal versus time, where m is the number of samples stored as a row vector, and n is the number of time points in each sample. The signal can be from an electrochemical sensor, such as, for example, a steady-state or transient signal analysis from a transient measurement system, which may benefit from the particular high-speed sampling of data, as described in more detail elsewhere herein.

[0387] Singular value decomposition is, for example, the SVD function of Matlab (registered trademark): M=USV i The signal template is implemented using available subroutines such as . V is an n x n matrix containing the singular vectors in order of decreasing contribution to the overall signal. In other words, the first column of V will contain the most salient features in the training set, the second column will contain the next most salient features, etc. In one approach, the most salient features, i.e., the first singular vector, are converted into a signal template. In another approach, the signal template is a linear combination of the singular vectors. Other approaches will be understood as well.

[0388] Yet another approach uses a physical or mathematical model of the system to generate a template for the sensor response. An example would be a mathematical model for a sensor response based on compression artifacts. Such a model may be used to generate a dynamic response seen in typical compression artifacts. Another example would be a mathematical model for a sensor response based on diffusion rates. Such a model may be used to generate a dynamic response for a reliable sensor or for a sensor slowed down by bioattachment or encapsulation. Other mathematical models may be generated for other such signals, e.g., from step or periodic voltage cycles (AC or DC), intermittent exposure to the sample, etc.

[0389] In the second step, matching the template to the data, the sensor signal can be sensor response versus time or may be pre-processed to filter out electronic noise or other data acquisition artifacts. Each sample of the incoming data is then projected onto one or more templates to determine its correlation with (or deviation from) the template and detect specific features and failure modes. The result of the projection yields the contribution of that particular template shape to the overall shape of the sensor signal.

[0390] In another implementation of the second step of matching the template to the data, the expected sensor response may be shifted in time to compensate for allowable manufacturing, operational, and physiological variations that change fluid volume. For example, the time shift may result from changes in catheter volume or sensor position that affect dead volume. The shifted sensor response may then be matched to the template.

[0391] Yet another approach similarly allows the sensor response to be stretched or compressed in time to compensate for acceptable manufacturing, operational, and physiological variations that may occur. For example, such signal variations may result from peristaltic pump efficiency or sensor response changes with temperature. An example of the method of FIG. 35A is described below.

[0392] Referring to Figure 34B, a signal is shown that generally illustrates a compression artifact. A typical embodiment includes a "regular" signal portion 411 before compression, a "regular" signal portion 413 after compression, a steep downward slope 415, a steep upward slope 439, and a flat area 437 between the slopes. The downward slope generally indicates the onset of compression, and the upward slope indicates the relief of compression. The time between the two may vary, but is generally a few minutes to tens of minutes.

[0393] Compression artifacts generally have the shape illustrated, and therefore a template may be generated accordingly, which may be used as a reference against which the incoming signal is evaluated. For example, with reference to FIG. 34C, an exemplary signal template 449 is illustrated having horizontal bands 411' and 413', a downward sloping band 415', and an upward sloping band 439'. A flat sub-band 437' is also illustrated, and it will be understood that the length (in time) of this band may vary depending on how long the compression occurs. As long as the incoming signal is within the bands, i.e. the data is clinically relevant. situation As long as the criteria are met, compression failure, i.e. clinical situation For example, several slopes of downward and upward signal waveforms, 441 and 443, respectively, are illustrated, which correspond to clinically relevant information set by bands. situation However, the signal slope 447 is illustrated below and the signal slope 447' is illustrated above, which do not fit within the band and therefore are not clinically relevant for compression failure. situation would not meet the criteria.

[0394] It will be appreciated that numerous variations are possible. For example, the positions of bands 411' and 413' within template 449 may vary significantly in their vertical (signal value) positions. The widths of bands 411' and 413' may be greater than or the same as the width of band 437'. Other variations, and ways of providing templates without the use of such bands, e.g. other mathematical models such as correlation analysis against a curve (template), etc. will be appreciated as well.

[0395] To determine the template, an SVD routine may be run over a large training set of individual compression artifacts collected from a variety of different sensor hosts that have experienced compression artifacts. The known compression artifact signatures provide a tool against which each possible compression artifact may be evaluated. Multiple such templates may be created, and each time a compression artifact is detected, the measured signal may be projected onto each of the templates to obtain their contribution to the overall shape. Although compression artifacts are illustrated herein, the same principles of creating and using templates for comparison against any known signature (e.g., EOL, drop recovery, transient signals obtained during a self-diagnostic period, or any other waveform generated by the sensor by any known methodology) may be applicable, as recognized by those skilled in the art.

[0396] 35A-35C illustrate some examples of signals that may be compared against templates. FIG. 35A illustrates an exemplary compression artifact. FIG. 35B illustrates an artifact that may be compared against a template. In this plot, the overall shape is somewhat consistent with a compression artifact, but not with additional noise, which may still be quantified as yes / no and / or determined in terms of a confidence factor. FIG. 35C illustrates some other types of signal artifacts other than compression. The signal is not well described by either a typical compression waveform or normal glucose behavior, but the signal correlates more closely with normal glucose behavior. Thus, the compression artifact may not be discerned in the particular scenario of FIG. 35C, but a prompt may be sent to the user to provide more information and / or other processing may be applied.

[0397] Thus, the use of templates generated from the training set allows the algorithm to detect specific failure modes that manifest as specific waveforms. Given a sufficiently large training data set, more templates can be easily generated for other failure modes that require discrimination. The results of template matching can be used in clinical trials to discriminate failure modes and thus faults. situation It may be combined with information.

[0398] Generally, templates are created empirically across large data sets derived from different patients, although in certain implementations other methods of establishing templates may be used, including the use of templates established from a single patient.

[0399] Systems and methods have been disclosed for dynamically and iteratively providing fault discrimination and response processing. Various methods have been disclosed for performing fault discrimination and for subsequent processing, including corrective action.

[0400] Variations will be apparent to those skilled in the art given this teaching. situation Data is clinical situation Clinical to develop information situation Specific clinical outcomes compared against a standard situation As explained above, the above clinical situation It will be understood that the above is an exemplary and not exhaustive list. For example, the sensor insertion site may also be determined based on clinical situation Certain sensor insertion sites may be more susceptible to failures such as drop recovery, water ingress, and compression, and therefore may be more susceptible to failures such as drop recovery, water ingress, and compression. Situation By taking into account, the discrimination or determination of a fault may be made more accurately. In another variation, while in vivo sensors and measurements are generally described above, in some implementations, ex vivo sensors and measurements may also be used. In yet another variation, while continuous measurements are generally described above, certain implementations may utilize periodic or intermittent measurements. Other clinical situation Variations and variations of the above will be understood as well.

[0401] As yet another example of a variant, techniques have been described for determining and responding to compression faults. Compression faults can cause CGM devices to be inaccurate when tissue around the sensor is compressed. It is believed that tissue compression causes reduced perfusion of glucose to the sensor and / or oxygen around the sensor (and results in a compressed signal). The effect typically occurs over a short period of time, such as 5 or 20 minutes to 60 minutes, up to several hours. Accuracy returns when the patient adjusts position and no longer compresses the sensor. In addition to (or instead of) other methods for detecting compression artifacts, a compression sensor can be placed in the transmitter or disposable sensor. The compression sensor can directly indicate if tissue is compressed in that area. When activated, several actions can be taken, for example, an alarm can be sounded, data can be blanked by the receiver, or the patient may be alerted via a light shock, vibration, etc. If the system is connected to a pump, a specific action or no action can be taken, such as suspending insulin. Additionally, the sensor may be designed to cause discomfort to the patient if the patient presses against the sensor, discouraging them from lying on the sensor.

[0402] For example, referring to FIG. 36, an integrated force sensor 436 and transmitter are illustrated. In this implementation, a small pressure transducer is placed directly under the sensor transmitter. Specifically, the transmitter 436 is formed by combining two parts: a rigid base 444 carrying a printed circuit board and a rigid cover 437. A compressible gasket 454 is provided that allows the cover to move to and from the base in response to an externally applied force. A pressure sensing element 442 is provided on the printed circuit board, which is coupled to the cover 437, for example, by a contact pin 438. When a force is applied to the transmitter, the cover is depressed and the pressure can be sensed using the pressure sensor. In an alternative embodiment, the compressible gasket is omitted, in which case the cover may comprise a thin flexible area that flexes when a force is applied.

[0403] Pressure transducer measurements may be used to aid in the identification of false hypoglycemic values ​​associated with compression applied directly to the sensor and transmitter pads. Exemplary pressure transducers include miniature piezoelectric pressure transducers, strain gauges, springs, those using capacitance measurements, and the like. Falsely low blood glucose readings associated with compression will not result in alarms or other unwanted therapeutic actions. Special algorithms may combine pressure transducer data with the previous 60-90 minutes of glucose trend data to further aid in discerning actual hypoglycemic events from false readings induced by compression at the transmitter and sensor sites. In this manner, the phenomenon of "alarm fatigue" is minimized, increasing the likelihood that the user will respond to other alarms.

[0404] In addition, the CGM system may be used as part of an automatic insulin infusion system or artificial pancreas system. Such systems will generally include automatic interruption of insulin infusion in response to actual or impending hypoglycemia being detected. As noted above, if the detection of hypoglycemia is correct, interruption of the insulin pump is justified. However, if the hypoglycemia detected by the sensor is false, there is a risk that automatic pump interruption could lead to severe hyperglycemia and possibly diabetic ketoacidosis. Using the system of FIG. 36, an independent method may be used to determine if a sensor reading is anomalous by using data from the real-time pressure transducer, significantly improving the accuracy of the reading and therefore the treatment to the patient.

[0405] In another variation, although various types of sophisticated response processing techniques have been disclosed, another method for handling a failure or malfunction is to notify the patient of the problem and to configure the system to go into a fail-safe mode or shut off the sensor.

[0406] In yet another variation, in the above implementations where a prediction or forecast is proposed, pattern analysis, clinical situationAny method of forecasting or predicting using historical and current data values ​​may be applied, including methods involving the use of linear regression, etc. However, simple linear regression may also be applied. In this example, a certain amount of data is used for linear regression, which is used to calculate the latest values ​​using the regression-determined line. For example, data may be taken every 30 seconds over a 5 minute period, which may be used for regression analysis. This technique may also help smooth the data and remove time lags. The residuals around the line may be used as an estimate of the noise level. In an improved technique, limits may be placed on the slope of the calculated line to reflect appropriate physiological limits. Limits may also be placed on how much the slope of the line may change during each 5 minute interval. In a particular implementation, a linear regression is taken over 10 samples, and a predicted value is calculated for the end points of the line, significantly reducing the amount of noise and filtering time delays.

[0407] This idea is illustrated in Figure 37, where points 456 delineate the beginning and end of different 5 minute sampling periods. Ten samples are taken in each of the 5 minute periods, corresponding to 30 second intervals. Using linear regression, the estimated glucose values ​​for the end points of the lines are calculated in a particularly fast manner. This provides a faster or more adaptive way to perform response processing.

[0408] The connections between the elements shown in the figures illustrate exemplary communication paths. Additional communication paths may be included, either direct or via intermediaries, to further facilitate the exchange of information between the elements. The communication paths may be bidirectional communication paths that enable the elements to exchange information.

[0409] The various operations of the methods described above may be performed by any suitable means capable of performing the operations, such as, for example, various hardware and / or software component(s), circuits, and / or module(s). Generally, any operation illustrated in a figure may be performed by a corresponding functional means capable of performing the operation.

[0410] The various illustrative logic blocks, modules, and circuits described in connection with this disclosure (such as the blocks of FIGS. 2 and 4) may be implemented or performed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array signal (FPGA) or other programmable logic device (PLD), discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any commercially available processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, such as, for example, a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.

[0411] In one or more aspects, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. Storage media may be any available medium that can be accessed by a computer. By way of example, and not limitation, such computer-readable media may include various types of RAM, ROM, CD-ROM, or other optical disk storage, magnetic disk storage, or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Any connection is also properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technology such as infrared, radio, and microwave, the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technology such as infrared, radio, and microwave are included in the definition of medium. Disk and disc, as used herein, include compact disc (CD), laser disc, optical disk, digital versatile disk (DVD), floppy disk, and Blu-ray disc, where disks typically reproduce data magnetically, whereas discs reproduce data optically using lasers. Thus, in some aspects, computer readable media may include non-transitory computer readable media (e.g., tangible media). In addition, in some aspects, computer readable media may include transitory computer readable media (e.g., signals). Combinations of the above should also be included within the scope of computer readable media.

[0412] The methods disclosed herein include one or more steps or actions for achieving a desired method. The method steps and / or actions may be interchanged with one another without departing from the scope of the claims. In other words, unless a specific order of steps or actions is specified, the order and / or use of specific steps and / or actions may be modified without departing from the scope of the claims.

[0413] Certain aspects may comprise a computer program product for performing the operations presented herein. For example, such a computer program product may comprise a computer-readable medium having instructions stored (and / or coded) thereon, which instructions are executable by one or more processors to perform the operations described herein. For certain aspects, the computer program product may include packaging materials.

[0414] Software or instructions may also be transmitted over a transmission medium. For example, if the software is transmitted from a website, a server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless techniques such as infrared, radio, and microwave, the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless techniques such as infrared, radio, and microwave are included in the definition of transmission media.

[0415] Furthermore, it should be appreciated that modules and / or other suitable means for implementing the methods and techniques described herein may be downloaded and / or otherwise obtained by a user terminal and / or base station, where applicable. For example, such devices may be coupled to a server to facilitate the transfer of means for implementing the methods described herein. Alternatively, the various methods described herein may be provided via a storage means (e.g., a physical storage medium such as RAM, ROM, a compact disk (CD) or a floppy disk, etc.) such that the user terminal and / or base station may acquire the various methods upon coupling or providing the storage means to the device. Furthermore, any other suitable technique for providing the methods and techniques described herein to a device may be utilized.

[0416] It should be understood that the cla...

Claims

1. 1. A method for determining a fault type in an in-vivo continuous analyte monitoring system, comprising: receiving a signal from the analyte monitor; Receiving clinical status data; evaluating the clinical status data against clinical status criteria to determine clinical status information; determining the fault type based on both the received signal from the analyte monitor and the clinical context information; and and performing a response action based at least on the determined fault type.

2. 1. A method for determining a fault type in an in-vivo continuous analyte monitoring system, comprising: receiving a signal from the analyte monitor; Receiving clinical status data; evaluating the clinical status data against clinical status criteria to determine clinical status information; determining the fault type based solely on the received signal; and and performing a response action based on the determined fault type and the determined clinical context information.

3. 1. A method for performing a response process in response to a fault in an in vivo continuous analyte monitoring system, comprising: receiving a signal from the analyte monitor; Receiving clinical status data; evaluating the received clinical status data against clinical status criteria to determine clinical status information; and performing a response action based at least on the received signal and the determined clinical situation information.

4. 1. A method for determining a fault type in an in-vivo continuous analyte monitoring system, comprising: receiving a signal from the analyte monitor; Receiving clinical status data; converting said clinical context data into clinical context information; determining the fault type based on both the received signal from the analyte monitor and the clinical context information; and and performing a response action based at least on the determined fault type.

5. 1. A method for determining a fault type in an in-vivo continuous analyte monitoring system, comprising: receiving a signal from the analyte monitor; evaluating the received signals against fault criteria to determine fault information; determining clinical status information; determining the fault type based on both the fault information and the clinical condition information; and performing a response action based at least on the determined fault type.

6. The method of claim 1 , 2, 4, or 5, wherein the determining comprises classifying the fault based on the received signal, the clinical context information, or both.

7. 6. The method of claim 1, 2, 4, or 5, wherein the determining comprises classifying the fault based on the received signal, the clinical context information, or both, and the classifying the fault comprises classifying the fault as a sensor environment fault or as a system error / artifact fault.

8. The method of claim 7 , wherein the determining comprises classifying the fault as a sensor environment fault and further comprises subclassifying the fault as a compression fault or an early wound response fault.

9. 6. The method of claim 1, 2, 4, or 5, wherein the determining comprises determining whether the received signal or the received data matches or meets a predetermined criterion.

10. The method of claim 1 , 2, 4, or 5, wherein the determining comprises analyzing the signal using a time-based technique.

11. The method of claim 1 , 2, 4, or 5, wherein the determining comprises analyzing the signal using a frequency-based technique.

12. The method of claim 1 , 2, 4, or 5, wherein the determining comprises analyzing the signal using a wavelet-based technique.

13. The method of claim 1 , 2, 4, or 5, wherein the determining comprises raw signal analysis.

14. The method of claim 1 , 2, 4, or 5, wherein the determining comprises residual signal analysis.

15. The method of claim 1 , 2, 4, or 5, wherein the determining comprises pattern analysis.

16. The method of claim 1 , 2, 4, or 5, wherein the determining comprises slow versus fast sampling.

17. The method of claim 1 , 2, 4, or 5, wherein the determining comprises projecting the received signal onto a plurality of templates, each template corresponding to a failure mode.

18. The method of claim 1 , 2, 4, or 5, wherein the determining comprises a variability analysis.

19. The method of claim 1 , 2, 4, or 5, wherein the determining comprises a fuzzy logic analysis.

20. 6. The method of claim 1, wherein the received clinical status data is selected from the group consisting of age, anthropometric data, medications currently acting on the patient, temperature compared to a standard, the patient's malfunction history, the patient's activity level, the patient's exercise level, the patient's level of interaction with a glucose monitor, a pattern of glucose signal values, clinical glucose values ​​and their derivatives, a range of a patient's glucose levels over a period of time, the duration that a patient's glucose level has been maintained in a range, a patient's glucose state, a glycemic urgency index, time, or pressure.

21. The method of any one of claims 1 to 5, further comprising processing said signal.

22. The method of any one of claims 1 to 5, further comprising processing the signal, said processing removing or filtering noise from the signal.

23. The method of any one of claims 1 to 5, further comprising receiving a further signal.

24. The method of any one of claims 1 to 5, further comprising receiving a further signal, said further signal being a sensor temperature signal, an impedance signal, an oxygen signal, a pressure signal, or a background signal.

25. The method of any one of claims 1 to 5, wherein the clinical status information corresponds to data regarding the patient other than a signal value measured by a sensor associated with the analyte monitor.

26. The method of claims 1 to 3, wherein the clinical situation criteria include predefined values ​​or ranges of parameters selected from the group consisting of medications currently acting on the patient, temperature, the patient's malfunction history, the patient's activity level, the patient's exercise level, the patient's level of interaction with a glucose monitor, a pattern of glucose signal values, clinical glucose values ​​and their derivatives, a range of a patient's glucose levels over a period of time, the duration that a patient's glucose level is maintained within a range, a patient's glucose state, a glycemic urgency index, time of day, or pressure.

27. The method of claim 1, 2 or 4, wherein the clinical condition data includes temperature, the clinical condition criteria includes a pattern of temperature, the evaluating determines the clinical condition information as a user being in contact with water at the sensor site, and the determining the fault type includes determining the fault type as water intrusion.

28. The method of claim 1, 2 or 4, wherein the clinical condition data includes a patient's activity level or time of day, the clinical condition criteria includes a pattern of the patient's activity level, the evaluating determines the clinical condition information as the user applying pressure to the sensor site, and the determining the fault type includes determining the fault type as pressure.

29. The method of claim 1, wherein the clinical condition data includes an implantation time and the clinical condition criteria includes a range of implantation times during which a dip and recover failure may occur, the evaluating step determines the clinical condition information as the sensor was recently implanted, and the determining the failure type includes determining the failure type as a dip and recover failure.

30. The method of any one of claims 1 to 5, wherein the clinical status data comprises a clinical glucose value and data selected from the group consisting of age, anthropometric data, activity, exercise, clinical use of data, or patient interaction with a monitor.

31. The method of any one of claims 1 to 5, wherein the response process includes providing an indication to a user, the indication including a warning, a warning, an alarm, a confidence indicator, a range of values, a predicted value, or a blank screen.

32. The method according to any one of claims 1 to 5, wherein performing said response processing comprises adjusting a level of filtering of said received signal.

33. The method according to any one of claims 1 to 5, wherein performing the response processing comprises performing a prediction of future signal values ​​based on the received signals.

34. The method of any one of claims 1 to 5, wherein performing the response process includes performing a self-diagnostic routine.

35. The method according to any one of claims 1 to 5, wherein performing said response processing comprises performing a compensation step.

36. The method of any one of claims 1 to 5, wherein performing the response process comprises switching from a first treatment mode to a second treatment mode.

37. A system for carrying out the method according to any of claims 1 to 36.

38. 20. A device substantially as hereinbefore shown and / or described in the present specification and / or drawings.

39. 20. A system substantially as hereinbefore shown and / or described in the present specification and / or drawings.

40. 20. A method substantially as hereinbefore described and / or as illustrated in the present specification and / or drawings.

41. 1. An electronic device for monitoring data associated with a physiological condition, comprising: a continuous analyte sensor configured to substantially continuously measure a concentration of an analyte in a host and provide continuous sensor data associated with the analyte concentration in the host; A processor module configured to perform any one of the methods according to claims 1 to 36.

42. 42. The device of claim 41, wherein the analyte is glucose.

43. 1. An electronic device for delivering a drug to a host, comprising: a drug delivery device configured to deliver a drug to the host, the drug delivery device operably connected to a continuous analyte sensor, the continuous analyte sensor configured to substantially continuously measure a concentration of an analyte in the host and provide continuous sensor data associated with the analyte concentration in the host; A processor module configured to perform any one of the methods according to claims 1 to 36.

44. 44. The device of claim 43, wherein the analyte is glucose and the drug is insulin.

Images