End-of-life detection for analyte sensors experiencing gradual sensor degradation

Abstract

The present application provides systems and methods for processing sensor data from a continuous analyte sensor and detecting end of life of the sensor. The method includes receiving sensor data from the analyte sensor, evaluating a plurality of risk factors associated with an end of life condition of the analyte sensor, the risk factors including downward drift of sensor sensitivity over time, an amount of asymmetric non-stationary noise, and a duration of the noise, determining an end of life status of the analyte sensor based at least in part on the evaluating, and providing an output related to the end of life status of the analyte sensor.

Description

[Technical field]

[0001] Related Applications This application claims the benefit of U.S. Provisional Application No. 63 / 265,324, filed December 13, 2021, which is incorporated by reference in its entirety herein.

[0002] SUMMARY OF THE DISCLOSURE The embodiments described herein generally relate to systems and methods for processing sensor data from continuous analyte sensors and detecting end of life of the sensors. [Background technology]

[0003] Diabetes mellitus is a disease in which the pancreas cannot make enough insulin (Type I or insulin-dependent) and / or insulin is not effective (Type 2 or non-insulin-dependent). In the diabetic state, the victim suffers from hyperglycemia, which can cause many physiological disorders associated with deterioration of small blood vessels, such as kidney failure, skin ulcers, or bleeding into the vitreous of the eye. A hypoglycemic reaction (hypoglycemia) can be precipitated by inadvertent overdosing of insulin or after normal administration of insulin or glucose-lowering agents accompanied by abnormal exercise or inadequate food intake.

[0004] Traditionally, people with diabetes carry self-monitoring blood glucose (SMBG) monitors, which typically require an uncomfortable finger prick to obtain a blood sample for measurement. Due to the lack of comfort and convenience associated with finger pricking, people with diabetes typically measure their glucose levels only two to four times per day. Unfortunately, if the time interval between measurements is too far apart, the diabetic patient may not know about a hyperglycemic or hypoglycemic condition until it is too late, sometimes resulting in dangerous side effects. Not only will a diabetic patient not be able to measure their SMBG values ​​in a timely manner, but the traditional method also increases the likelihood that they will not know if their blood glucose levels are rising (high) or falling (low). Thus, a diabetic patient may not be able to make educated insulin treatment decisions.

[0005] Another device that some diabetics use to monitor blood glucose is a continuous analyte sensor. A continuous analyte sensor typically includes a sensor that is placed subcutaneously, transcutaneously (e.g., transdermally), or intravascularly. The sensor measures the concentration of a given analyte in the body and generates a raw signal that is transmitted to electronics associated with the sensor. The raw signal is converted to an output value that is displayed on a display. The output value resulting from the conversion of the raw signal is typically expressed in a form that provides meaningful information to a user, such as blood glucose expressed in mg / dL.

[0006] One of the main perceived advantages of continuous analyte sensors is the ability of these devices to be used continuously for several days, e.g., 1, 3, 5, 6, 7, 9, 10, 14 days or more. Although these various devices may be approved for a certain number of days and may be used "off-label" beyond their approved days, the performance of the sensor is known to degrade over its lifespan. And since no two environments are the same for any two sensors, the lifespan of any particular sensor may actually be less than the approved lifespan of the sensor. Therefore, it would be beneficial to know the condition or time when a sensor is nearing the end of its lifespan so that the user can be informed that the sensor should be replaced. Summary of the Invention [Means for solving the problem]

[0007] In a first exemplary aspect, a method for determining end of life of a continuous analyte sensor is provided, the method including receiving sensor data from the analyte sensor; evaluating a plurality of risk factors associated with end of life symptoms of the analyte sensor, the risk factors including downward drift of sensor sensitivity over time, an amount of asymmetric non-stationary noise, and a duration of the noise; determining an end of life status of the analyte sensor based at least in part on the evaluating; and providing an output related to the end of life status of the analyte sensor.

[0008] In an exemplary embodiment of the first aspect, determining the end of life status further includes mapping each of the risk factors to an end of life risk factor metric.

[0009] In an exemplary embodiment of the first aspect, the mapping is performed using a different substitution function for each of the risk factors.

[0010] In an exemplary embodiment of the first aspect, at least one of the substitution functions is a logistic regression function.

[0011] In an exemplary embodiment of the first aspect, assigning a weight to each of the end-of-life risk factors and determining a weighted average from the weighted risk factor metrics, and determining the end-of-life status of the analyte sensor is based at least in part on the weighted average.

[0012] In an exemplary embodiment of the first aspect, the determining further includes comparing the weighted average to a threshold and determining that the sensor end of life is likely when the weighted average exceeds the threshold.

[0013] In an exemplary embodiment of the first aspect, the output recommends terminating sensor usage when the weighted average exceeds a threshold.

[0014] In an exemplary embodiment of the first aspect, the plurality of risk factors further comprises a rate of change of downward drift of the sensor sensitivity.

[0015] In an exemplary embodiment of the first aspect, determining the end of life status further comprises mapping each of the risk factors to an end of life risk factor.

[0016] In an exemplary embodiment of the first aspect, the mapping is performed using a common substitution function that receives each of the risk factors.

[0017] In an exemplary embodiment of the first aspect, the substitution function is a logistic regression function.

[0018] In an exemplary embodiment of the first aspect, determining the downward drift of the sensor sensitivity by examining the ratio of short-term sensitivity to long-term sensitivity.

[0019] In an exemplary embodiment of the first aspect, determining downward drift in sensor sensitivity by down-sampling the sensor signal, periodically determining a slope between selected data points in the down-sampled signal to obtain a series of calculated slopes, periodically comparing the currently calculated slope to a previously calculated slope, and determining that the sensor sensitivity is decreasing if a ratio of the currently calculated slope to the previously calculated slope decreases over time.

[0020] In an exemplary embodiment of the first aspect, determining a downward drift in the sensor sensitivity by: filtering the sensor data with a slow low-pass filter having a lower cutoff frequency than the fast low-pass filter to obtain slow filtered sensor data; filtering the sensor data with the fast low-pass filter to obtain fast filtered sensor data; and examining a ratio of a difference between the fast filtered sensor data and the slow filtered sensor data to the slow filtered sensor data, the more negative the ratio, the more likely the downward drift in the sensor sensitivity.

[0021] In an exemplary embodiment of the first aspect, determining a downward drift in sensor sensitivity by examining a change in the magnitude of the sensor sensitivity.

[0022] In an exemplary embodiment of the first aspect, determining the downward drift of the sensor sensitivity by examining the rate of change of the sensor sensitivity.

[0023] In an exemplary embodiment of the first aspect, determining the rate of change of the sensor sensitivity includes filtering the sensor signal with a low pass filter to obtain a filtered sensor signal; periodically applying moving windows of specified duration to the filtered sensor signal; calculating, for each of the moving windows, a gradient between a local maximum and a local minimum in the filtered sensor signal; periodically comparing the gradient calculated for a current one of the moving windows with a gradient calculated for a previous one of the moving windows; and determining a rate of change of the sensor sensitivity based on the comparison.

[0024] In an exemplary embodiment of the first aspect, determining a rate of change of the sensor sensitivity based on the comparison includes determining that the sensor sensitivity degradation is increasing when the periodic comparison reveals that the calculated slope is increasing over time.

[0025] In a second exemplary aspect, a system for determining end of life of a continuous analyte sensor is provided, the system comprising sensor electronics configured to be operatively connected to the continuous analyte sensor, the sensor electronics configured to receive a sensor signal from the analyte sensor; evaluate a plurality of risk factors associated with end of life symptoms of the analyte sensor, the risk factors including downward drift of sensor sensitivity over time, an amount of asymmetric non-stationary noise, and a duration of the noise; determine an end of life status of the analyte sensor based at least in part on the evaluating; and provide an output related to the end of life status of the analyte sensor.

[0026] In an exemplary embodiment of the second aspect, the sensor electronics are further configured to determine an end-of-life status by mapping each of the risk factors to an end-of-life risk factor metric.

[0027] In an exemplary embodiment of the second aspect, the mapping is performed using a different substitution function for each of the risk factors.

[0028] In an exemplary embodiment of the second aspect, at least one of the substitution functions is a logistic regression function.

[0029] In an exemplary embodiment of the second aspect, the sensor electronics is further configured to assign a weight to each of the end-of-life risk factors and determine a weighted average from the weighted risk factor metrics, and determining the end-of-life status of the analyte sensor is based at least in part on the weighted average.

[0030] In an exemplary embodiment of the second aspect, the sensor electronics are further configured to compare the weighted average to a threshold and determine that the sensor end of life is likely when the weighted average exceeds the threshold.

[0031] In an exemplary embodiment of the second aspect, the output recommends terminating sensor usage when the weighted average exceeds a threshold.

[0032] In an exemplary embodiment of the second aspect, the plurality of risk factors further comprises a rate of change of downward drift of the sensor sensitivity.

[0033] In an exemplary embodiment of the second aspect, the sensor electronics are further configured to determine an end-of-life status by mapping each of the risk factors to an end-of-life risk factor.

[0034] In an exemplary embodiment of the second aspect, the mapping is performed using a common substitution function that receives each of the risk factors.

[0035] In an exemplary embodiment of the second aspect, the substitution function is a logistic regression function.

[0036] In an exemplary embodiment of the second aspect, the sensor electronics is further configured to determine a downward drift in the sensor sensitivity by examining a ratio of short-term sensitivity to long-term sensitivity.

[0037] In an exemplary embodiment of the second aspect, the sensor electronics are further configured to determine a downward drift in the sensor sensitivity by down-sampling the sensor signal; periodically determining a slope between selected data points in the down-sampled signal to obtain a series of calculated slopes; periodically comparing the currently calculated slope to a previously calculated slope; and determining that the sensor sensitivity is decreasing if a ratio of the currently calculated slope to the previously calculated slope decreases over time.

[0038] In an exemplary embodiment of the second aspect, the sensor electronics are further configured to determine a downward drift of the sensor sensitivity by filtering the sensor data with a slow low pass filter having a lower cutoff frequency than the fast low pass filter to obtain slow filtered sensor data, filtering the sensor data with the fast low pass filter to obtain fast filtered sensor data, and examining a ratio of a difference between the fast filtered sensor data and the slow filtered sensor data to the slow filtered sensor data, where the more negative the ratio, the more likely the downward drift of the sensor sensitivity over time.

[0039] In an exemplary embodiment of the second aspect, the sensor electronics is further configured to determine a downward drift in the sensor sensitivity by examining a change in the magnitude of the sensor sensitivity.

[0040] In an exemplary embodiment of the second aspect, the sensor electronics is further configured to determine a downward drift in the sensor sensitivity by examining a rate of change of the sensor sensitivity.

[0041] In an exemplary embodiment of the second aspect, the sensor electronics are further configured to determine a rate of change of the sensor sensitivity by filtering the sensor signal with a low pass filter to obtain a filtered sensor signal; periodically applying moving windows of specified duration to the filtered sensor signal; calculating, for each of the moving windows, a gradient between a local maximum and a local minimum in the filtered sensor signal; periodically comparing the gradient calculated for a current one of the moving windows with a gradient calculated for a previous one of the moving windows; and determining a rate of change of the sensor sensitivity based on the comparison.

[0042] In an exemplary embodiment of the second aspect, the sensor electronics are further configured to determine a rate of change of the sensor sensitivity based on the comparison by determining that the sensor sensitivity degradation is increasing if the periodic comparison reveals that the calculated slope is increasing over time.

[0043] In a third exemplary aspect, a method for determining end of life of a continuous analyte sensor is provided, the method including receiving sensor data from the analyte sensor; comparing the sensor signal to one or more predefined end of life signature patterns associated with end of life symptoms of the analyte sensor, at least one of the predefined end of life signature patterns including a noise pattern representative of asymmetric non-stationary noise; determining an end of life status of the analyte sensor based at least in part on the comparing; and providing an output related to the end of life status of the analyte sensor.

[0044] In an exemplary embodiment of the third aspect, the method further includes determining a noise risk factor associated with the end of life condition of the analyte sensor based on the comparison, determining at least one additional risk factor associated with the end of life condition of the analyte sensor, and evaluating the noise risk factor and the at least one additional risk factor, wherein determining an end of life status of the analyte sensor based at least in part on the comparing includes determining an end of life status of the analyte sensor based on the evaluating.

[0045] In an exemplary embodiment of the third aspect, the at least one additional risk factor comprises at least one risk factor selected from the group consisting of: downward drift of sensor sensitivity over time, duration of noise, and rate of change of downward drift of sensor sensitivity.

[0046] In an exemplary embodiment of the third aspect, the at least one additional risk factor includes a downward drift of the sensor sensitivity over time, a duration of the noise, and a rate of change of the downward drift of the sensor sensitivity.

[0047] In an exemplary embodiment of the third aspect, determining the end-of-life status further includes mapping each of the risk factors to an end-of-life risk factor metric.

[0048] In an exemplary embodiment of the third aspect, the mapping is performed using a different substitution function for each of the risk factors.

[0049] In an exemplary embodiment of the third aspect, at least one of the substitution functions is a logistic regression function.

[0050] In an exemplary embodiment of the third aspect, the method further includes assigning a weight to each of the end-of-life risk factors and determining a weighted average from the weighted risk factor metrics, wherein determining the end-of-life status of the analyte sensor is based at least in part on the weighted average.

[0051] In an exemplary embodiment of the third aspect, the determining further includes comparing the weighted average to a threshold and determining that the sensor end of life is likely when the weighted average exceeds the threshold.

[0052] In an exemplary embodiment of the third aspect, the output recommends terminating sensor usage when the weighted average exceeds a threshold.

[0053] In a fourth exemplary aspect, a system for determining end of life of a continuous analyte sensor is provided, the system comprising sensor electronics configured to be operatively connected to the continuous analyte sensor, the sensor electronics configured to: evaluate a plurality of risk factor metrics associated with end of life symptoms of the sensor, each of the plurality of risk factors indicative in part of end of life of the sensor based on a comparison of the risk factor to one or more criteria, the risk factors including a downward drift of sensor sensitivity over time, an amount of asymmetric non-stationary noise, and a duration of the noise; and determine an end of life status of the sensor based on the evaluation of the plurality of risk factors, where if at least two of the plurality of risk factors are determined to each satisfy one or more criteria, then a combination of the at least two variables is indicative of end of life of the sensor and provides an output related to the end of life status of the sensor.

[0054] In an exemplary embodiment of the fourth aspect, the plurality of risk factors further comprises a rate of change of downward drift of the sensor sensitivity.

[0055] In an exemplary embodiment of the fourth aspect, the plurality of risk factors further includes a number of days the sensor has been used, and assessing the number of days the sensor has been used includes at least one of assessing an amount of time since the sensor was initialized, assessing an amount of time since the sensor was implanted, or assessing an amount of time since the sensor was first calibrated.

[0056] In an exemplary embodiment of the fourth aspect, the sensor electronics are further configured to determine an end-of-life status by mapping each of the risk factors to an end-of-life risk factor metric.

[0057] In an exemplary embodiment of the fourth aspect, the mapping is performed using a different substitution function for each of the risk factors.

[0058] In an exemplary embodiment of the fourth aspect, at least one of the substitution functions is a logistic regression function.

[0059] In an exemplary embodiment of the fourth aspect, the sensor electronics is further configured to assign a weight to each of the end-of-life risk factors and determine a weighted average from the weighted risk factor metrics, and determining the end-of-life status of the analyte sensor is based at least in part on the weighted average.

[0060] In an exemplary embodiment of the fourth aspect, the sensor electronics are further configured to compare the weighted average to a threshold and determine that the sensor end of life is likely when the weighted average exceeds the threshold.

[0061] In an exemplary embodiment of the fourth aspect, the output recommends terminating sensor usage when the weighted average exceeds a threshold.

[0062] In an exemplary embodiment of the fourth aspect, the sensor electronics are further configured to determine an end-of-life status by mapping each of the risk factors to an end-of-life risk factor.

[0063] In an exemplary embodiment of the fourth aspect, the mapping is performed using a common substitution function that receives each of the risk factors.

[0064] In an exemplary embodiment of the fourth aspect, the substitution function is a logistic regression function.

[0065] In an exemplary embodiment of the fourth aspect, the sensor electronics is further configured to determine a downward drift in the sensor sensitivity by examining a ratio of short-term sensitivity to long-term sensitivity.

[0066] In an exemplary embodiment of the fourth aspect, the sensor electronics are further configured to determine a downward drift in the sensor sensitivity by down-sampling the sensor signal; periodically determining a slope between selected data points in the down-sampled signal to obtain a series of calculated slopes; periodically comparing the currently calculated slope to a previously calculated slope; and determining that the sensor sensitivity is decreasing if a ratio of the currently calculated slope to the previously calculated slope decreases over time.

[0067] In an exemplary embodiment of the fourth aspect, the sensor electronics are further configured to determine a downward drift in the sensor sensitivity by filtering the sensor data with a slow low pass filter having a lower cutoff frequency than the fast low pass filter to obtain slow filtered sensor data, filtering the sensor data with the fast low pass filter to obtain fast filtered sensor data, and examining a ratio of a difference between the fast filtered sensor data and the slow filtered sensor data to the slow filtered sensor data, where the more negative the ratio, the more likely the downward drift in the sensor sensitivity.

[0068] In an exemplary embodiment of the fourth aspect, the sensor electronics is further configured to determine a downward drift in the sensor sensitivity by examining a change in the magnitude of the sensor sensitivity.

[0069] In an exemplary embodiment of the fourth aspect, the sensor electronics is further configured to determine a downward drift in the sensor sensitivity by examining a rate of change of the sensor sensitivity.

[0070] In an exemplary embodiment of the fourth aspect, the sensor electronics are further configured to determine a rate of change of the sensor sensitivity by filtering the sensor signal with a low pass filter to obtain a filtered sensor signal; periodically applying moving windows of specified duration to the filtered sensor signal; calculating, for each of the moving windows, a gradient between a local maximum and a local minimum in the filtered sensor signal; periodically comparing the gradient calculated for a current one of the moving windows with a gradient calculated for a previous one of the moving windows; and determining a rate of change of the sensor sensitivity based on the comparison.

[0071] In an exemplary embodiment of the fourth aspect, the sensor electronics are further configured to determine a rate of change of the sensor sensitivity based on the comparison by determining that the sensor sensitivity degradation is increasing if the periodic comparison reveals that the calculated slope is increasing over time.

[0072] In a fifth exemplary aspect, a system for determining end of life of a continuous analyte sensor is provided, the system comprising sensor electronics configured to be operatively connected to the continuous analyte sensor, the sensor electronics configured to: evaluate a plurality of risk factor metrics associated with end of life symptoms of the sensor, the risk factors including downward drift of sensor sensitivity over time, an amount of asymmetric non-stationary noise, and a duration of the noise; replace an output of the plurality of risk factor evaluation with an end of life risk factor value; determine a likelihood of recovery based on the end of life risk factor value; determine an end of life status of the sensor based on the likelihood of recovery; and provide an output related to the end of life status of the sensor.

[0073] In an exemplary embodiment of the fifth aspect, the plurality of risk factors further comprises a rate of change of downward drift of the sensor sensitivity.

[0074] In an exemplary embodiment of the fifth aspect, the sensor electronics is further configured to determine a downward drift in the sensor sensitivity by examining a ratio of short-term sensitivity to long-term sensitivity.

[0075] In an exemplary embodiment of the fifth aspect, the sensor electronics are further configured to determine a downward drift in the sensor sensitivity by down-sampling the sensor signal; periodically determining a slope between selected data points in the down-sampled signal to obtain a series of calculated slopes; periodically comparing the currently calculated slope to a previously calculated slope; and determining that the sensor sensitivity is decreasing if a ratio of the currently calculated slope to the previously calculated slope decreases over time.

[0076] In an exemplary embodiment of the fifth aspect, the sensor electronics are further configured to determine a downward drift of the sensor sensitivity by filtering the sensor data with a slow low pass filter having a lower cutoff frequency than the fast low pass filter to obtain slow filtered sensor data, filtering the sensor data with the fast low pass filter to obtain fast filtered sensor data, and examining a ratio of a difference between the fast filtered sensor data and the slow filtered sensor data to the slow filtered sensor data, where the more negative the ratio, the more likely the downward drift of the sensor sensitivity is.

[0077] In a sixth exemplary aspect, a method for determining an end-of-life status of a continuous analyte sensor is provided, the method including receiving sensor data from the analyte sensor; evaluating a plurality of risk factor metrics associated with end-of-life symptoms of the analyte sensor, at least a portion of the plurality of risk factor metrics being based on the sensor data, the plurality of risk factor metrics including downward drift of sensor sensitivity over time, an amount of asymmetric non-stationary noise, and a duration of the noise; determining an end-of-life status of the analyte sensor based at least in part on the evaluating; and providing an output related to the end-of-life status of the analyte sensor.

[0078] In an exemplary embodiment of the sixth aspect, the method further includes replacing a first risk factor metric of the plurality of risk factor metrics with a first end-of-life risk factor value, the first end-of-life risk factor value indicating a likelihood that the first risk factor metric indicates that the continuous analyte sensor is in an end-of-life condition.

[0079] In an exemplary embodiment of the sixth aspect, the method further includes replacing a second risk factor metric of the plurality of risk factor metrics with a second end-of-life risk factor value, wherein replacing the first risk factor metric is performed using a first substitution function and replacing the second risk factor metric is performed using a second substitution function different from the first substitution function.

[0080] In an exemplary embodiment of the sixth aspect, at least one of the first substitution function or the second substitution function is a logistic regression function.

[0081] In an exemplary embodiment of the sixth aspect, the method further includes assigning a first weight to the first end-of-life risk factor value to determine a first end-of-life risk factor value, assigning a second weight to the second end-of-life risk factor value to determine a second end-of-life risk factor value, and determining a weighted average using the first end-of-life risk factor value and the second end-of-life risk factor value.

[0082] In an exemplary embodiment of the sixth aspect, the determining further includes determining that the weighted average meets a threshold value and determining that the continuous analyte sensor is in an end-of-life condition based on determining that the weighted average meets the threshold value.

[0083] In an exemplary embodiment of the sixth aspect, the method further includes initiating a termination of sensor use.

[0084] In an exemplary embodiment of the sixth aspect, the plurality of risk factor metrics further comprises a rate of change of downward drift of the sensor sensitivity.

[0085] In an exemplary embodiment of the sixth aspect, the method further includes substituting the plurality of risk factor metrics with an end-of-life likelihood using a common substitution function.

[0086] In an exemplary embodiment of the sixth aspect, the substitution function is a logistic regression function.

[0087] In an exemplary embodiment of the sixth aspect, the method further includes determining, at least in part, the downward drift of the sensor sensitivity over time by examining a ratio of short-term sensitivity to long-term sensitivity.

[0088] In an exemplary embodiment of the sixth aspect, the method further includes determining, at least in part, a downward drift of the sensor sensitivity over time by downsampling the sensor data to generate a downsampled signal; periodically determining a slope between selected data points in the downsampled signal to obtain a series of calculated slopes; periodically comparing the currently calculated slope to a previously calculated slope; and determining that the sensor sensitivity is decreasing if a ratio of the currently calculated slope to the previously calculated slope decreases over time.

[0089] In an exemplary embodiment of the sixth aspect, the method includes filtering the sensor data with a slow low-pass filter having a lower cutoff frequency than the fast low-pass filter to obtain slow filtered sensor data; The method further includes determining the downward drift of the sensor sensitivity over time, at least in part, by filtering the sensor data with a fast low pass filter to obtain fast filtered sensor data, and examining a ratio of a difference between the fast filtered sensor data and the slow filtered sensor data to the slow filtered sensor data, where the more negative the ratio, the more likely the downward drift of the sensor sensitivity over time.

[0090] In an exemplary embodiment of the sixth aspect, the method further includes applying the classifier model to generate an end-of-life likelihood for the continuous analyte sensor based at least in part on the plurality of risk factor metrics, and an end-of-life status of the analyte sensor is based at least in part on the end-of-life likelihood.

[0091] In an exemplary embodiment of the sixth aspect, the method further includes applying multiple sequential processing layers to the sensor data, where applying a first processing layer of the multiple processing layers includes applying a first filter to the sensor data to generate a first filter output and downsampling the first filter output to generate the first layer output, where applying a second processing layer of the multiple processing layers includes applying a second filter to the first layer output to generate a second filter output and downsampling the second filter output to generate a second layer output, where at least one of the multiple risk factor metrics is based at least in part on the second layer output.

[0092] In an exemplary embodiment of the sixth aspect, the method further includes determining, at least in part, the downward drift of the sensor sensitivity over time by examining a rate of change of the sensor sensitivity, where examining the rate of change of the sensor sensitivity includes filtering the sensor data with a low pass filter to obtain filtered sensor data; determining, for a first moving window, a first gradient between a local maximum of the filtered sensor data in the first moving window and a local minimum of the filtered sensor data in the first moving window; determining, for a second moving window after the first moving window, a second gradient between a local maximum of the filtered sensor data in the second moving window and a local minimum of the filtered sensor data in the second moving window; determining a rate of change of the sensor sensitivity based on the first slope and the second slope.

[0093] In an exemplary embodiment of the sixth aspect, determining the rate of change of the sensor sensitivity includes determining that the downward drift of the sensor sensitivity is increasing when the second slope is greater than the first slope.

[0094] In an exemplary embodiment of the sixth aspect, the method further includes determining a noise metric describing the sensor data; determining that the sensor data does not produce a reliable output based at least in part on the end-of-life status of the analyte sensor and the noise metric; and suspending display of at least one value based on the sensor data in response to determining that the sensor data does not produce a reliable output.

[0095] In an exemplary embodiment of the sixth aspect, the method further includes receiving second sensor data from the analyte sensor after pausing the display of the at least one value based on the sensor data; evaluating a second plurality of risk factor metrics, at least a portion of the second plurality of risk factor metrics based at least in part on the second sensor data; determining a second end of life status of the analyte sensor based at least in part on the evaluating; determining a second noise metric describing the second sensor data; determining that the second sensor data produces a reliable output based at least in part on the second end of life status of the analyte sensor and the second noise metric; and displaying the at least one value based on the second sensor data in response to determining that the second sensor data produces a reliable output.

[0096] In a seventh exemplary aspect, a continuous analyte sensor system is provided. The continuous analyte sensor system can include a processor circuit programmed to perform operations including receiving sensor data from the analyte sensor, evaluating a plurality of risk factor metrics associated with end-of-life conditions of the analyte sensor, at least a portion of the plurality of risk factor metrics being based on the sensor data, the plurality of risk factor metrics including a downward drift of sensor sensitivity over time, an amount of asymmetric non-stationary noise, and a duration of the noise, determining an end-of-life status of the analyte sensor based at least in part on the evaluating, and providing an output related to the end-of-life status of the analyte sensor. [Brief description of the drawings]

[0097] The details of the present disclosure, both as to its structure and operation, can be understood in part by studying the accompanying drawings, in which like reference numerals refer to like parts, and in which the drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the present disclosure. [Figure 1A] 1 illustrates a schematic diagram of sensor sensitivity as a function of time during a sensor session, according to various embodiments. [Figure 1B] 1B shows a schematic diagram of regression-derived sensitivity and baseline at different periods of a sensor session according to the embodiment of FIG. 1A. [Figure 2A] 10A-10D collectively illustrate different embodiments of a process for generating a sensor sensitivity profile. [Figure 2B] 10A-10D collectively illustrate different embodiments of a process for generating a sensor sensitivity profile. [Figure 2C] 10A-10D collectively illustrate different embodiments of a process for generating a sensor sensitivity profile. [Figure 2D] 10A-10D collectively illustrate different embodiments of a process for generating a sensor sensitivity profile. [Diagram 3]1 shows a diagram illustrating different types of information that may be input into a sensor system to define a sensor sensitivity profile over time. [Figure 4] 1 illustrates a schematic diagram of sensor sensitivity as a function of time between completion of sensor fabrication and start of a sensor session, according to various embodiments. [Diagram 5] 1 illustrates a diagram of a calibration process for forming a transformation function using various inputs, according to various embodiments. [Figure 6] 1 shows a flowchart illustrating a process for determining or detecting outlier matched data pairs, according to an embodiment. [Figure 7] 1 illustrates a diagram of a calibration set having six matched data pairs, according to various embodiments. [Figure 8] 1 shows a diagram of a calibration line plotted using Ordinary Least Square (OLS), according to an embodiment. [Figure 9] 1 shows a flowchart illustrating a process for dynamically and intelligently determining end of life for a sensor, according to an embodiment. [Figure 10] 10 shows a flowchart illustrating exemplary inputs and outputs for the flowchart of FIG. [Figure 11A] 1 shows a sensor signal with loss of sensitivity (downward slope of the signal) indicative of an end of life condition. [Figure 11B] 13 illustrates the use of a spike detector to detect an increase in downward spikes in the signal generated by the sensor, which is a particular risk factor for end of life. [Figure 12A] 11B shows the power spectral density of the sensor signal shown in FIG. 11A indicative of an end of life condition. [Figure 12B] The same type of graph as FIG. 12A is shown, but from a sensor signal that did not exhibit end-of-life symptoms. [Figure 13A] In one embodiment, a substitution function that maps noise durations to end-of-life risk factor values ​​is shown. [Figure 13B]In one embodiment, a substitution function that maps errors in calibration to end-of-life risk factor values ​​is shown. [Figure 13C] In one embodiment, a substitution function that may be used by the processor circuit to replace the combined end-of-life index calculated from the various risk factors with an end-of-life score and / or end-of-life status is shown. [Figure 14] 1 shows a flowchart illustrating a process for determining associated sensor reuse, according to an embodiment. [Figure 15] 1 shows a diagram illustrating noise duration associated with end of life, according to an embodiment. [Figure 16] 1 illustrates an exemplary process flow that may be performed by a processor circuit associated with a continuous analyte sensor to determine an end-of-life condition for the continuous analyte sensor. [Figure 17] 13 illustrates another exemplary process flow that may be performed by a processor circuit associated with a continuous analyte sensor to determine an end-of-life status for the continuous analyte sensor. [Figure 18] 13 illustrates another exemplary process flow that may be performed by a processor circuit associated with a continuous analyte sensor to determine an end-of-life condition of the continuous analyte sensor. [Figure 19] For example, one embodiment of a process flow that may be applied to sensor data, such as estimated glucose value (EGV), to reduce the impact of glycemic management on a host / patient is shown. [Figure 20] 1 shows a graph of end-of-life likelihood versus noise measure with a sample of sensor data plotted. [Figure 21] 21 illustrates one example of a process flow 2100 that may be executed by a processor circuit to suspend display of sensor data based on an end-of-life likelihood in combination with at least one other sensor characteristic, such as a noise metric. [Figure 22] FIG. 1 illustrates an example of an environment including an analyte sensor system. [Diagram 23] FIG. 23 illustrates an example of a medical device system including the analyte sensor system of FIG. 22. [Figure 24] FIG. 1 is a side view of an exemplary analyte sensor that may be implanted in a host. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0098] definition To facilitate understanding of the embodiments described herein, several terms are defined below.

[0099] As used herein, the term "analyte" 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 special or customized meaning) and refers to, but is not limited to, a substance or chemical component 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 heads, devices, and methods disclosed herein is glucose.

[0100] As used herein, the terms "continuous analyte sensor" and "continuous glucose sensor" are broad terms and are to be given their ordinary and customary meaning to those of skill in the art (and are not to be limited to any special or customized meaning), and refer to, but are not limited to, a device that continuously or continuously measures analyte / glucose concentration and / or calibrates the device (e.g., by continuously or continuously adjusting or determining the sensitivity and background of the sensor) at time intervals ranging from, for example, a fraction of a second up to, for example, 1, 2, or 5 minutes or longer.

[0101] As used herein, the terms "raw data stream" and "data stream" 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 special or customized meaning) and refer to, but are not limited to, an analog or digital signal directly related to an analyte concentration measured by an analyte sensor. In one embodiment, a raw data stream is digital data in counts converted by an A / D converter from an analog signal (e.g., voltage or amperes) representing the analyte concentration. These terms broadly encompass multiple time-spaced matched data pairs from a substantially continuous analyte sensor, including individual measurements taken at time intervals ranging from a fraction of a second up to, for example, one, two, or five minutes or longer.

[0102] As used herein, the term "sensor data" 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 special or customized meaning), further referring to, but not limited to, any data associated with a sensor, such as a continuous analyte sensor. Sensor data includes raw data streams or simply data streams of analog or digital signals directly related to the analyte measured from the analyte sensor (or other signals received from another sensor), as well as calibrated and / or filtered versions of the raw data. In one embodiment, the sensor data includes digital data of "counts" converted by an A / D converter from an analog signal (e.g., voltage or amperes) and includes one or more data points representing glucose concentration. Thus, the terms "sensor data point" and "data point" generally refer to a digital representation of sensor data at a particular time. These terms broadly encompass multiple time interval data points from a sensor, such as a substantially continuous glucose sensor, including individual measurements taken at time intervals ranging from a fraction of a second up to, for example, one, two, or five minutes or longer. In another embodiment, the sensor data includes an integrated digital value representing one or more data points averaged over a period of time. The sensor data may include calibration data, smoothed data, filtered data, transformed data, and / or any other data associated with the sensor.

[0103] As used herein, the term "counts" is a broad term given its ordinary and customary meaning to those of skill in the art (and is not limited to any special or customized meaning) and refers to, but is not limited to, a unit of measurement for a digital signal. In one embodiment, the raw data stream measured in counts is directly related to a voltage (e.g., converted by an A / D converter) that is directly related to a current from the working electrode.

[0104] As used herein, the terms "matched data pair" or "data pair" are broad terms given their ordinary and customary meaning to those of skill in the art (and are not limited to any special or customized meaning), and further refer to, but are not limited to, substantially time-corresponding sensor data (e.g., one or more sensor data points) and matched reference data (e.g., one or more reference analyte data points).

[0105] As used herein, the term "sensor electronics" is a broad term given its ordinary and customary meaning to those skilled in the art (and is not limited to any special or customized meaning) and refers to, but is not limited to, the components of a device (such as hardware and / or software) configured to process data. In the case of an analyte sensor, the data includes biological information obtained by the sensor regarding the concentration of an analyte in a biological fluid. U.S. Patent Nos. 4,757,022, 5,497,772, and 4,787,398 describe suitable electronic circuitry that may be utilized with certain embodiment devices.

[0106] As used herein, the term "calibration" 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 special or customized meaning) and refers to, but is not limited to, the process of determining the relationship between raw sensor data (e.g., analog (nA) or digital units (counts) vs. clinically meaningful units (e.g., mg / dL or mmol / L of glucose)). In some embodiments, the process involves determining the relationship by pairing the sensor data with corresponding reference data, although other calibration techniques (that do not pair sensor data with time-corresponding reference data) may also be used. In some embodiments, i.e., for continuous analyte sensors, the calibration may be updated or recalibrated over time (e.g., once or more than once a day or when reference data is provided by the host) as changes in the relationship between the sensor data and the reference data occur due to, for example, changes in sensitivity, baseline, transport, metabolism, etc.

[0107] As used herein, the terms "calibrated data" and "calibrated data stream" 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 special or customized meaning), and further refer to, but are not limited to, data that has been transformed from a raw state to another state using a function, e.g., a transformation function (also called a modification function), to provide a meaningful value to a user.

[0108] As used herein, the term "calibration set" 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 special or customized meaning) and refers to a set of data that includes information useful for calibration, but is not limited to such a set. In some embodiments, a calibration set is formed from one or more matched data pairs that are used to determine a relationship between reference data and sensor data. However, other data derived pre-embedding, either externally or internally, may also be used.

[0109] As used herein, the term "sensitivity" or "sensor sensitivity" is a broad term and is to be given its ordinary and customary meaning to one of skill in the art (and is not to be limited to any special or customized meaning), and refers to, but is not limited to, the amount of signal generated by a particular concentration of a measured analyte or a measured species (e.g., HO) associated with the measured analyte (e.g., glucose). For example, in one embodiment, the sensor has a sensitivity of about 1 to about 300 picoamps of current per mg / dL of glucose analyte.

[0110] As used herein, the terms "sensitivity profile" and "sensitivity curve" 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 special or customized meaning), and refer to, but are not limited to, a representation of change in sensitivity over time.

[0111] overview Conventional in vivo continuous analyte sensing techniques have typically relied on reference measurements performed during a sensor session to calibrate the continuous analyte sensor. The reference measurements are matched with substantially time-corresponding sensor data to create matched data pairs. Regression is then performed on some or all of the matched data pairs (e.g., by using least squares regression) to generate a transformation function that defines the relationship between the sensor signal and the estimated glucose concentration.

[0112] In critical care settings, calibration of continuous analyte sensors is often performed by using a calibration solution with a known concentration of analyte as a reference. This calibration procedure can be cumbersome since a calibration bag separate from (and in addition to) the IV bag is typically used. In ambulatory environments, calibration of continuous analyte sensors is traditionally performed by capillary blood glucose measurements (e.g., fingerstick glucose tests) through which reference data is obtained and input into the continuous analyte sensor system. This calibration procedure typically involves frequent fingerstick measurements, which can be inconvenient and painful.

[0113] In certain embodiments, a continuous analyte sensor includes one or more working electrodes and one or more reference and / or counter electrodes that operate together to measure a signal associated with the concentration of an analyte in a host. The output signal from the working electrode is typically a raw data stream that is calibrated, processed, and used to generate an estimated analyte (e.g., glucose) concentration. In certain embodiments, a continuous analyte sensor may measure additional signals associated with the baseline and / or sensitivity of the sensor, thereby allowing monitoring of the baseline and / or additional monitoring of sensitivity changes or drift that may occur over time in a continuous analyte sensor.

[0114] In some embodiments, the sensor extends through a housing, which maintains the sensor on the skin and provides an electrical connection of the sensor to the sensor electronics. In one embodiment, the sensor is formed from a wire. For example, the sensor may include an elongated conductive body, such as a bare elongated conductive core (e.g., metal wire) or an elongated conductive core coated with one, two, three, four, five, or more layers of material, each of which may or may not be conductive. The elongated sensor 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. Other embodiments of elongated conductive bodies are disclosed in U.S. Patent Application Publication No. 2011 / 0027127, which is incorporated herein by reference in its entirety. Preferably, the membrane system is deposited over at least a portion of the electroactive surfaces of the sensor (including the working electrode and optionally the reference electrode) to provide protection of exposed electrode surfaces from the biological environment, diffusion resistance (limitation) of analytes as needed, catalyst to enable enzymatic reactions, limit or block interferents, and / or hydrophilicity at the electrochemically reactive surfaces of the sensor interface. Disclosures regarding different membrane systems that may be used with the embodiments described herein are described in U.S. Patent Application Publication No. 2009-0247856 A1, which is incorporated herein by reference in its entirety. In addition to wire-based sensors, the systems and methods described herein may be applicable to other invasive, minimally invasive, or non-invasive sensor technologies, such as planar substrate-based sensor technologies.

[0115] Calibrating sensor data from a continuous analyte sensor generally involves defining a relationship between the sensor-generated measurements (e.g., in units of nA or digital counts after A / D conversion) and clinically relevant values ​​(e.g., in units of mg / dL or mmol / L). In certain embodiments, one or more reference measurements taken immediately after the analyte sensor is manufactured and before the sensor is used for calibration. The reference measurements may be taken in many forms. For example, in certain embodiments, the reference measurements may be determined from a ratio or correlation between the sensitivity of a sensor (e.g., from a particular sensor lot) for in vivo analyte concentration measurements over a particular time period and the sensitivity of another sensor (e.g., from the same lot made in substantially the same manner under substantially the same conditions) for in vitro analyte concentration measurements. By providing a continuous analyte sensor with a predetermined in vivo-to-in vitro ratio and a predetermined sensitivity profile (as described in more detail elsewhere herein), self-calibration of the sensor may be achieved along with a high level of sensor accuracy.

[0116] Determining Sensor Sensitivity In certain embodiments, self-calibration of an analyte sensor system may be performed by determining the sensor sensitivity based on the sensitivity profile (and measured or estimated baseline) such that the equation y=mx+b can be solved, where y represents the sensor signal (counts), x represents the estimated glucose concentration (mg / dL), m represents the sensor sensitivity to the analyte (counts / mg / dL), and b represents the baseline signal (counts). From this equation, a transformation function can be formed, which converts the sensor signal to an estimated glucose concentration.

[0117] It has been found that the sensitivity of a sensor to an analyte concentration during a sensor session often changes or drifts as a function of time. FIG. 1A illustrates this phenomenon, providing a plot of the sensor sensitivity 110 of a group of consecutive glucose sensors as a function of time during a sensor session. FIG. 1B provides three plots of the transformation function at three different periods of the sensor session. As shown in FIG. 1B, the three transformation functions have different slopes, each of which corresponds to a different sensor sensitivity. Thus, the difference in slope over time indicates that a change or drift in sensor sensitivity occurs over the sensor session.

[0118] Referring back to the study associated with FIG. 1A, the sensors were made in substantially the same manner under substantially the same conditions. The sensor sensitivities associated with the y-axis of the plot are expressed as a percentage of the substantially steady-state sensitivity reached about 3 days after the start of the sensor session. In addition, these sensor sensitivities correspond to measurements obtained from a standard reference glucometer (e.g., manufactured by Yellow Springs Instruments). As shown in the plot, the measured sensitivity of each sensor (expressed as a percentage of the steady-state sensitivity) is very close to the sensitivity of the other sensors in the group at any given time of the sensor session. Without being bound by theory, it is believed that the observed upward trend in sensitivity (over time), which is particularly noticeable in the initial part of the sensor session, may be due to conditioning and hydration of the sensing area of ​​the working electrode. It is also believed that the glucose concentration of the fluid surrounding the glucose sensor continuously during the sensor startup may also affect the sensitivity drift.

[0119] With the sensors tested in this study, the change in sensor sensitivity over time defined by a sensor session (expressed as a percentage of substantially steady-state sensitivity) resembled a logarithmic growth curve. It should be understood that other continuous analyte sensors fabricated using different techniques, with different specifications (e.g., different film thicknesses or compositions), or under different manufacturing conditions may exhibit different sensor sensitivity profiles (e.g., those associated with a linear function). Nevertheless, with improved control over the operating conditions of the sensor fabrication process, a high level of reproducibility has been achieved such that the sensitivity profiles exhibited by individual sensors of a sensor population (e.g., a sensor lot) are substantially similar, and sometimes nearly identical.

[0120] It has been discovered that the change or drift in sensitivity over a sensor session is not only substantially consistent between sensors manufactured in a substantially similar manner under substantially the same conditions, but modeling can be performed via a mathematical function that can accurately estimate this change or drift. As shown in FIG. 1A, an estimation algorithm function 120 can be used to define the relationship between time during a sensor session and sensor sensitivity. The estimation algorithm function can be generated by testing a sample set (including one or more sensors) from a sensor lot under in vivo and / or in vitro conditions. Alternatively, the estimation algorithm function can be generated by testing each sensor under in vivo and / or in vitro conditions.

[0121] In some embodiments, the sensor may be subjected to an in vitro sensor sensitivity drift test, in which the sensor is exposed to changing conditions (e.g., step changes in glucose concentration in a solution) and an in vitro sensitivity profile of the sensor is generated over a period of time. The period of the test may correspond substantially to the entire sensor session of the corresponding in vivo sensor, or may include a portion of the sensor session (e.g., the first day, first two days, first three days, etc. of the sensor session). It is envisioned that the above-mentioned test may be performed on each individual sensor, or alternatively, on one or more sample sensors of a sensor lot. From this test, an in vitro sensitivity profile may be created, from which an in vivo sensitivity profile may be modeled and / or formed.

[0122] From the in vivo or in vitro test, one or more data sets including matched data pairs, each relating sensitivity to time, can be generated and plotted. A sensitivity profile or curve can then be fitted to the matched data pairs. If the curve fit is determined to be satisfactory (e.g., the standard deviation of the generated matched data pairs is below a certain threshold), the sensor sensitivity profile or curve can be deemed to have passed quality control and suitable for release. From there, the sensor sensitivity profile can be transformed into an estimated algorithmic function, or alternatively, into a lookup table. The algorithmic function or lookup table can be stored, for example, in a computer-readable memory and accessed by a computer processor.

[0123] The estimation algorithm function may be formed by applying curve fitting techniques that recursively fit a curve to the matched data pairs by adjusting the function (e.g., by adjusting the parameters of the function) until an optimal fit to the available matched data pairs is obtained. In simple terms, a "curve" (e.g., a function that may also be referred to as a "model") is fitted to generate a curve that relates one data value to one or more other data values, and the parameters of the curve are selected such that the curve estimates the relationship between the data values. By way of example, the selection of the parameters of the curve may include the selection of coefficients of a polynomial function. In some embodiments, the curve fitting process may include evaluating how closely the curve determined in the curve fitting process estimates the relationship between the data values ​​to determine the optimal fit. As used herein, the term "curve" 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 special or customized meaning) and refers to a function or a graph of the function, which may include, but is not limited to, rounded curves or straight lines, e.g., lines.

[0124] The curve may be formed by any of a variety of curve fitting techniques, such as, for example, linear least squares fitting, nonlinear least squares fitting, Nelder-Mead simplex, Levenberg-Marquardt, and variations thereof. Additionally, the curve may be fitted using any of a variety of functions, including, but not limited to, linear (including constant functions), logarithmic, quadratic, cubic, square root, power, polynomial, rational, exponential, sinusoidal, and variations and combinations thereof. For example, in some embodiments, the estimation algorithm includes a linear function component that is given a first weight w1, a logarithmic function component that is given a second weight w2, and an exponential function component that is given a third weight w3. In further embodiments, the weights associated with each component may vary as a function of time and / or other parameters, while in alternative embodiments, one or more of these weights are constant as a function of time.

[0125] In certain embodiments, the correlation (e.g., R value) of the estimation algorithm function is a measure of the quality of the curve fit to the matched data pair with respect to the data obtained from the sample sensor and may be one metric used to determine if the function is optimal. In some embodiments, the estimation algorithm function formed from the curve fitting analysis may be adjusted to account for other parameters, such as other parameters that may affect or provide additional information regarding the sensor sensitivity. For example, the estimation algorithm function may be adjusted to account for the sensitivity of the sensor to hydrogen peroxide or other chemical species.

[0126] The estimation algorithms formed and used to accurately estimate the sensitivity of an individual sensor at any time during a sensor session can be based on factory calibration and / or based on a single early reference measurement (e.g., using a single-point blood glucose monitor). In some embodiments, sensors across a population of consecutive analyte sensors manufactured in substantially the same manner under substantially the same conditions exhibit a substantially fixed in vivo vs. in vitro sensitivity relationship. For example, in one embodiment, the in vivo sensitivity of a sensor at a certain time after the start of sensor use (e.g., t=about 5, 10, 15, 30, 60, 120, or 180 minutes after sensor use) is consistently equal to the measured in vitro sensitivity of the sensor or a comparable sensor. From this relationship, an initial value of the in vivo sensitivity can be generated, from which an algorithmic function corresponding to the sensor sensitivity profile can be formed. In other words, from this initial value (representing one point in the sensor sensitivity profile), the remainder of the entire sensor sensitivity profile can be determined and plotted. The initial value of the in vivo sensitivity can be associated with any portion of the sensor sensitivity profile. In certain embodiments, multiple initial values ​​of in vivo sensitivity that are separated in time and correspond to multiple in vitro sensitivity may be calculated and combined together to generate a sensor sensitivity profile.

[0127] In some embodiments, as shown in FIG. 2A, the initial value 210 of in vivo sensitivity is associated with a time corresponding to (near) the beginning of the sensor session. Based on this initial value 210, the remainder of the sensor sensitivity profile 220 is plotted (e.g., plotted back and forth across an x-axis corresponding to time), as shown in FIG. 2B. However, as shown in FIG. 2C, in some embodiments, the initial value 210' may be associated with any other time of the sensor session. For example, as shown in FIG. 2C, in one embodiment, the initial value 210' of in vivo sensitivity is associated with a time at which the sensitivity has reached a substantially steady state (e.g., about day 3). From the initial value 210', the remainder of the sensor sensitivity profile 220' is plotted, as shown in FIG. 2D.

[0128] In other embodiments, the in vivo to in vitro sensitivity relationships were not equal, but the relationship nevertheless included a consistently fixed relationship or function. By having a substantially fixed in vivo to in vitro sensitivity relationship, some of the sensors described herein may be factory calibrated by evaluating the in vitro sensitivity characteristics of sensors from a particular sensor lot at a manufacturing facility (e.g., one or more sensitivity values ​​measured over a particular time period), defining the in vivo sensitivity characteristics of other sensors in the same sensor lot based on the relationship to the measured in vitro sensitivity characteristics, and storing this calculated in vivo sensitivity characteristic in electronics associated with the sensor (e.g., a computer memory of the sensor electronics, further described elsewhere herein, configured to be operatively coupled to the sensor during sensor use).

[0129] Thus, using information acquired prior to a sensor session related to the in vivo vs. in vitro sensor sensitivity relationship and a given sensor sensitivity profile, factory calibration is achieved with a high level of sensor accuracy. For example, in some embodiments, the sensor was able to achieve an accuracy corresponding to an average absolute relative difference of about 10% or less over a sensor session of at least about 3 days, and sometimes at least about 4, 5, 6, 7, or 10 days. In some embodiments, the sensor was able to achieve an accuracy corresponding to an average absolute relative difference of about 7%, 5%, or 3% or less over a sensor session of at least about 3 days. Factory calibration may eliminate the need for recalibration or may otherwise be required only in certain circumstances, such as in response to detection of a sensor failure.

[0130] Referring back to the study associated with FIG. 1A, sensors were constructed with a working electrode configured to measure a glucose + baseline signal and a corresponding auxiliary electrode configured to measure only the baseline signal. The sensor electronics in the sensor system subtracted the baseline signal from the glucose + baseline signal to obtain a signal that is entirely or substantially entirely associated with the glucose concentration. In addition, an algorithmic function was generated and stored in the sensor electronics associated with the sensors to estimate the sensitivity over the lifetime of these sensors. This algorithmic function is plotted in FIG. 1A and shown closely overlaid on the measured sensor sensitivity of the sensor. The determination of the baseline and sensitivity at any given time during the life of the sensor forms a transformation function, which converts the sensor signal to an estimated glucose concentration.

[0131] Although it has been found that individual sensors of a population of sensors manufactured under substantially identical conditions generally exhibit substantially similar or nearly identical sensor sensitivity profiles and have substantially similar or nearly identical in vivo to in vitro sensor sensitivity relationships, it has sometimes been found that actual sensor sensitivity (e.g., sensitivity expressed as an actual sensitivity value, rather than as a percentage of substantially steady-state sensitivity) can vary between sensors. For example, even if individual sensors are manufactured under substantially identical conditions, they may have different sensitivity characteristics during use if they are exposed to different environmental conditions (e.g., exposure to radiation, extreme temperatures, abnormal dehydration conditions, or any environment that may damage enzymes in the sensor membrane or other parts of the sensor) during the period between sensor fabrication and sensor use.

[0132] Thus, to compensate for potential effects resulting from these conditions, in certain embodiments, the continuous analyte sensor is configured to request and accept one or more reference measurements (e.g., from a fingerstick glucose measurement or from a calibration solution) at the beginning of a sensor session. For example, the request for one or more reference measurements may be made approximately 15 minutes, 30 minutes, 45 minutes, 1 hour, 2 hours, 3 hours, etc. after activation of the sensor. In some embodiments, the sensor electronics are configured to process and use the reference data to generate (or adjust) a sensor sensitivity profile in response to input of one or more reference measurements to the sensor. For example, if a reference measurement of glucose concentration is taken and input to the sensor at time=x, an algorithmic function of the sensor sensitivity may be generated by matching the sensor sensitivity profile at time=x to the reference measurement. The use of one of the one or more reference measurements at the start of the sensor in conjunction with a predefined sensor sensitivity profile allows for self-calibration of the sensor without or reducing the need for further reference measurements.

[0133] FIG. 3 is a diagram illustrating different types of information that may be input to a sensor system to define a sensor sensitivity profile over time, in one embodiment. The input information may include information acquired prior to the sensor session 310 and information acquired during the sensor session 320. While in the embodiment shown in FIG. 3, both information acquired prior to the sensor session 310 and information acquired during the sensor session 320 are used to generate, adjust, or update a function associated with the sensor sensitivity profile 330, in another embodiment, the sensor system may be configured to use only information acquired prior to the sensor session. In certain embodiments, the formation of an initial sensor sensitivity profile may occur prior to the sensor session, at the beginning of the sensor session, or immediately after the beginning of the sensor session. Additionally, in some embodiments, the sensor sensitivity profile may be continuously adjusted, regenerated, or updated to account for parameters that may affect the sensor sensitivity or provide additional information regarding the sensor sensitivity during the sensor session. Information acquired prior to the sensor session may include, for example, a sensor sensitivity profile generated prior to or at the beginning of the sensor session, as described above. It may also include sensitivity values ​​associated with a substantially fixed in vivo to in vitro sensor sensitivity relationship, as described above.

[0134] Alternatively, instead of a fixed sensitivity value, the in vivo vs. in vitro sensor sensitivity relationship may be defined as a function of time between the completion of sensor fabrication (or the time a calibration check was performed on sensors from the same lot) and the start of a sensor session. As shown in FIG. 4, it has been discovered that the sensitivity of a sensor to an analyte concentration may change as a function of time between the completion of sensor fabrication and the start of a sensor session. FIG. 4 illustrates this phenomenon through a plot that resembles a decreasing trend in sensitivity over time between the completion of sensor fabrication and the start of a sensor session. Similar to the discovered change or drift in sensitivity over time of a sensor session, this change or drift in sensitivity over time between the completion of sensor fabrication and the start of a sensor session is generally consistent between sensors that were not only manufactured under substantially the same conditions and in substantially the same manner, but also avoided exposure to certain conditions (e.g., exposure to radiation, extreme temperatures, abnormal dehydration conditions, or any environment that may damage enzymes in the sensor membrane or other parts of the sensor, etc.). Thus, the change or drift in sensitivity over time between the completion of sensor fabrication and the start of a sensor session may also be modeled through a mathematical function 420 that accurately estimates this change or drift. The estimation algorithm function 420 may be any of a variety of functions, such as, for example, linear functions (including constant functions), logarithmic functions, quadratic functions, cubic functions, square root functions, power functions, polynomial functions, rational functions, exponential functions, sine functions, and combinations thereof.

[0135] The information acquired prior to the sensor session may also include information regarding the characteristics or properties of a particular sensor. By way of example and not limitation, the information acquired prior to the sensor session may include the particular material used to fabricate the sensor (e.g., the material used to form the sensor membrane), the thickness of the sensor membrane, the permeability of the membrane to glucose or other chemical species, an in vivo or in vitro sensor sensitivity profile of another sensor made in substantially the same manner under substantially the same conditions, and the like. In certain embodiments, the information acquired prior to the sensor session may include information regarding the process conditions under which the sensor is fabricated. This information may include, for example, the temperature at which the sensor membrane was cured, the length of time the sensor was immersed in a particular coating solution, and the like. In other embodiments, the information acquired prior to the sensor session may relate to physiological information of the patient. For example, the patient's age, body mass index, sex, and / or historical patient sensitivity profile may be used as parameters to form the sensor sensitivity profile. Other information acquired prior to the sensor session that may also be used includes information regarding sensor insertion, such as, for example, the location (e.g., abdominal vs. dorsal) or depth of sensor insertion.

[0136] In general, the sensor sensitivity function may be developed by theoretical or empirical methods, or both, and stored as a function or look-up table, thereby enabling sensor self-calibration that eliminates (or substantially reduces) the need for reference measurements. The sensor sensitivity function may be generated at a manufacturing facility and shipped with the system, or may be generated by the system immediately prior to (or during) use.

[0137] calibration An exemplary calibration process according to some embodiments will now be described with reference to Figure 5. Figure 5 illustrates an exemplary calibration process 500 that may use one or more of pre-implantation information 502, internal diagnostic information 504, and external reference information 506 as inputs to form or modify a transformation function 508. The transformation function 508 may be used to convert the sensor data (e.g., in units of current or counts) to an estimated analyte value 510 (e.g., in units of analyte concentration). Information representing the estimated analyte value may then be output 512, such as displayed on a user display, transmitted to an external device (e.g., an insulin pump, a PC computer, a mobile computing device, etc.), and / or otherwise further processed. The analyte may be, for example, glucose.

[0138] In process 500, pre-embedded information 502 may refer to information generated prior to implantation of the sensor currently being calibrated. Pre-embedded information 502 may include any of the following types of information: A priori information including sensitivity values ​​and ranges from in vitro or in vivo testing, a predefined sensitivity profile associated with a currently used (e.g., implanted) sensor, such as a predicted profile of the sensor's sensitivity change over time; a previously determined relationship between a particular stimulation signal output (e.g., an output indicative of the impedance, capacitance, or other electrical or chemical property of the sensor) and sensor sensitivity (e.g., determined from previous in vivo and / or ex vivo studies), such as described in U.S. Patent Application Publication No. 2012-0265035, which is incorporated herein by reference in its entirety; A previously determined relationship between a particular stimulation signal output (e.g., an output indicative of the impedance, capacitance, or other electrical or chemical property of the sensor) and sensor sensitivity (e.g., determined from previous in vivo and / or ex vivo studies); Sensor data obtained from previously implanted analyte concentration sensors, such as sensors from the same lot and / or one or more different lots of the sensor being calibrated; the calibration code associated with the sensor being calibrated; Patient-specific relationships between sensors and sensitivity, baseline, drift, impedance, impedance / temperature relationships (e.g., determined from previous studies of the patient or other patients with characteristics in common with the patient); The specific relationship of the sensor implantation site (abdomen, arm, etc.) (different sites may have different vascular densities); Time since sensor manufacture (e.g., the time the sensor has been stored, the date the sensor was manufactured and / or shipped, the time between when the sensor was manufactured and / or shipped and when the sensor was implanted), Exposure of the sensor during storage to temperature, humidity and external factors.

[0139] In process 500, internal diagnostic information 504 may refer to information generated by the sensor system in which the implanted analyte sensor (whose data is being calibrated) is used. Internal diagnostic information 504 may include any of the following types of information: a stimulation signal output of the sensor (e.g., the output may be indicative of the impedance of the sensor) using any of the stimulation signal techniques described herein (the stimulation signal output may be acquired and processed in real time); sensor data measured by the implanted sensor (real-time data and / or sensor data previously generated using the currently implanted sensor) indicating an analyte concentration; Temperature measurement using an implanted sensor or an auxiliary sensor (such as a thermistor) co-located with or separate from the implanted analyte sensor; Sensor data from multi-electrode sensors (e.g., where one electrode of the sensor is designed to determine the baseline signal), sensor data generated by redundant sensors, where one or more of the redundant sensors are designed to be substantially identical to at least some, if not all, of the other redundant sensors (e.g., having the same sensor membrane type); Sensor data generated by one or more auxiliary sensors having a different modality (e.g., optical, thermal, capacitive, etc.) that are co-located with the analyte sensor or located remotely from the analyte sensor; The amount of time since the sensor has been embedded and / or connected (e.g., physically or electronically) to the sensor electronics of the sensor system; Data representing the pressure on a sensor / sensor system, e.g. generated by a pressure sensor (e.g. to detect compression artifacts), Data generated by the accelerometer (e.g., indicative of the host's movement / motion / activity); Measurement of moisture ingress (e.g., to indicate the integrity of the moisture seal in the sensor system), A measure of the noise in the analyte concentration signal (which in some embodiments can be referred to as the residual between the raw signal and the filtered signal).

[0140] In process 500, external reference information 506 may refer to information generated from a source while the implanted analyte sensor (for which the data is being calibrated) is in use. External reference information 506 may include any of the following types of information: real-time and / or prior analyte concentration information obtained from a reference monitor (e.g., analyte concentration values ​​obtained from a separate sensor such as a fingerstick glucose meter); Reference meter type / brand (different meters may have different bias / accuracies), information indicating the carbohydrates consumed by the patient, Information from drug pens / pumps such as on-board insulin, insulin sensitivity, on-board glucagon, etc. Glucagon sensitivity information, Information gathered from population-based data (e.g., based on data collected from sensors with similar characteristics, such as sensors from the same lot).

[0141] Detect outliers A useful feature of calibration is the ability to determine the presence of outlier data points or outliers in the calibration information (e.g., matched data pairs in a calibration set). As can be easily understood, the inclusion of insufficient data points, such as reference glucose outliers, can adversely affect any calibration, for example, introducing bias or error into the calibrated sensor data that stems from errors from the calibration information. For example, an erroneous finger stick is an obvious cause of a poor reference data point. An erroneous finger stick can result from different sources, including user error and sensor BG mismatch.

[0142] Examples of user error include incorrect input into the receiver (e.g., 109 mg / dl vs. 190 mg / dl) and poor fingerstick technique (e.g., "sugar on finger"). Examples of sensor BG discrepancies include time lag and sensor not tracking (e.g., BG trending up, sensor trending down).

[0143] Incorrect calibration can result in poorly drawn calibration lines and introduce errors into the transformation functions used to convert sensor-generated data into clinical units. Thus, the ability to identify and / or remove or reduce the presence of outliers ensures that carried-forward calibration sets are more accurate and therefore result in more accurate calibration lines.

[0144] 6 is a flow chart 600 illustrating a process for dynamically and intelligently determining or detecting outlier data points according to one embodiment of the present disclosure. External BG reference data (e.g., blood glucose values ​​from a finger stick check) may or may not be useful in calibration depending on the quantity error associated therewith and / or the reason for any error or discrepancy between the BG reference data and the time-corresponding sensor data. As explained above, detecting outlier data points is highly desirable as the exclusion of such data points can minimize and / or prevent errors in the sensor calibration, making the calibration more reliable.

[0145] At block 610, the processor circuit may be configured to iteratively evaluate multiple subsets of data in the calibration data set. In some embodiments, the processor circuit determines a best data subset from the calibration data set. Generally, the iterative evaluation may include iteratively evaluating different calibration data from a single calibration set during a particular calibration, or iteratively evaluating different calibration data from an aggregate calibration set at different calibrations, thereby identifying trends in errors.

[0146] In some embodiments, the subset may be at least 1 / N of the matched data pairs, or at least 0.5 * The subset may be between N and N-1, where N is the number of matched data pairs in the full set. In some embodiments, the subset may be at least 1 / 2 the matched data pairs. In some embodiments, the calibration data set includes at least 5 matched data pairs. However, some calibration data sets include 10 or more matched data pairs. As used herein, a data point in a calibration data set may include a matched data pair that includes a reference data value (e.g., BG value) and a sensor data value (e.g., count).

[0147] In some embodiments, it may be advantageous to reduce the number of subsets evaluated to improve computational efficiency. For example, heuristic methods may be applied by using prior information (e.g., information from previous outlier detection evaluations) to better identify the best data pair (or subset) and / or the worst data pair (or subset). In other words, trends or history associated with each matched data pair or subset may be used to limit the iterative evaluation of the subsets. Advantageously, the heuristic method for determining which subset to evaluate takes advantage of the fact that matched pairs in the subsets used in previous iterations represent points close to the regression line. This is expected to yield good results despite a reduction in the number of subsets examined. One heuristic method involves "pre-screening" the latest BG values ​​by comparing the newly formed matched data pairs with the best subsets determined from previous iterations of outlier detection. This may allow immediate decision making to be made before a full iterative process.

[0148] In some embodiments, matched data pairs may be assigned weights based on the likelihood of being outliers. The processor circuitry may use any subset and calculate the residuals, and then each point may be assigned a weight based on how far the point is from the line (e.g., the residuals). These weights may be combined to indicate how likely the point is to be an outlier. For example, the smaller the weight, the better the point matches with the rest of the matched pair, and vice versa. For example, the weight may be the average of the squared residuals obtained from all subsets for which it was evaluated. In some embodiments, the processor circuitry may use only the subsets with the lowest weights (e.g., below a threshold) to effectively select the points that are best correlated with most other points.

[0149] In some embodiments, each subset may be used to draw a line (e.g., using regression analysis). In some embodiments, multiple lines associated with the dataset are generated to determine the best subset. For example, generating multiple lines may result in a best line associated with the best subset. In some embodiments, the best line includes the line that has the best fit to all of the points used to generate the line. The best fit may be determined using one or more functions selected from the following list: For example, regression techniques useful for determining statistical relevance of datasets may be used, such as linear regression (e.g., Least Quantile of Squares (LQS), Least Trimmed of Squares (LTS)), non-linear regression, rank correlation, least mean squared fit, mean absolute deviation, and mean absolute relative difference.

[0150] In some embodiments, each subset defines a convex hull or convex envelope. In some embodiments, multiple convex envelopes associated with the dataset are generated to determine the best subset. For example, generating multiple convex envelopes may result in a best convex envelope associated with the best subset. In some embodiments, the best convex envelope includes the convex envelope that has the best fit for all points used to generate the convex envelope. The best fit may be determined using any standard algorithm for convex hull computation, such as one or more functions selected from the following list: Convex hull algorithms such as Graham scan (e.g., O(NlogN) complexity), QuickHull (e.g., O(NlogN) average complexity), and Monotone chain (e.g., O(NlogN)), and / or a rotating caliper algorithm for computing the diameter and width of the convex hull (e.g., the convex hull is computed (O(N) time).

[0151] In some embodiments, it is desirable to use the convex hull according to a first heuristic to reduce the number of subsets, since any outliers or outlier data points lie on the convex hull (e.g., are part of the line that defines the convex hull). Thus, when considering the subsets, a convex hull (taking NlogN times) may be constructed, and only the subsets that are completely inside the convex hull may be used to draw the calibration line. Once the calibration line is drawn, the points on the hull may be included in the subsets one by one, using each time the point that is closest to the current line. If, after including a sufficient number of points, any of the remaining points on the convex hull are found to be far from the line, such points may be picked out as outliers.

[0152] In block 620, the processor circuitry may be configured to identify a boundary or confidence interval associated with the best subset. In some embodiments, the boundary or confidence interval may be a boundary line of the best convex envelope.

[0153] In some embodiments, the boundary or confidence interval of the step may be an acceptable deviation or scatter from the best line. For example, each point used to generate the best line may be compared to the best line to determine a residual value for each point (e.g., the most accurate point relative to the best line). In some embodiments, the boundary or confidence interval is a predetermined multiple of a scale. The scale is defined herein as the average value of the residuals multiplied by some consistency factor. The boundary interval may then be defined as the average residual multiplied by a consistency factor (e.g., multiplied by a factor such as 2). In other embodiments, the boundary or confidence interval is a set value, e.g., pre-set as part of a factory setting. In some embodiments, the boundary or confidence interval is calculated and updated in real time. The residual value of each point may then be compared to the boundary or confidence interval to determine whether the residual value of each point falls within the boundary or confidence interval.

[0154] In some embodiments, the diameter and width of the convex hull may be calculated using a second heuristic. As used herein, the diameter is the maximum distance between two points in the convex hull, and the width is the minimum distance between two parallel support lines. In general, if the diameter is much larger than the width, the points in the set will cluster around the line. In some embodiments, the ratio of the area of ​​the convex hull to the diameter may be compared. In some embodiments, iterating over the points on the convex hull and finding the residual with respect to the diameter may identify any point outside the given residual as an outlier.

[0155] At block 630, the processor circuitry may be configured to identify all values ​​(e.g., data points) outside the boundaries or confidence intervals as possible outliers. For example, if any residual values ​​are outside the boundaries or confidence intervals, then each point associated with a residual value that is outside the boundaries or confidence intervals may be identified as a possible outlier.

[0156] Several examples of methods for implementing blocks 610-630 are provided below.

[0157] Example 1: Least quartile of squares ("LOS") In one exemplary embodiment, the Least Quartile of Squares ("LQS") or Least Median of Squares ("LMS") robust regression technique allows for the identification of 0% to 50% contaminated data, meaning that outliers can be detected when up to 50% of the data is contaminated. For example, the highest destructive value LQS is 50%, since it is at this point that good data cannot be distinguished from bad data. LQS can be thought of as a sampling algorithm (e.g., trying all permutations in the calibration data set and drawing a line). Ultimately, LQS operates by minimizing the residual around a desired quantile.

[0158] Example 2: Minimum Square Cropping ("LTS") In another exemplary embodiment, the least trimmed squares ("LTS") is a robust statistical method for fitting a function to a set of data without being affected by outliers. In some exemplary embodiments, a modified version such as the FAST-LTS algorithm may be used. The function first calculates all possible combinations of lines based on a subset of the calibration set, as shown by the following equation [1] (e.g., if the calibration set has 6 points, one of the lines can be drawn using points 1, 3, 4, 6).

[0159] [Number]

[0160] If there are 6 points in the calibration set, the number # of lines drawn is 15. In some exemplary embodiments, the FAST_LTS algorithm starts with all combinations of 2 points and then adjusts the line to include h points. In some embodiments, the FAST_LTS algorithm may start with all combinations of h points. In some embodiments, when the calibration data is sparse, there is no advantage in terms of computational efficiency between starting with 2 points or h points. Note that h can be any number as long as 0.5 * It should be noted that h can be any number as long as N < h < N - 1 is satisfied.

[0161] In some embodiments, the calibration set must have 5 or more points in order to execute the function. However, the input to the function can be a subset of the calibration set (e.g., all points in the last 48 hours if there are at least 5 or more, or the last 5 points if the sensor is out of calibration are input to the device). Once all combinations of h points are calculated, the function loops and calculates the ordinary least squares ("OLS") line using each combination of h points. In other words, even if there are 6 points in the calibration set, only 4 of them are used to calculate the line. An example of a 6-point calibration set (Figure 7) and one of the OLS lines drawn using h points (Figure 8) are provided.

[0162] Still referring to FIG. 8, each line may be bounded by one or more acceptable boundaries for the slope and baseline, which may be static or dynamic. In some embodiments, the boundaries may include confidence intervals. In some embodiments, the acceptable boundaries include upper and lower limits defined by minimum and maximum slope and baseline, against which certain calibration information (e.g., calibration lines and / or match pairs) are compared.

[0163] An estimate of the expected signal may be calculated using all the BG values ​​in the calibration set and the regression line drawn. The residuals in the signal (raw data points) may be calculated and then the absolute value of each of the residuals may be taken and then sorted from smallest to largest. When using the wedge parameters, if there is no line that fits within the wedge parameters, the function ends since no actual line has been drawn. In some embodiments, information related to failure to pass the wedge parameters may be used to drive changes in the limits or wedge parameters used in the processor circuitry or communicated to the system's algorithms (e.g., calibration algorithms) to suggest that some points in the calibration set should be dropped.

[0164] In some embodiments, once the best fit line is selected, the residuals may be calculated, squared, and then sorted from smallest to largest, thereby allowing an estimate of the scale (variance around the line) to be calculated. The scale value is greater than 0, but may be limited by another value greater than 0 (e.g., lower bound, upper bound, or both) to avoid further calculations, as seen below. For example, if the upper bound of the scale is consistently achieved (e.g., scale > upper bound), such information may be communicated to other algorithms of the system (e.g., calibration), as such information may indicate that the calibration set is not sufficiently linear and / or that there is too much scatter in the calibration set that one or more data pairs should be removed. After calculating the scale, an estimate of possible outliers (counts) may be made. If the ratio of the residual to the scale exceeds a threshold (e.g., residual > upper bound), an estimate of possible outliers (counts) may be made. i All points with a scale above 2.25 may be flagged as outliers. One or more levels of scale (e.g., preliminary, intermediate, and final scales) may be calculated and processed as described above. As will be appreciated by those skilled in the art, other diagnostics may be applied to the scales to improve accuracy.

[0165] In some embodiments, a possible outlier may be immediately removed from the calibration set. In some embodiments, a possible outlier may be retroactively removed from the calibration set at a calibration after the calibration that resulted in the flagging. In some other embodiments, the flagging of a possible outlier may signal an algorithm to check the clinical relevance of the error before removing the point from the calibration set. In some other embodiments, only the relevance of the error may be used to remove a point from the calibration set.

[0166] Referring now to block 640, the processor circuitry may be configured to determine clinical relevance by evaluating the relevance of the possible outliers identified in block 630 to identify root causes of errors in the possible outliers, trends in the outlier information, etc., resulting in useful outlier information for further processing (block 650). Although several embodiments for flagging potential outlier points are described above (blocks 610-630), the relevance assessment (640) and response process (650) may be combined with or applied to any method for determining outlier information. While not wishing to be bound by any particular theory, errors that cause the flagging of possible outliers may be caused by a variety of different circumstances. For example, a phenomenon that may be described as "soak and recover" that occurs during the initial implantation of a subcutaneous sensor may be determined. As another example, physiological changes in a compartment of a test sample (e.g., capillary blood vs. interstitial fluid) may be inherently different at a particular time. Additionally, errors may be flagged as outliers based on statistical analysis or residuals as assessed on raw counts, but the errors may not be clinically relevant; for example, errors in blood glucose reference values ​​compared to a regression line (best-fit or calibration line) capture clinical errors.

[0167] In some embodiments, the processor circuit may evaluate a possible outlier only when the most recent BG is determined to be a possible outlier. In some embodiments, the relevance determination includes examining one or more factors from the following list, without limitation: time since sensor implantation, trend of outlier evaluation, amplitude of the data point's error relative to the (best) line (e.g., evaluated in clinically relevant units such as mg / dL or mmol / L), direction of the data point's error relative to the (best) line (e.g., evaluated in clinically relevant units, the direction of error is less than or greater than 0 when the BG reference value is higher or lower than the sensor value, respectively), clinical risk (e.g., static or dynamic risk) of the data at the timestamp of the matched data pair, rate of change of the analyte concentration or derivative of the sensor data associated with the matched data pair, acceleration (or deceleration) of the analyte concentration or second derivative of the sensor data associated with the matched data pair, etc. In some embodiments, the relevance determination evaluates the impact of the possible outlier on the regression line (best or calibration line). For example, sometimes in a regression analysis, data points have a disproportionate impact on the slope of the regression equation. One way to test for the influence of outliers is to calculate the regression equation with and without the outliers. A threshold, such as greater than 25 pA / mg / dL, can be applied to determine if the influence is disproportionate.

[0168] In some embodiments, a relevance check may be performed because a highly correlated calibration set may flag a point as a statistical outlier, but the blood glucose may be, for example, only 5 mg / dl, which is well within the error of a standard blood glucose meter. For example, clinical relevance may be calculated based on estimated glucose values ​​("EGVs"), which may also be referred to as calibrated glucose values, as shown by the following equation [2]:

[0169]

number

[0170] The EGV may then be checked against the matching BG. In some embodiments, if the matching BG is below a predefined threshold, such as ≦75 mg / dl, the error check utilizes a first criterion, such as absolute difference, and if the matching BG is greater than a predefined threshold, such as >75 mg / dl, the error check utilizes a second criterion (different from the first criterion), such as absolute relative difference (ARD). In one exemplary embodiment, if the error check is absolute difference (BG≦75 mg / dl), the error threshold is 20 mg / dl (approximately 3 sigma of the BG error threshold), and if the error check is ARD (BG>75 mg / dl), the error threshold is 25% (approximately 3 sigma of the BG error threshold). Other thresholds and metrics known to those skilled in the art may be applied. As a result of the clinical relevance determination, errors associated with possible outliers may be determined to be clinically irrelevant, and the matching data pair may remain in the calibration set. For example, a point initially identified as a possible outlier (eg, using a statistical analysis of the error in the y-direction) may be re-identified as clean if the calculated error is within a predetermined threshold.

[0171] However, if the outcome of the relevance determination test is a determination that the error is relevant, processing may move to block 650. For example, the relevance determination test may compare a parameter associated with the error described above to a quantitative outlier criterion (e.g., a threshold value), and errors that meet the criteria for a clinically relevant outlier (e.g., outside a predetermined threshold) are flagged in the outlier information.

[0172] In some embodiments, the relevance determination test evaluates additional criteria such as time since implantation. Time since implantation may indicate phenomena known to occur during the life of a sensor. For example, transcutaneous and implantable sensors are affected by in vivo properties and physiological responses in the surrounding tissue. For example, a decrease in sensor accuracy after the sensor is implanted is one common phenomenon commonly observed. This phenomenon is sometimes referred to as a "soaking and recovery" process. Without being bound by theory, it is believed that soaking and recovery is caused by trauma from the insertion of the implantable sensor and possibly from stimulation of nerve bundles near the implanted area, which results in the nerve bundles reducing blood flow to the implanted area. Alternatively or additionally, soaking and recovery may be related to damage to nearby blood vessels, resulting in a vasospastic event. Any local cessation of blood flow in the implanted area for a period of time results in a decrease in the amount of glucose in the area of ​​the sensor. During this time, the sensor becomes less sensitive and is unable to track glucose accurately. Thus, soaking and recovery manifests as a suppressed glucose signal. Suppressed signals from soaking and recovery often appear within the first day after signal implantation, most commonly within the first 12 hours after implantation. Soaking and recovery are typically expected to resolve within 6-8 hours. Identification of soaking and recovery may provide information to the patient, physician, or other user that the implant does not need to be removed because the sensor is only temporarily affected by a short-term physiological response and normal function is likely to return within a few hours. In one embodiment, the combination of suppressed signals (e.g., detection of a downward shift in sensor sensitivity) during a predefined period (e.g., the first 36 hours after implantation) may be used to identify the root cause of the outlier as a soaking and recovery phenomenon.

[0173] Other physiological responses to an implantable sensor may also affect the performance of the implantable sensor. For example, during wound healing and foreign body reactions, the surface of an implantable sensor may be coated with proteins or other biological materials to the extent that the sensor is unable to accurately track blood glucose. This phenomenon is sometimes referred to as "biofouling", and biofouling often manifests as a downward shift in sensor sensitivity over time. Similarly, an implantable sensor may become encapsulated by biological material to the extent that the sensor is unable to provide glucose data and is considered to be effectively at end of life. In some cases, an implantable device may be programmed to correct for errors associated with biofouling and end of life, whereby identification of these phenomena helps provide more accurate glucose data. In one example, a combination of the downward rate of change of sensor sensitivity over a predetermined period of time (e.g., at least 5 days after sensor implantation) may be used to identify the root cause of outliers such as those associated with biofouling or end of life.

[0174] Another physiological effect observed may be referred to as a "compartment effect." This effect results from differences in actual physiological glucose levels in different compartments within the body, for example, differences in actual physiological glucose levels in capillary blood compared to interstitial fluid. Identification of these physiological phenomena also generally provides useful information in determining whether the error associated with the flagged outlier is related to the baseline BG value or sensor data, and may be processed accordingly. Identification may include additional inputs such as time and / or timing from implantation, location and / or amount of insulin injected. In one example, when the rate of change of the host's glucose exceeds a threshold and / or a meal has been recently consumed, and when the outlier does not follow a similar error trend, the processor circuit may identify the root cause as a transient compartment effect.

[0175] Other factors that may be evaluated to identify the root cause of the outlier may include secondary (e.g., redundant) sensors, trend information associated with outliers (outlier information) from previous iterations of outlier detection, etc. Trend information may be particularly useful when data points (matched pairs) are consistently indicative of a particular error mode, perhaps not enough to meet the outlier criteria of evaluation block 630, but perhaps enough based on the trend criteria (exceeding a threshold of x consecutive outlier evaluations, such as 2, 3, 4, 5, 6, 7, or 8). In general, any number of conditions and / or criteria may be evaluated to determine the relevance (e.g., root cause and / or clinical relevance) of a possible outlier.

[0176] In block 650, the processor circuit may be configured to process data in response to the outlier information determined in block 640. In general, matched data pairs flagged as outliers (outlier information) after the relevance determination test (block 640) may undergo additional processing. Outliers may be removed prospectively or retroactively, temporarily or permanently, from the calibration set. Outliers may be processed according to block 650 with or without the relevance determination check of block 640, and / or may be processed iteratively at each outlier detection check. Additionally or alternatively, other data, such as information sent to the insulin delivery device, prompts on a user interface, or other related processing, may be included herein.

[0177] In one example, when successive iterations of outlier detection indicate a trend in sensor error, the processor circuitry may increase the number of reference data requested and / or relied upon for calibration, e.g., when biofouling and / or end of sensor life are detected, as described in more detail elsewhere herein. In another example, when successive iterations of outlier detection indicate a trend in reference blood glucose error, the processor circuitry may decrease the number of reference data and / or not rely on the reference data for sensor calibration, e.g., when the sensor is stable and shows no signs of drift in sensor sensitivity. Such processing may enable a continuous glucose sensor system to switch the degree of reliance (or not) on blood glucose reference measurements based on a measure of accuracy of the sensor data versus the reference blood glucose measurements over time.

[0178] For example, the processor circuit may flag an outlier, but retain the flagged outlier in the calibration data set until the next data point is collected. In such an embodiment, flagging the outlier allows for an iterative process of confirming (or disproving) the outlier based on additional data, such as the next reference data point and the resulting matched data pair, to avoid false positive identification of the outlier and the consequences on the user experience. In other words, once an outlier is flagged, the root cause of the outlier may be associated with either the sensor data point or the reference data point (BG), since the flagged outlier is affected by both the sensor and reference inputs. In some embodiments, once the next reference data point is input (after the flagged outlier), its respective matched data pair may confirm that the sensor performance has changed, and the flagged outlier is indeed more reliable than other matched data pairs in the calibration set, resulting in removal of some or all other data pairs in the calibration set except for the two flagged outliers and / or the most recent matched data pair. On the other hand, the next matched data pair (after flagging of the outlier) may not be identified as an outlier, thereby confirming that the previously flagged outlier should be removed from the calibration set. Criteria for evaluating the next matched data pair with respect to the previously flagged outlier may include using an iterative process as described herein with respect to dynamically and intelligently determining or detecting outlier data points, or other criteria and analyses specific to determining root cause and / or subsequent interpretation of the flagged outlier.

[0179] In one embodiment, if the most recent BG baseline is identified as a possible outlier (block 630), the direction of the error is checked using a clinical relevance test (block 640). For example, if the direction of the error is less than 0 (e.g., the BG baseline is reading a higher value than the sensor value), then in block 650, all points flagged as outliers are removed from the calibration set except for the most recent BG baseline. On the other hand, if the most recent BG baseline is identified as a possible outlier and the direction of the error is checked as positive (e.g., the sensor value is higher than the BG baseline) using a clinical relevance test, no points are removed from the calibration set. Other methods for inclusion and / or exclusion of matched data pairs may be used to complement and / or supplement the outlier detection method described in flowchart 600.

[0180] This embodiment utilizes a clinical relevance test that considers the direction of error associated with a particular BG reference value (e.g., a matched data pair that includes BG), whereby when the most recent BG reads higher than the sensor value, it is assumed that the sensor is tracking well and that the BG is contaminated by user error, e.g., "sugar on the finger," whereby when the sensor value reads a threshold higher than the BG value, it is assumed that the BG is flagged as an outlier due to drift (e.g., change in sensitivity of the sensor over time). In such an embodiment, the calibration set may be screened to remove insufficient data points (e.g., criteria that ensure that at least about 1 / 2, 2 / 3, or 3 / 4 of the matched data pairs in the calibration data set are removed). With a clean calibration set, a calibration line may be drawn using one or more preferred functions. If the system falls out of calibration, a modified process of outlier detection may be implemented. As used herein, "falling out of calibration" refers to a condition in which sensor data is not converted due to unreliability of the calibration information. An exemplary modified outlier detection using a calibration set that does not include the BG value that caused the system to go "out of calibration" may evaluate the following: A) If the most recent BG (BG1) is flagged as an outlier and BG1 is a higher reading than the sensor, the calibration set (CAL1) before BG1 was entered is stored and the system waits for the next BG (BG2) to be entered. When BG2 is entered it is passed to outlier detection with CAL1. If BG2 is flagged as an outlier, CAL1 is cleared and BG1+BG2 becomes the new calibration set (CAL2). If BG2 is clean, BG1 is discarded and CAL1+BG2 becomes the new calibration set (CAL2'); B) BG1 is a lower reading than the sensor, the calibration set is cleared, BG1 is stored and the system waits for a new BG to start a new calibration set.

[0181] In some embodiments, information associated with the flagged outlier may be communicated to and / or used to trigger another algorithm of the system. For example, the outlier may be specifically identified as indicating a downward sensitivity shift that meets a criterion (e.g., based on an analysis of the error between the BG and the corresponding EGV of the matched pair), and such information may be a risk factor indicative of end of life, and may be used as an input to an end of life decision function and / or may initiate an end of life algorithm. As another example, the outlier may be specifically identified as being associated with a "soak and recover" event based on the time since implantation, the host's history with continuous glucose sensing, and / or the like. This information may thus be useful to trigger a pause in glucose data by the processor circuit for a predetermined period of time (e.g., a "soak and recover" known to have a finite duration) and / or to inform the user on a user interface that some implant effect may cause some inaccuracy over the next x minutes or y hours (e.g., 2, 4, 6, 8, 10, 12, 16, or 24 hours).

[0182] In some embodiments, host-device interaction may be a useful input, which, together with knowledge of flagged outliers, may indicate user frustration with the device and / or the host may need additional guidance. In some embodiments, when the host is periodically (e.g., outside of a normal calibration scheme) inputting BG values ​​at a rate that is more than periodic (e.g., more than X BG entries in the last Y minutes, e.g., more than 2, 4, or 6 entries in the last 15, 30, or 60 minutes), it may be an indication that the host is aware of an error and may be attempting to correct the sensor by "feeding" BG values ​​to the device. The resulting process may include trusting the BG values ​​and providing feedback in the form of messages and / or glucose values ​​that indicate to the host that the device is receiving and responding to the BG input. One example may include questioning the host via prompts on the user interface to determine the root cause. Another embodiment may involve relying on user input over outlier detection testing, thereby "unflagging" so-called flagged outliers, either temporarily or permanently, so that the so-called flagged outliers actually affect the estimated glucose value displayed to the user, temporarily or permanently, and appear to more closely align with the user's BG baseline.

[0183] If the most recent BG is not flagged as an outlier, it can be assumed that the outlier detection was successful, the outlier detection line (best fit OLS line) is within the wedge of the line logic function (the line drawing logic and outlier detection may have different wedges), and the system is calibrating and uses the line drawn by the outlier detection.

[0184] In some embodiments, processing responsive to outlier detection determines whether and / or how much of the pre-implant information 502, internal diagnostic information 504, and / or external reference information 506 may be relied upon for any given calibration. In one example where there is a lack of confidence in the reference glucose data resulting from repeated failure of matched data pairs to pass one or more evaluations of the process of flowchart 600 (for previous calibrations), the processor circuit may determine that the calibration should rely solely on the pre-implant information 502 and / or the internal diagnostic information 504. In some embodiments, the use of the pre-implant information 502, internal diagnostic information 504, and / or external reference information 506 may be selected and / or selectively weighted based on the outlier detection process and / or other diagnostic information, such as the internal diagnostic information. In some embodiments, the use of the external reference information may be adaptively determined and applied to the calibration depending on other parameters, such as time since implantation or historical sensor performance for a particular patient. In one example, a sensor session may begin (initial calibration) by relying only on pre-implant information 502 (e.g., a priori sensitivity information), and then rely more or less on the pre-implant information 502 as the sensor session progresses and additional internal diagnostic information 504 and / or external reference information 506 is acquired. In another example, the sensor session may rely at least in part on the external reference information 506 during the sensor session, but rely less and less on the external reference information as the sensor session progresses, with an increasing confidence in the pre-implant calibration information and / or internal diagnostic information being determined as the sensor session progresses. As another example, when pre-implant information in the form of patient history information is available (i.e., because the patient has previously worn the sensor), the sensor may rely more on the pre-implant information during subsequent sensor sessions compared to the first one or several sensor sessions. Other combinations and selective uses of pre-implantation information 502, internal diagnostic information 504, and / or external reference information 506 for calibration of the sensor (i.e., adaptively and / or progressively over a sensor session) may be envisioned as well.

[0185] Referring back to flowchart 600, in one exemplary embodiment, the iterative evaluation may be performed on the entire calibration set, e.g., using linear regression, rather than on a subset of the calibration set. In other words, the end result may be the same (e.g., looking for trends in the errors to ascertain their clinical relevance) whether one iteratively evaluates multiple subsets on a single calibration set during a particular calibration, or iteratively evaluates the entire calibration set (which may be thought of as a subset of all data pairs collected during a sensor session) at different calibrations. For example, a traditional linear regression analysis (e.g., ordinary least squares "OLS") may be used, with homoscedasticity assumed. In other words, the standard deviation of the error terms is constant and independent of the x values ​​(ε~N(0,o+)). OLS may identify outliers through iterations of the errors.

[0186] It should be appreciated that block 610 covers repeated evaluation of a particular matched data pair in different calibrations. Block 620 describes identifying the boundaries or confidence intervals associated with the best subset. In some embodiments, it may be possible and desirable to track (e.g., trended with each calibration evaluation) which data pairs consistently exhibit errors above a threshold based on a regression line (e.g., drawn from the calibration set or a subset of the calibration set). For example, there may not be enough error to know if a data pair should be removed using a single evaluation, but when looking at trends across multiple calibration evaluations (e.g., repeated over one or more days), the data pair may exhibit a consistent trend of error. The boundaries or confidence intervals used to monitor these trends may be the same as or different (e.g., less strict) than the boundaries or confidence intervals associated with other aspects such as flagging possible outliers. For example, at initialization, the system may start counting errors and look for frequency of errors over time. Errors may be calculated by recalculating the EGV for each of the count values ​​for each data pair in the calibration set using the slope (M) and intercept / baseline (B). This can provide a normalized comparison for each data pair in the calibration set. Using the BGs in the calibration set, the error in the calculated EGV can be determined as shown by the following equation [3]:

[0187]

number

[0188] This formula advantageously allows the error to be determined in units of blood glucose for each data pair (hence better clinical relevance of the error). In some embodiments, the relevance may be determined based on the trend of the error (e.g., a data pair identified as a possible outlier 5 out of the last 6 calibrations). In other words, both the frequency of the error and the fact that the error is evaluated after converting the counts (for each data pair) to units of blood glucose may be used to evaluate the relevance (e.g., clinical relevance) of the possible outlier. In this embodiment, when a data pair is tracked with a clinical error (e.g., error in units of blood glucose) above a certain threshold for at least x out of the last y calibrations / assessments, it may be affecting the calibration error in a clinically relevant manner. In some embodiments, this "trend" error may be flagged as an outlier, for example. In some embodiments, outliers or points identified as having some moderate level of error may be flagged). For example, this embodiment qualitatively illustrates that when a particular error is seen frequently (x times within a paste time) with the same data pair, the data pair is removed from the calibration set and a new calibration line (e.g., a dummy version) is tested to determine if it passes other tests for accuracy in the calibration (e.g., wedge parameters). If the new calibration line passes the tests or is as good as or better than the last one (e.g., based on predetermined criteria), the removed data pair is permanently removed (e.g., culled or discarded) from the calibration set.

[0189] End of Life Detection The glucose sensor embodiments 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. To prevent use beyond the useful life, some embodiments notify the user to replace the sensor after it is determined that the sensor should no longer be used. Various methods may be used to determine whether the sensor should no longer be used, such as a predetermined amount of time that has passed since the sensor was first used (e.g., when it was first implanted in the user or when it was first electrically connected to the sensor electronics) or a determination that the sensor is defective (e.g., due to membrane rupture, unstable sensitivity, etc.). Once it is determined that the 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 shutting down the display or ceasing to display new (or real-time) sensor data on the display.

[0190] In some embodiments, a continuous glucose monitor may exhibit signs of progressive sensor decline (PSD) near the end of its life. Signs of PSD may be recognized and total sensor end of life and resulting user safety or inconvenience may be prevented. In some embodiments, the present disclosure describes distinct sensor failure signatures and how they may be reliably recognized and detected.

[0191] 9, a flow chart 700 is shown illustrating a process for dynamically and intelligently determining or detecting the end of life of a sensor in accordance with an embodiment of the present disclosure. As explained above, detecting the end of life of a sensor is desirable since an accurate reading of a user's blood glucose level is important in diabetes monitoring.

[0192] At block 710, the processor circuit may be programmed to evaluate a plurality of risk factor metrics. The risk factor metrics may describe risk factors that indicate whether the continuous analyte sensor has entered an end-of-life state. For example, the processor circuit may execute sensor end-of-life risk factor instructions, algorithms, and / or functions. In general, the processor circuit may be programmed to execute one or more functions to evaluate a plurality of individual risk factor metrics, each of which may provide, in part, an indication of the end of sensor life. In general, EOL symptoms are progressive, e.g., not all symptoms (or episodes) are indicative of sensor failure. Each of the risk factor metrics may be evaluated periodically or intermittently as frequently as sensor data is received (e.g., every 5 minutes), or more intermittently (e.g., every few hours or daily). The risk factor metrics may be determined iteratively, averaged, or trended over time, and the results may be used in subsequent processing. In some embodiments, the evaluation of one or more risk factor metrics may be triggered by another event, such as a trend error in BG (e.g., from outlier detection) that meets one or more criteria.

[0193] In some embodiments, certain risk factor metrics are evaluated more frequently or less frequently than other risk factors. In some embodiments, when one or more risk factor metrics meet one or more predetermined criteria, either a) a process for dynamically and intelligently determining or detecting that the continuous analyte sensor has entered an end-of-life condition (700), b) a process for evaluating one or more other risk factor metrics (710), or c) a process for determining an end-of-life status (720) may be initiated in response thereto.

[0194] In some embodiments, end of life detection may be accomplished using a combination of methods, each of which individually detects an end of life signature or risk factor. A combination of methods or signatures may result in improved specificity (e.g., fewer false positives). It should be understood that an end of life determination method or algorithm may use a combination of risk factors in determining end of life.

[0195] In some embodiments, suitable risk factor metrics may be selected from a list including, but not limited to: number of days the sensor has been used (e.g., implanted), sensor sensitivity, or whether there has been a degradation 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 standards), noise analysis (e.g., EOL noise factors (skewness, kurtosis, and rotation)), duration, magnitude and history, spectral content analysis, pattern recognition), oxygen (e.g., whether there is a concentration and / or predefined oxygen concentration pattern), glucose pattern (e.g., meal characteristics such as mean, variability, peak-to-peak fluctuation, expected vs. unexpected behavior such as after a meal when glucose is not rising as expected), error between baseline BG and EGV sensor values, including the direction of the error (whether BG or EGV is reading high compared to the other), and a measure of the linearity (or lack of linearity) of the sensor. Sensor linearity refers to the consistency of the sensor's sensitivity over a particular measurement range (e.g., 40-400 mg / dL for glucose sensors). For example, if the sensor signal reads low at low BG and high at high BG, versus if the sensor signal reads low at low BG but does not read high at high BG (does not change or increase beyond a certain BG value), linearity can be assumed and non-linearity can be assumed (based on the error between the reference BG value and the EGV sensor value).

[0196] One risk factor metric that may be useful in determining end of life is the number of days the sensor has been used (e.g., implanted). The processor circuitry may be configured to determine how many days the sensor has been used (e.g., implanted). In some embodiments, the number of days the sensor has been used is determined based in part on using initial calibration data, sensor initialization, operative connection of the sensor with the sensor electronics, user input data, etc. In some embodiments, the processor circuitry detects sensor reboots and uses the reboot information in determining the number of days since implantation.

[0197] In some embodiments, when a certain threshold, e.g., a certain number of days, is met, a particular variable associated with the threshold may be automatically used in the end of life function. For example, if the number of days the sensor was used is determined to be at least 4 days, the number of days the sensor was used is automatically used and / or a simple "yes / no" indicator that the threshold has been met is used. In some embodiments, the number of days the sensor was used is automatically used if the number of days the sensor was used is at least 1 / 3 of the number of days the sensor was approved for use. In other embodiments, the number of days the sensor was used is automatically used if the number of days the sensor was used is at least 1 / 2, 2 / 3, or 3 / 4, etc. of the number of days the sensor was approved for use. In some embodiments, the actual number of days the sensor was used is always used in the end of life evaluation. In some embodiments, the end of life evaluation is performed after a predetermined number of days of sensor use.

[0198] Additionally or alternatively, the time elapsed since insertion may be mapped to an end-of-life risk factor value (e.g., likelihood of future recovery or probability of sensor failure) since the longer the sensor has been in use since implantation, the more likely changes in the sensor-tissue interface (e.g., biofouling) are to affect sensor function. The translation of the end-of-life risk factor metric into an indication of whether a continuous analyte sensor is in an end-of-life state is described in more detail herein, for example, with reference to block 720 and / or FIG. 10. In one example, the end-of-life risk factor metric value maps to approximately 1.0 between days 1 and 5, gradually decreasing beyond day 5, reaching 0.5 at day 8, 0.2 at day 10, and approximately 0.1 at day 14. As can be appreciated by one of ordinary skill in the art, other values ​​and thresholds may be used.

[0199] Another risk factor metric that may be useful in determining end of life is whether there has been a decline in sensor sensitivity, or 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 circuitry may be configured to determine whether there has been a decline in signal sensitivity. For example, for some sensors, their sensitivity drifts up or remains relatively flat over a large portion of the sensor's life, e.g., 3, 5, or 7 days. Towards end of life, the sensor's sensitivity to changes in glucose may decrease. This decline may be due to (a) a change in sensitivity (e.g., min raw_signal=m * A decrease in sensitivity that occurs monotonically over a period of several hours (e.g., 12 hours) can be recognized by determining either (a) the decrease in the sensor raw count signal (glucose + baseline) or (b) the decrease in the sensor raw count signal. For example, the following equation [4] may be used:

[0200]

number

[0201] In some embodiments, other forms of signal descriptive statistics related to signal sensitivity (e.g., median, percentiles, interquartile range, etc.) may be used as risk factor metrics for detecting end of life. In some embodiments, determining whether there has been a decrease in signal sensitivity involves comparing the measured signal sensitivity to 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 from previous in vitro and / or in vivo testing of the sensor. In some embodiments, if the measured signal sensitivity is outside of an acceptable range, the signal sensitivity may be automatically used in the end of life function. In some embodiments, the measured signal sensitivity, the change in sensitivity, and / or a predefined indicator of decreased sensitivity may be used as an input or variable in the end of life function.

[0202] In some embodiments, the sensitivity variable in the end of life 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 degradation 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 of the measured signal sensitivity is within an acceptable range. The acceptable range may be determined by a priori information, such as from previous in vitro and / or in vivo testing of the sensor. In one example, a change of more than 20% over a day is an indicator of end of life and may be useful as an input in the end of life detection function. In one example, an acceleration (e.g., rate of decrease in sensitivity) of more than 20% over a 12-hour period is an indicator of end of life and may be useful as an input in an end of life detection algorithm.

[0203] In some embodiments, the rate of change of signal sensitivity may be determined based in part on a slow moving average of 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. Thus, the change in the average of sensor data (e.g., uncalibrated (raw or filtered)) over time (e.g., 2, 3, 4, 5, 6, 7, or more days) may be interpreted as a change in the sensitivity of the sensor over time. The result of the slow moving average can be a quantifiable amount and / or a simple "yes / no" indicator of sensitivity degradation, which may be useful as one input or variable into an end-of-life function.

[0204] For example, the processor circuitry 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 exponential forgetting factor to evaluate the sensor sensitivity over time. In one embodiment of an alpha filter with exponential forgetting, an "alpha" may be used as given by the following equation [5]:

[0205]

number

[0206] Here, alpha defines how much history we want to remember (how quickly we forget). If α is 0.01, then in 1 / 0.01 (i.e., 100 time constants) samples, 63% of the previous information will be forgotten. Thus, if the sampling rate is 12 samples / hour, then 63% of the signal will be forgotten by 100 samples, e.g., about 8 hours. In such an embodiment, with three time parameters or constants, which is about one day, 5% (i.e., 0.37%) of the signal remaining from the previous day will be forgotten. * 0.37 *0.37=0.05) will remain. In the above formula, alpha is the "forgetting factor". Alpha can vary from 0 to 1, and its value indicates how quickly old measurements are forgotten by the model. If the value of alpha is closer to 1, the model adapts to recent measurements more quickly. If the value of alpha is closer to 0, the model adapts to recent measurements more slowly. The value of alpha may depend on the time since the sensor was implanted. The calculation may be recursive or non-recursive.

[0207] In some embodiments, sensitivity loss can be a risk factor metric indicative of end of life. Sensitivity loss can occur toward the end of a sensor's life due to physiological wound healing and foreign body mechanisms surrounding the sensor, or other mechanisms including reference electrode capacitance, enzyme depletion, membrane changes, etc.

[0208] 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 begins 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 long-term (48 hours) average sensor sensitivity. If the ratio of short-term sensitivity to long-term sensitivity is less than 70%, the sensor may be at risk of losing sensitivity. Loss of sensitivity may be translated into an end-of-life risk factor value, for example, with a value of about 1 until the ratio is about 70%, reducing to 0.5 at 50% and <0.1 at 25%.

[0209] FIG. 11A shows a sensor signal with loss of sensitivity (downward slope of the signal) indicative of an end-of-life condition (x-axis: time in hours, y-axis: sensor signal in counts). It can be seen that at approximately 170 hours (day 8), the sensor signal became significantly noisier with a large downward spike and gradually decreased in amplitude over the next few hours to days due to loss of sensitivity. The end-of-life detection algorithm identified these two risk factors. For loss of sensitivity, the algorithm first calculated the maximum value of the count output by the sensor over the last few days and used this maximum value as the final steady-state sensitivity of the sensor. The algorithm compared the short-term (average of the past 12 hours) average count and then calculated. If the short-term average count was lower than normal variation over an 8-hour period (i.e., 2 times the standard deviation in the last 8 hours), the algorithm flagged the sensor as at risk for end-of-life. The risk factor metric value was calculated as follows: If the short-term average was within 80% of the long-term average, the end-of-life risk factor metric value was 1. When the short-term to long-term ratio falls below 0.6, the end-of-life risk factor metric value is below 0.5, and approaches 0.1 when the ratio falls below 0.4. Alternative calculations of end-of-life risk related to sensitivity may use external standards such as glucose fingerstick readings. In either case, the particular estimated sensitivity loss may be transformed into an end-of-life risk factor metric value using functions described elsewhere herein.

[0210] In some embodiments, the sensor sensitivity may be calculated by comparing the sensor data (e.g., calibrated sensor data) to a reference blood glucose (BG). For example, a calibration algorithm adjusts the glucose estimate based on a systematic bias between the sensor and the reference BG. An end-of-life algorithm may use this bias, called the error at calibration or downward drift, to quantify or qualify end-of-life symptoms. The error at calibration may be normalized to account for irregular calibration times and may be smoothed (e.g., moving average or exponential smoothing) to give more weight to recent data. In some embodiments, an end-of-life risk factor metric value is determined based on the smoothed error obtained at calibration. In such an embodiment, the end-of-life risk factor value is 1 for all values ​​of error at calibration > -0.3, reduces to 0.5 for errors at calibration = -0.4, and reduces to < 0.1 for errors at calibration = -0.6.

[0211] In some embodiments where a factory calibration is employed such that the need for recalibration is eliminated, the loss of sensor sensitivity over time (downward slope of the sensor signal) can be estimated by filtering the raw sensor data (e.g., counts) using a low pass filter to smooth any fluctuations in the signal that may arise from the patient's blood glucose fluctuations resulting from meals or exercise. In this way, an average trend of the sensor data, and therefore the sensor sensitivity, over time can be obtained. That is, filtering the raw sensor data allows long-term trends that may be associated with sensor sensitivity decay to be distinguished from short-term trends.

[0212] For example, some embodiments employ a low pass filter to estimate the change in the magnitude of the sensor sensitivity. This may be accomplished by first filtering the signal and then sampling points along the filtered signal at appropriate time intervals (e.g., every 12 hours), thereby obtaining a downsampled signal. In some cases, the smoothing parameters used in the low pass filter may be optimized in real time to minimize the mean absolute error between the low pass filtered signal and the raw sensor data. Since the data points in the downsampled signal have been filtered, their values ​​would not be expected to undergo large deviations as a result of changes in glucose levels, since it may be assumed that the glucose levels are relatively stable over these longer periods or fluctuate only by a small amount that may be represented by a mean deviation that may be considered. The slope between various ones of the points in the downsampled signal may be determined on some continuing basis (e.g., every 24 hours) and compared to a previously calculated slope. If the ratio of the current estimated slope to the previously estimated slope shows a downward trend over a period of time, it is assumed that the magnitude of the sensor sensitivity is decreasing over time, and therefore this risk factor is more likely to be indicative of end of sensor life.

[0213] Because this technique of using a low pass filter to estimate the change in sensitivity of a factory calibrated sensor is computationally intensive and therefore may be updated, for example, only once a day, in some alternative embodiments, a technique using two low pass filters may be used to estimate the downward slope of the sensor signal. One filter, referred to herein as the slow filter, has a lower cutoff frequency than the other filter, referred to herein as the fast filter. That is, the slow filter performs more smoothing of the sensor data than the fast filter. In this way, the fast filter responds more quickly to downward trends in the slope of the sensor signal than the slow filter. However, the fast filter is also more susceptible to glycemic fluctuations. Thus, if a downward trend initially identified by the fast filter is due to a long-term trend such as that resulting from sensor sensitivity degradation and not due to short-term fluctuations, it should subsequently be identified by the slow filter. In one particular embodiment, a risk factor metric indicative of sensor sensitivity degradation may be the ratio of the difference between the fast filtered sensor data and the slow filtered sensor data to the slow filtered sensor data. The more negative this ratio is, the more likely it is due to sensor sensitivity degradation. This technique for estimating the sensitivity loss of a factory-calibrated sensor is computationally efficient and therefore can be updated on much more frequent time scales (e.g., every 5 minutes) than the previously described technique that uses only a single low-pass filter.

[0214] In some embodiments, a low pass filter may also be used to determine the aforementioned risk factor metric that looks at the rate of change of sensor sensitivity degradation, rather than simply the magnitude of sensor sensitivity degradation. Because the filtered sensor data removes fluctuations due to short-term glycemic fluctuations and thus represents the patient's average glucose level, the filtered signal is likely to have only a few (e.g., 3-5) local maxima and minima per day. A moving window having a specified duration (e.g., 6-12 hours) may be applied to the filtered signal over several days' worth of data. Within each window, local maxima and minima may be identified and a slope calculated. The slope associated with the most recent window may be compared to the slope associated with the previous window. If the slope, which represents the rate of change of sensor sensitivity degradation, increases over time as the window moves, for example, over several days, the sensor sensitivity degradation is likely increasing more rapidly and may therefore be more indicative of end of sensor life.

[0215] Another risk factor metric that may be useful in determining end-of-life is EOL noise, based on noise analysis, e.g., end-of-life (EOL) noise factors (skewness, kurtosis, and rotation), duration, magnitude and history, spectral content analysis, pattern recognition, etc. In some embodiments, the processor circuitry may be configured to evaluate the noise (e.g., amplitude, duration, and / or pattern) and determine whether there is a predefined noise pattern indicative of EOL. For example, a typical sensor end-of-life 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).

[0216] In some embodiments, the duration of the noise may indicate end of life. FIG. 15 shows a diagram illustrating noise duration associated with end of life. 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, light noise, moderate noise, or severe noise based on the relative magnitude of the sensor and filtered sensor counts and their derivatives. The EOL algorithm described in block 710 may use this information to replace the severe noise duration (e.g., the amount of sensor data in a severe noise condition) with a metric that reflects end of life risk. The assumption behind the calculation of this metric is that the end of life of a sensor will appear as an episode if continuous noise is detected rather than intermittent noise of a few samples. Thus, the end of life determination may penalize noise of longer duration more, as shown in FIG. 15. Thus, at each sample time, the total duration of noise up to that point is used to calculate the risk factor metric value at that point. In some cases, the relative duration of noise may be used to calculate an end-of-life risk factor. That is, the risk factor metric may evaluate the amount of time that the sensor data exhibited noisy behavior (e.g., severe or moderate noise) versus the amount of time that the sensor data exhibited mild or no noise within some immediately preceding time period.

[0217] In some embodiments, determining whether a predetermined end-of-life signature (noise pattern) is present involves evaluating the measured signal using a pattern recognition algorithm to identify the end-of-life signature in the sensor signal. This may be accomplished, for example, by comparing the measured sensor signal to end-of-life noise pattern characteristics to determine whether the recorded noise pattern resembles a predicted noise pattern. For example, in some embodiments, pattern recognition may also be used to perform noise analysis to identify certain types of noise artifacts. For example, the sensor data may be examined to determine the presence or absence of asymmetric (e.g., non-Gaussian) non-stationary noise in the sensor data.

[0218] In other embodiments, determining whether a predetermined noise pattern (end of life signature) is present includes comparing the measured signal to a predetermined noise pattern to determine whether the recorded noise pattern resembles the predetermined noise pattern. For example, the predetermined noise pattern may include a series of specific negative spikes in a short time frame. The predetermined noise pattern may include an increase in spike activity for a given time frame.

[0219] In one embodiment, threshold detection of the rate of change may be used to detect up or down spikes. Spikes may be detected as would be understood by one of ordinary skill in the art, e.g., point-to-point difference and thresholding, sharpness filters, etc. For example, an algorithm or function may output +1 for up spikes and -1 for down spikes. Using this spike data time series, one may use either an up spike (positive) spike detection algorithm or a down (negative) spike detection algorithm or a full spike detection (e.g., positive or negative spike time series) algorithm.

[0220] In some embodiments, end of life detection using these spike detection functions may be accomplished using a negative threshold on the moving average of the spike time series (e.g., twice as many negative spikes as positive spikes), or a threshold on total spike activity that indicates a 3-4 fold increase in total spike activity (e.g., 3 or 4). Other forms of spike detection such as a least squares acceleration filter may be employed. In some embodiments, the end of life risk factor value may be determined to be 1 for values ​​of spike metric <1, reduced to 0.5 for spike metric >2, reduced to <0.1 for spike metric >5, etc.

[0221] FIG. 11B illustrates the use of a spike detector to detect an increase in downward spikes in the signal generated by the sensor, which is a specific risk factor metric for end of life (x-axis, time in hours; left y-axis: sensor signal in counts; right y-axis: output of the spike detection filter). In some embodiments, noise occurring during sensor end of life may show large (e.g., more than about 30% drop in amplitude) downward spikes (spikes with an overall negative change from the average) among some random noise. To detect this type of end of life noise, the spike detection filter first identified positive and negative spikes as those with a point-to-point change in signal of more than 33%. Next, negative spikes were identified by looking at the average signal within the last 30 minutes (6 samples) compared to the last 12 samples. The output of these steps was a signal that was +1 for negative spikes and 0 everywhere else. The algorithm then calculated a 2-hour moving average of this signal (zeros and ones) to get the final output shown. Values ​​greater than 2 are considered to indicate end of life risk.

[0222] Another exemplary risk factor metric that can be used in end-of-life condition detection includes high-frequency activity or patterns. For example, an end-of-life signature pattern may show a significant increase in high-frequency activity when a power spectral density 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). Therefore, a high-pass or band-pass filter may be used to detect end-of-life patterns associated with high-frequency activity.

[0223] FIG. 12A shows the power spectral density of the sensor signal shown in FIG. 11A showing end-of-life symptoms. The power spectral density is the energy in each discrete frequency of the signal and was calculated using the Matlab pwelch function. Other standard functions implementing FFT can also be used. The energy content of the signal was calculated over a window of about 5 hours, exceeding the expected energy by more than 10 times (i.e., frequencies >0.2mHz), resulting in the algorithm identifying the risk factor value. The solid trace shows the power spectral density of the signal for the first 7 days when no end-of-life symptoms were observed, and the dashed trace is the power spectral density when end-of-life symptoms were observed. FIG. 12B shows the same type of graph as FIG. 12A, but from a sensor signal that did not show end-of-life symptoms. The sensor performed well, as shown by the lack of change in energy content at higher frequencies even after 7 days. In these examples, the power spectral density of the sensor signal was continuously monitored. In FIG. 12A, the measured power spectral density increases from days 7 to 11 compared to the expected power spectral density (measured from the first 7 days), indicating an end-of-life risk factor. Meanwhile, in FIG. 12B, the measured power spectral density (days 7 to 11) tracks more closely with the expected power spectral density (from the first 7 days), indicating no significant risk of end-of-life. The expected power spectral density may be calculated using a priori sensor knowledge (e.g., by using sensor data from the first few days of the sensor's life for the same or different patient or sensor session). In some embodiments, the end-of-life risk factor value is 1 for sensors whose short-term power spectral density (at specific frequencies above 0.2 mHz) is about 1-2 times the expected power spectral density, decreasing to about 0.5 when the short-term power spectral density is >5 times the long-term, and further decreasing to <0.1 when the short-term power spectral density is close to 10 times the expected power spectral density.

[0224] In some embodiments, a slowly varying long-scale average signal can be used to normalize the data and improve the reliability of the detection method, e.g., signal sensitivity or noise pattern, for example, by using the following definitions: Long_time _scale = Long-term (1-2 days) moving average or filtered raw glucose data Signature = short-term (approximately 4-6 hours) filtered (any, including spike detection) data Normalized_Signal=Signature / Long_time_scale

[0225] A threshold on the normalized signal and a duration constraint can be applied to detect end-of-life signatures. Thus, end-of-life can be detected when: Normalized_Signal>threshold (above a certain duration).

[0226] In some embodiments, the threshold and duration may be optimized to achieve a particular sensitivity and specificity. Alternatively, having a short duration constraint may be used to detect oxygen noise instead of end of life.

[0227] 9 and 10, any noise detection algorithm may be used to quantify the duration of noise of a certain amplitude (740), for example as described in more detail in Patent No. 8,260,393, which is incorporated by reference in its entirety. This noise analysis may also be used to determine whether to display the sensor data to a user. For severe noise (above a predetermined level), the algorithm may further process the signal to determine an end of life risk factor metric based on the noise.

[0228] In some embodiments, EOL noise (block 750) may be determined to be characteristic of a sensor end of life based on various algorithms that evaluate known end of life failure modes identifiable on the signal. It may have a large (>30% point drop) negatively skewed downward spike over the duration of the episode, with intermittent rapid rotations or oscillations, e.g., multiple peaks and valleys or number of derivative sign changes. Noise identification may use these features to identify whether the sensor is exhibiting end of life symptoms and, depending on the magnitude and duration, may calculate an end of life risk factor value from the episode, sometimes called a noise factor.

[0229] Another risk factor metric that may be useful in determining end of life is oxygen (e.g., concentration and / or whether there is a predetermined oxygen concentration pattern). For example, in some embodiments, the processor circuit may be configured to determine whether there is a predetermined oxygen concentration and / or trend or pattern associated with the oxygen concentration. Any oxygen sensor useful for quantifying oxygen concentration may be useful here, either separate from the sensor or integrated with the sensor. For electrochemical sensors, including potentiostats, pulsed amperometric detection may be employed to determine oxygen measurements. Pulsed amperometric detection involves switching, cycling, or pulsing the voltage of the working electrode (or reference electrode) in the electrochemical system between, for example, a positive voltage (e.g., +0.6 for detecting glucose) and a negative voltage (e.g., -0.6 for detecting oxygen). In some embodiments, oxygen deficiency may be seen at the counter electrode when insufficient oxygen is available for reduction, thereby affecting the counter electrode in that it cannot balance the current coming from the working electrode. If there is insufficient oxygen available to the counter electrode, the counter electrode may be driven to its most negative value, which may be ground or 0.0 V, in the electrochemical search for electrons, which shifts the reference and reduces the bias voltage, as explained in more detail below. In other words, a common consequence of ischemia is seen as a drop in the sensor current (e.g., lower sensitivity) as a function of glucose concentration. This occurs because the working electrode, due to the reduced bias, no longer oxidizes all of the H2O2 that reaches its surface.

[0230] In some embodiments, a non-enzymatic electrode or sensor may be used as an oxygen sensor. In an exemplary dual working electrode sensor having an enzymatic and a non-enzymatic working electrode, the non-enzymatic electrode may be used as an oxygen sensor by changing the bias potential from a positive value (e.g., 600mV-800mV) to a negative value (e.g., negative 600mV-800mV). At this potential, dissolved oxygen is reduced, resulting in 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 the positive and negative biases shown. 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.

[0231] Glucose oxidase-based sensors are known to be limited by the amount of oxygen present. When the oxygen level falls below a threshold, the enzyme electrode current drops ("oxygen starvation"), but the glucose concentration remains constant. This oxygen starvation can result in a loss of accuracy, as a lower glucose level may be reported than is actually the case. Oxygen starvation can occur later in the life of the sensor, such as when the sensor is encapsulated in a subcutaneous environment. Therefore, being able to measure oxygen allows for the detection of this encapsulation and end of life of the sensor.

[0232] In some embodiments, determining whether a predetermined oxygen concentration pattern exists involves reviewing the oxygen concentration pattern to see if the oxygen concentration is adequate. For example, an oxygen concentration pattern that indicates a decrease in oxygen availability over time may indicate the end of life of a sensor.

[0233] Another risk factor metric that may be useful in determining end of life is glucose pattern (e.g., meal characteristics such as average, variability, peak-to-peak fluctuations, expected vs. unexpected behavior such as after meals when glucose is not elevated as expected).

[0234] Yet another risk factor metric that may be useful in determining end of life 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 high compared to the other) and / or utilizing flagged outliers, as described in more detail elsewhere herein. In some embodiments, the processor circuitry may be utilized to identify a discrepancy between the reference value (e.g., BG) and the sensor value (e.g., EGV). For example, as described above in outlier detection, 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 some embodiments, a large discrepancy between the reference value and the sensor value may indicate the end of the sensor life. Without wishing to be bound by a particular theory, it is believed that this is because the sensor is reading 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 end of life risk factor metric, as described with reference to block 710. Yet another risk factor metric that may be useful in determining end of life is a measure of sensor linearity (or lack thereof). As noted above, 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).

[0235] In some embodiments, the processor circuitry is configured to evaluate various risk factors to provide an end-of-life risk factor metric, 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., the output of various tests). If the weights used in the algorithm are modified, the risk factor metric may be scaled.

[0236] In some embodiments, the processor circuitry is configured to perform a probability function to determine the probability of end of life and / or the likelihood of recovery for one or more of the plurality of end of life risk factors. In some embodiments, the risk factor metrics are mapped to a score (e.g., 0 to 1) based on one or more parameters. The score may be mapped by a function such as that shown in FIG. 13A and FIG. 13B, which replaces a particular risk factor metric or set of risk factor metrics with an end of life likelihood, for example, indicating the likelihood that the sensor will recover from the particular risk factor from end of life. As will be appreciated by one of ordinary skill in the art, other methods of replacing risk factor metrics with end of life likelihoods may be used, for example, by using one or more criteria, algorithms, functions, or equations.

[0237] In some embodiments, the risk factor metrics are fuzzified using predefined membership functions to quantify the tendency to indicate end of life. As used herein, a membership function defines the degree to which a condition is met or a value belongs to a fuzzy set defined by the function. In binary logic, a number either meets a condition completely or not at all. In fuzzy logic, a number can meet a condition to some degree described by the membership function.

[0238] As an example of a binary indicator function, the noise level is compared to a hard threshold such as "5". Any value below 5 (such as 4.9) is treated as being noise-free, and any value above 5 (such as 5.1) is treated as having an unacceptable level of noise. As an example of a fuzzy membership function, a sigmoid 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 very similarly here. Fuzzification is the determination of the degree to which a value belongs to a fuzzy set defined by a particular membership function.

[0239] In some embodiments, each of the plurality of risk factor metrics is indicative of end of life of the sensor in part if each variable is determined to meet a threshold value. In some embodiments, a combination of at least two of the plurality of risk factor metrics is indicative of end of life of the sensor if at least two of the plurality of risk factor metrics are determined to meet a threshold value.

[0240] At block 720, the processor circuitry is configured to determine whether the continuous analyte sensor is in an end-of-life condition. In one embodiment, likelihood or probability analysis may be used to determine whether the continuous analyte sensor is in an end-of-life condition. The output of the risk factor metric evaluated at block 710 becomes an input to an end-of-life determination process (block 720). As described in more detail elsewhere herein, the output of the risk factor metric at 710 may be mapped to an end-of-life risk factor metric value, e.g., a value between 0 and 1, an end-of-life probability or likelihood score, an actual value (output from the risk factor evaluation), etc. The end-of-life risk factor metric then becomes an input to an end-of-life determination function, whereby the risk factors may be weighted or otherwise processed by the processor circuitry using probability analysis, decision matrices, various subroutines, etc. to determine an actual end-of-life indicator, a probability (or likelihood) of end-of-life, a predicted time to end-of-life, etc. The probability function, decision function, various subroutines, etc. may be implemented as an end-of-life determination function, as will be understood by those skilled in the art.

[0241] In one embodiment, decision fusion may be used as a function of the various inputs (e.g., to block 720) 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. As described in more detail above with reference to block 710, appropriate risk factors are measured and fused to determine whether the sensor has reached end of life (EOL). Based on each individual risk factor, a decision may be made as to whether EOL is "yes" or "no." For example, if the sensor sensitivity has decreased over a certain amount of time t, then EOL is "yes," otherwise "no," or if the sensor had severe noise (above a predefined threshold level) for more than 12 of the last 24 hours, then EOL is "yes," otherwise "no."

[0242] The individual decisions can be combined into a single Bayesian likelihood value that 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 into a likelihood value using the following equation [6]:

[0243]

number

[0244] where d is a binary decision of 0 or 1 (no or yes), H1 is when the EOL is present, H0 is when the EOL is not present, and P() is a probability function. In practice, this means that for a "yes" decision, A=sensitivity / (1-specificity), and for a "no" decision, λ=(1-sensitivity) / specificity. For individual variable tests with high sensitivity and specificity, A is very high for a 1 decision and very small for a 0 decision.

[0245] 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 are given greater weight in the final decision making. A threshold value can be empirically determined for the final fusion likelihood value to achieve the best separation of EOL and non-EOL.

[0246] In some embodiments, linear discriminant analysis (LDA) may be used as the end-of-life decision function by taking input variables and providing an output decision.

[0247] In some embodiments, when the risk factor metrics or variables are fuzzified using a predefined membership function, the resulting degree of membership for all data quality metrics is scaled according to a predefined weight and combined to generate an indicator of the overall quality of the calculated glucose value. The weight may be applicable to all metrics and may indicate how the metric indicates end of life. These embodiments may use fuzzy logic concepts for each data quality metric using several fuzzy logic concepts such as membership functions and fuzzification, as described above. It should be understood that the fuzzification and membership functions may be applied to a processor circuit. The result of end of life detection (block 720) may be a confidence indicator that determines the likelihood of end of life beyond a simple pass / fail criterion.

[0248] In some embodiments, processing moves immediately to block 730 to provide an output related to (e.g., associated with and / or responsive to) the end of life status determined in 720 (e.g., alerting the user that the sensor is at the end of its life or is predicted to expire at some point). In some embodiments, the end of life status may be determined based on the likelihood that the sensor will not recover from the event, rather than the occurrence of the event. The likelihood that the sensor will not recover may be defined as a state when the sensor may no longer be accurate or has a long episode of noise (e.g., based on a risk factor assessment). The end of life indicator may also indicate the likelihood of recovery (e.g., when the episode may be transient rather than terminal). In some embodiments, the processor circuitry is 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 end of life condition (e.g., the likelihood of the sensor providing accurate data to the user in the next 24 hours). In some embodiments, a sensor is only determined to be at end of life if a high probability is determined that the sensor will not track glucose in the future (e.g., 24 hours) or will not show any glucose for several hours (e.g., 12 hours) (e.g., inaccuracy can be determined by comparison of EGV to a reference BG using a standard (e.g., within 20% or 20 mg / dL)).

[0249] Either integral to block 720 or thereafter, the processor may be configured to monitor the risk factor metrics (e.g., more frequently after the end of life indicator determines the likelihood of end of life) to determine whether the sensor is likely not to recover from the end of life determination. Suitable functions or algorithms for determining whether the sensor will recover from EOL may be selected from those known by those skilled in the art. For example, determining whether the sensor will recover may be a scaling from 0 to 1 based on an evaluation of one or more risk factors.

[0250] In some embodiments, the processor may be configured to determine the likelihood that the sensor will recover from the end of life determination based on the recent history. For example, the EOL determination function may determine that an end of life status is likely if the sensor is likely not to track glucose in the future, or if the sensor is likely not to detect glucose at all for an extended duration. The extended duration may include a period of more than 12 hours. In some embodiments, the processor circuitry 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 circuitry may be configured to re-allow the display of the sensor data if it is determined that the sensor has recovered from the end of life condition. However, if it is determined that the sensor is likely not to recover from the end of life condition, the process moves to block 730.

[0251] At block 730, the processor circuitry may be configured to provide an output related to the end-of-life status of the sensor. For example, a message related to the end-of-life status may be provided to a user (e.g., via a display). In some embodiments, the message to the user is provided on the sensor display itself. In other embodiments, the message is provided to a remote device, such as a laptop or mobile phone (e.g., a smartphone). In some embodiments, the end-of-life status (e.g., or information related thereto) is transmitted to a closed-loop algorithm that controls an insulin delivery device, e.g., an insulin pump. For example, the end-of-life status or score may be compared to one or more criteria that direct processing or output accordingly.

[0252] It should be appreciated that the goal of end-of-life detection is to reduce risks to performance that occur after the sensor has been in use for many days. In some embodiments, the user may be notified. For example, if an end-of-life shutoff occurs after 5 days (e.g., for a 7-day sensor), the user may be notified twice. For example, an alarm may be triggered immediately after BG entry.

[0253] In some embodiments, advantages of detecting the end of a sensor's life include that recognition of sensor failure allows for replacement of the sensor such that the accuracy of glucose estimation is not significantly affected, that end-of-life pattern recognition can be used in preventing closed-loop systems (e.g., artificial pancreas) from erroneous dosing, and that extended use of the sensor beyond the sensor's designed life can be recognized. As explained above, replacing the sensor such that the accuracy of glucose estimation is not significantly affected is desirable because it allows the user to rely on the sensor readings with confidence. Similarly, recognition of extended use of the sensor is desirable because it allows the sensor system to adjust the expectation or glucose estimate according to known contributing factors (e.g., drift, oxygen depletion, etc.), or, for example, inform the user that the sensor is not being used in accordance with FDA regulations, or prompt the user to change the sensor (e.g., by providing a message to the user), or deactivate / disable the sensor, etc., as explained in more detail below.

[0254] In some embodiments, intermittent signs of end of life may be used to turn on advanced signal filtering techniques, such as those described in more detail in Patent No. 8,260,393, which is incorporated herein by reference in its entirety.

[0255] In some embodiments, the message provided to the user related to the end of life status includes instructions to change the sensor if the sensor is in an end of life state or is predicted to be within a predetermined time frame. The predetermined time frame may be based on a predicted time until end of life, may be based on a predetermined period of about 1, 2, 3, 4 hours or more, etc. In other embodiments, the message related to the end of life status includes a warning that the sensor will shut down within a predetermined time frame if end of life is determined to have occurred. The predetermined time frame may be immediate or within minutes or hours.

[0256] In other embodiments, the output provided in block 730 may include an instruction, command, or set of instructions to interrupt what is being shown on the display (e.g., to a user or host) to avoid showing inaccurate data on the display based on an end of life score or status compared to a baseline, for example.

[0257] In still other embodiments, the output provided in block 730 may include an instruction, command, or set of instructions, such as to shut down the sensor, disable the display of real-time sensor data, instruct the user to remove the sensor, etc., based on the end-of-life score or status compared to a baseline. In some embodiments, the instruction may shut off the display and / or shut off the monitor itself. Advantageously, the system is configured to not allow the user to resume the session until a new sensor insertion is verified.

[0258] 10, a flowchart is presented illustrating an exemplary process associated with the flowchart of FIG 9. In some embodiments, the goal of the end of life (EOL) algorithm is to assess the sensor's ability to be accurate and reliable in the future. In this exemplary embodiment, the algorithm monitors noise and offset from BG (error at calibration) in the sensor data, replaces them with risk factor metrics, and determines the end of life status from the current condition state (e.g., noise or large error state).

[0259] 10, inputs to the algorithm may include various risk factor metrics including, for example, duration of noise (740), EOL noise (750), and downward drift of sensor sensitivity (760) determined based on an analysis of errors in calibration. These inputs may be determined iteratively or trended over time, and the results used in the combined algorithm of block 770. In this example, the output for each of the risk factor assessments may be substituted with a risk factor value based on the current condition or state (e.g., a value between 0 and 1), which is fed into the combined algorithm of block 770.

[0260] As described in detail above, there are several risk factor metrics associated with end of life, including, for example, (1) days of sensor life, e.g., time since insertion, (2) noise amplitude and duration, (3) noise type, and (4) sensitivity loss. It should be noted that the various risk factor metrics described herein may be used in conjunction with one another in any desired combination. Examples of some specific exemplary combinations that may be employed are provided below. It is described herein how some of these risk factor metrics may be quantified and how the likelihood of sensor recovery from these conditions may be calculated. It should be understood that the risk factors identified in blocks 740, 750, and 760 may all be evaluated in block 710 of FIG. 9.

[0261] In this example, the end of life algorithm is selectively turned on at a predetermined time (e.g., 5 days after implant) rather than time since implant being used as an input to the end of life determination algorithm 770. However, time since implant can further enhance the ability of the end of life algorithm to accurately determine end of life, since the longer the sensor is implanted, the more likely it is that it will reach end of life.

[0262] In this example, a noise detection algorithm, as described in more detail in Patent No. 8,260,393, incorporated herein by reference, is used to quantify the sensor data as clean or noisy (mild, moderate, or severe) based on the amplitude of the noise and the difference between the raw sensor signal and the filtered sensor signal. A noise duration (740) is determined based on the length of the noise of a particular severity. The occurrence of a noise episode of a particular severity (predetermined level of noise) longer than two hours affects the likelihood of recovery. FIG. 13A shows a substitution function that, in one embodiment, maps noise duration values ​​to risk factor values ​​(e.g., from 0 to 1). The substitution function may be in the form of a function that relates the likelihood of recovery to noise duration, or may use other mapping techniques, such as lookup tables, various subroutines, and / or the like. Although some sensors experience episodes of noise, an episode of continuous noise lasting less than two hours does not necessarily indicate the end of the sensor's life. Thus, in the example shown in FIG. 13A, any continuous noise duration less than 2 hours may have a very high value (e.g., a risk factor value of 1) indicating a high likelihood of recovery from a noise episode. As the duration increases, the likelihood that the sensor is normal decreases as shown. At about 5 hours, the likelihood that the sensor is normal decreases to about 50% (e.g., a 0.5 risk factor value), and above 8 hours, the risk factor value may shift to, for example, about 0.1. In some embodiments, the risk factor value is calculated continuously over a period of time, and the resulting risk factor value may be, for example, smoothed, for example, using exponential forgetting, etc. Duration may be tracked cumulatively over a sensor session, on a daily basis (e.g., hours / day), continuously, etc.

[0263] The EOL noise (block 750) is specific to and determined based on an algorithm that evaluates various aspects of the sensor signal related to noise, such as, for example, the skewness of a short duration of noise (e.g., 2 hours), the average rate of negative change of the signal within the episode, and the number of peaks and valleys (number of revolutions) within the episode. Once these parameters are calculated, a noise factor (e.g., 0 to 1) is calculated by combining each parameter, as can be understood by one of ordinary skill in the art. The parameters and / or EOL noise factor may be smoothed, for example, using an exponential forgetting factor. The EOL noise factor may be replaced with an end-of-life risk factor value, similar to FIG. 13A, e.g., a noise factor close to 0 is replaced with an end-of-life risk factor of 1, while a noise factor >0.5 has an end-of-life risk factor of <0.5. These values ​​may be scaled if the weights used in the algorithm are modified.

[0264] The sensitivity loss or drift (block 760) is determined based on an analysis of the error in the calibration (or receiving of the reference value). In this example, the error at calibration is calculated by taking the difference between the time corresponding to the calibrated sensor data and the reference blood glucose data and dividing the result by the time since the continuous blood glucose monitor was last calibrated, whereby an indication of downward drift in sensitivity can be inferred, where the calibrated sensor data is based on sensor data that has already been transformed using a previous calibration (transformation function). In this example, the error at calibration is normalized and smoothed. In this example, the likelihood of recovery is estimated based on the resulting smoothed error at calibration as shown in FIG. 13B, which is a substitution function that maps the error at calibration to a risk factor value. FIG. 13B shows the substitution function that defines the mapping from the error at calibration to the risk factor value. In this example, when the smoothed error at calibration is less than -20%, the risk factor value is high (close to 1.0). However, as the error in calibration decreases below -30%, the risk factor value decreases rapidly, reaching approximately 0.1 when the error is below -50%. This error is asymmetric because sensor sensitivity tends to decrease as it approaches end of life.

[0265] The processor circuit may then take the risk factor values ​​(740, 750, 760) and combine them into a metric (e.g., weighted average) or EOL index (block 770), which is then compared to an end-of-life criterion in block 780 to determine an end-of-life status (e.g., whether the continuous analyte sensor is in an end-of-life state). FIG. 13C illustrates a substitution function that may be used by the processor circuit in one embodiment to substitute the combined EOL index calculated from the various risk factors into an end-of-life score and / or end-of-life status. In this example, after obtaining the various risk factor values, they were summed using a weighted average to obtain a combined EOL index. This risk likelihood index was normalized by the sum of the weights as shown in FIG. 13C. An end-of-life score closer to 1 indicates a high likelihood of recovery from the various EOL symptoms (risk factors), while a value closer to 0 indicates a good possibility or likelihood that the sensor has reached the end of its life.

[0266] It should be understood that blocks 770 and 780 may be performed collectively in block 720 of FIG. 9. It should be noted that for all the above risk factors, the risk factor value may be calculated from a look-up table or function developed from a priori knowledge of the sensors and their behavior over time. Once these risk factors are combined into a metric (e.g., weighted average) or EOL index (block 770) using an appropriate algorithm, this is then used in block 780 of FIG. 10 to determine the end-of-life status of the sensor (e.g., if the sensor is likely to be useful and accurate in the future). For example, if the overall weighted average of the EOL index is close to 1, the sensor is likely to recover from the current state and is likely to be accurate and useful in the future (e.g., next 24 hours). If the weighted average is <0.5, the sensor is unlikely to be useful in the future. This value may be used to take different actions (e.g., provide a different output in block 730 of FIG. 9).

[0267] In another example of how various risk factor metrics can be combined to determine the end of sensor life, the processor circuit starts by determining that it is the 5th day and can evaluate a plurality of blood glucose entries over time to determine the end of life. According to block 710, the processor circuit evaluates a plurality of blood glucose entries over time and the number of days the sensor has been used. For example, for the first blood glucose entry (BG1), if the sensor data << reference data (BG1) exceeds a predetermined amount (e.g., outside a difference of 40% or 40 mg / dL), the resulting value as one of the results can be considered by the end-of-life determination function.

[0268] If the end-of-life criterion is met for BG1, the sensor system can prompt the user for a second blood glucose entry (BG2). If BG2 confirms that BG1 is the accurate blood glucose value (e.g., by being a comparable value), the BG value is considered to partially indicate the end of life. One skilled in the art can envision a subroutine useful for evaluating the BG vs. EGV difference and thereby mapping the likelihood of recovery.

[0269] As described above, the processor also considers the number of days the sensor has been used. For example, the end-of-life determination process may only start after a specific day after implantation and when an outlier detection has identified an outlier that meets a specific criterion. The number of days the sensor has been used can be determined or derived by examining the maximum possible sensor value based on the calibration history (4th day+). If the maximum sensor value is still << BG1 or BG2, the number of days the sensor has been used is considered to partially indicate the end of life.

[0270] Considering the BG value and the number of days the sensor has been used, in block 720, various subroutines are used to determine the end of life, determine the likelihood of recovery, or determine that it is zero. Thereafter, the sensor system can provide an instruction to block the sensor according to block 730.

[0271] In some embodiments, an assumption can be made using an end-of-life algorithm. For example, a blood glucose reading can be assumed to be accurate. If there are two incorrect blood glucose readings (e.g., sensor <<BG) as described above in outlier detection / screening, the blood glucose value can indicate, in part, the end of life (as described in this example).

[0272] As yet another example of how various risk factor metrics can be combined to determine sensor end of life, the processor circuit evaluates the signal for trends in noise (e.g., signal spikes) and trends in sensor sensitivity and provides risk factors to be processed by an end-of-life function or algorithm. This example is described below.

[0273] In block 710, the processor circuit compares the measured signal to a predetermined noise pattern to determine if the recorded noise pattern is similar to the predetermined noise pattern. For example, the predetermined noise pattern can include a series of specific negative spikes in a short time frame. The predetermined noise pattern can include an increase in spike activity for a given time frame. In this example, the processor circuit identifies a three-fold increase in signal spikes over the past 24 hours, which is a risk factor indicating end of life.

[0274] Similarly, the processor circuit evaluates sensitivity over time and either identifies a change in sensor sensitivity of more than 20% over 24 hours or compares it to a predicted profile as described with reference to FIGS. 1 and 2, which is a risk factor indicating end of life. As described above, sensor sensitivity can be determined using, for example, raw count analysis or comparison to a reference blood glucose value. In block 720, the end-of-life function uses instructions or conditions to process the spike trend and sensitivity trend to determine if end of life has occurred, is occurring, or will occur, or to determine the likelihood of recovery. If the likelihood of recovery is less than 50%, the sensor system can provide an instruction to block the sensor according to block 730.

[0275] Yet another example of how various risk factor metrics can be combined to determine sensor end-of-life is shown with respect to FIG. 16 , which shows a process flow 810 for determining an end-of-life condition for a continuous analyte sensor. The process flow 810 may be performed by a processor circuit associated with a continuous analyte sensor as described herein. In the example of FIG. 16 , the processor circuit may use raw sensor data as input data to evaluate three risk factor metrics. However, it will be understood that other exemplary configurations of the process flow 810 may evaluate more or less risk factor metrics than those shown in FIG. 16 . The three exemplary risk factor metrics shown in FIG. 16 include downward drift of sensor sensitivity over time (represented by downward drift risk factor metric 812), the presence, absence, and / or amount of asymmetric non-stationary noise in the sensor data (represented by asymmetric non-stationary noise risk factor metric 814), and the duration of the noise risk factor (represented by extended duration noise risk factor metric 816).

[0276] The processor circuitry may determine the risk factor metrics 812, 814, 816, for example, according to any of a variety of techniques described herein (e.g., pattern recognition, spectral content analysis). For example, the downward drift risk factor metric 812, which examines the magnitude of degradation in sensor sensitivity, may be determined by different techniques depending in part on whether the sensor is factory calibrated or periodically self-calibrated by the patient. In the case of factory calibration, for example, a technique using one or two low pass filters may be employed to estimate the slope of the sensor sensitivity over time.

[0277] The outputs of the risk factor metrics 812, 814, and 816 evaluated by the processor circuitry are directed to de-noising filters 820, 822, and 824, respectively, which smooth the risk factor data to remove noise that occurs in the process of estimating the risk factors themselves. The de-noising filters may, for example, smooth the various risk factor values ​​to reduce undesirable noise.

[0278] The three smoothed risk factors output by each denoising filter may be directed to a respective substitution function 830, 832, and 834. Each substitution function 830, 832, 834 may map the risk factor metrics to end-of-life risk factor values ​​on a normalized scale, such as 0 to 1, where, for example, 1 may represent a low likelihood of sensor end-of-life and 0 may represent a high likelihood of sensor end-of-life. In some embodiments, one, two, or all three of the substitution functions may be in the form of a logistic regression function. In some examples, each substitution function 830, 832, 834 may use a single one of the smoothed risk factors as a term. As a result, each substitution function may provide an output that is sensitive to only their corresponding risk factor.

[0279] The processor circuitry may apply wait functions 835, 836, 837 to the risk factor values ​​output by the substitution functions 830, 832, 834, respectively. It may assign a weight to each value. The assigned weights may be equal or unequal. For example, in a simple example where all weights are equal and the output from the substitution functions assigns values ​​between 0 and 1, each weighted value may have a maximum value of 1 / 3. The processor circuitry may sum the weighted outputs of the substitution functions 830, 832, 834 in a summing operation 840, resulting in a likelihood that an end-of-life condition exists for the continuous analyte sensor (e.g., an end-of-life likelihood). In some examples, the processor circuitry may compare the likelihood that an end-of-life condition exists to a threshold value. If the end-of-life likelihood is above the threshold value, the processor circuitry may take action to recommend termination of sensor use.

[0280] FIG. 17 illustrates another exemplary process flow 850 that may be performed by a processor circuit associated with a continuous analyte sensor to determine an end-of-life status for the continuous analyte sensor. In the example of FIG. 17, the processor circuit assesses the end-of-life status using four risk factor metrics: a downward drift risk factor metric 842, an asymmetric noise risk factor metric 844, an extended duration noise risk factor metric 846, and a rate of downward drift risk factor metric 848. In the example of FIG. 17, the risk factor metrics 842, 844, 846, 848 are determined by the processor circuit from a sensor signal, such as from raw counts generated by the continuous analyte sensor. Although four risk factor metrics 842, 844, 846, 848 are illustrated in FIG. 17, it will be understood that the process flow 850 of FIG. 17 may be implemented with more or fewer risk factor metrics. Also similar to FIG. 16, the processor circuitry applies noise removal filters 852, 854 to the outputs of the asymmetric non-stationary noise risk factor metric 844 and the extended duration noise risk factor metric 846.

[0281] 17 , a single logistic regression function 856 is applied to multiple risk factor metrics 842, 844, 846, 848. Instead of returning a risk factor value for each individual risk factor metric, the single logistic regression function 856 may receive each of the risk factor metrics 842, 844, 846, 848 as inputs and provide a single end-of-life likelihood 858 as an output. (If the risk factors are provided to a de-noising filter, such as risk factor metrics 844 and 846 being provided to respective de-noising filters 852, 854, the output of the de-noising filter may be provided to the logistic regression function 856.) In some examples, the logistic regression function 856 may be configured with one or more terms corresponding to each of the risk factor metrics 842, 844, 846, 848. For example, the logistic regression function 856 may include one or more terms corresponding to risk factor metric 842, one or more terms corresponding to risk factor metric 844, one or more terms corresponding to risk factor metric 846, one or more terms corresponding to risk factor metric 848, etc. The output of the logistic regression function 856, in some examples, is an end of life likelihood 858 indicating the likelihood that an end of life condition exists for the analyte sensor. For example, the end of life likelihood 858 may be on a scale from 0 to 1, as described herein. In some examples, the processor circuit may compare the end of life likelihood 858 of the logistic regression function 856 to a threshold value to determine whether an end of life condition exists for the continuous analyte sensor.

[0282] Using a common logistic regression function 856 to evaluate multiple risk factor metrics 842, 844, 846 may provide certain advantages. For example, this configuration may make it easier to tune the logistic regression function 856, may be more robust, and may provide a closed-form solution when optimizing. In some embodiments, the process flow 850 may be configured with dedicated logistic regression functions for one or more risk factor metrics in addition to the logistic regression function 856. Of course, the process flow 850 of FIG. 17 may alternatively employ dedicated transform functions for some or all of the individual risk factor metrics instead of the single function shown. Similar to FIG. 16.

[0283] The various risk factor metrics employed in the system of FIG. 17 may be determined according to any of the various techniques described herein (e.g., pattern recognition, use of one or more low pass filters, spectral content analysis, etc.).

[0284] Reusing the Sensor In some embodiments, sensor reuse may be detected. This may be accomplished by using an end-of-life sensor feature and comparing the profile of a used sensor to what a new sensor's profile should look like. Sensor reuse can be dangerous to a user as the sensor may provide inaccurate data that the user can rely on.

[0285] FIG. 14 is a flow chart of an exemplary process 900 for determining sensor reuse, according to one embodiment. At block 902, a sensor insertion event is triggered. The insertion event may be one of any number of possible events indicating that a new sensor has been implanted, such as a user providing input to the sensor system that a new sensor has been implanted, the sensor system detecting an electrical connection to the sensor, a predetermined time having passed since the system prompted the user to use a new sensor, etc. Next, at step 904, a data point or series of data points is collected from the analyte sensor being used, and at step 906, an initial calibration is performed using the collected data to generate an initial calibration. Then, at block 908, the data point or series of data points is used as an input to an end-of-life sensor function. At block 910, the end-of-life function generates a sensor profile. The sensor profile may include graphs showing noise, sensitivity, etc., as described above with respect to, for example, FIGS. 11-13 and 15. At block 912, the sensor profile from block 910 is compared to the standard sensor profile of the new sensor and a similarity profile is calculated. At step 914, a determination is made whether the calculated similarity profile indicates a new sensor or a reused sensor. If the similarity profile indicates a new sensor, the example process 900 ends at block 916. If the similarity profile indicates a reused sensor, at block 918, a notification may be provided.

[0286] The sensor reuse routine of block 918 may include triggering an audible and / or visual alarm to notify a user of improper sensor reuse. The alarm may also inform a user of reasons why sensor reuse may be undesirable, such as potentially providing inaccurate and unreliable sensor readings. The sensor reuse block 918 may alternatively or additionally cause the sensor system to completely or partially shut down and / or stop displaying sensor data on the sensor system's display.

[0287] Further embodiments and examples 18 shows a simplified block diagram illustrating another process flow 1000 that may be performed by a processor circuit associated with a continuous analyte sensor to determine an end-of-life condition of the continuous analyte sensor. In some embodiments, the configuration of FIG. 18 may enable real-time detection of gradual sensor degradation without sacrificing healthy sensor life.

[0288] The process flow 1000 may include a feature generation module 1002, a classifier model 1004, and an end-of-life state generation module 1006. The feature generation module 1002 may receive raw inputs associated with one or more risk factor metrics as described herein and may generate a representation of the one or more end-of-life risk factor metrics as features in a feature space. The features generated by the feature module 1002, collectively referred to as a feature set, may be provided to a classifier model 1004. The classifier model 1004 may be a trained machine learning model. The classifier model 1004 may receive the feature set and generate an output representing an end-of-life likelihood, or the likelihood that the continuous analyte sensor is in an end-of-life state. The end-of-life state generation module 1006 may receive the end-of-life likelihood and determine whether the continuous analyte sensor is in an end-of-life state.

[0289] In some examples, the feature generation module 1002 may receive some or all of the raw data stream generated by the continuous analyte sensor. In some examples, the feature generation module 1002 may also receive noise condition data describing the noise conditions of the continuous analyte sensor. In some examples, the feature generation module 1002 may derive the noise condition data of the continuous analyte sensor from the raw data stream. The feature generation module 1002 may also receive other inputs indicative of other risk factor metrics.

[0290] The feature generation module 1002 processes its various inputs to generate features that constitute a feature set that is provided to a classifier model 1004. The features in the feature set may correlate to progressive sensor decline (PSD) or otherwise indicate whether a continuous analyte sensor is in an end-of-life condition.

[0291] The processor circuitry may apply the classifier model 1004 to the generated feature set to determine an end of life likelihood. The end of life likelihood may indicate the probability that the continuous analyte sensor is in an end of life state based on the feature set based on the risk factor metric. In some embodiments, the end of life likelihood is a number between 0 and 1. An end of life state generation module 1006 may be executed to generate an end of life state or flag using the current end of life likelihood.

[0292] The feature generation model 1002, in some examples, may generate features from three exemplary categories: (i) noise features, (ii) downward drift features, and (iii) secondary features. The noise features may be designed to capture characteristics of noise that are often present when there are non-recoverable PSD events indicative of an end-of-life condition of a continuous glucose sensor. The downward drift features are correlated with the accumulated sensor model deviation (i.e., sensitivity / drift degradation). The secondary features may be determined based on the noise features, downward drift features, and their interactions, as described herein. The generated feature sets may capture different aspects of the PSD as it appears on the raw glucose signal from the clinical data.

[0293] In some embodiments, the classifier model 1004 is a logistic regression model that linearly combines all features with predefined coefficients and applies a non-linear function (e.g., sigmoid) to turn it into an end-of-life likelihood between 0 and 1. It will be appreciated that other types of classifier models may also be used, e.g., gradient boosting decision tree models, random forest models, etc. Prior to use, the classifier model 1004 may be trained using appropriate training data. The training data may, in some embodiments, be based on the behavior of the continuous analyte sensor associated with the processor circuitry and / or may be based on the behavior of other continuous analyte sensors. The classifier model 1004 is trained, in some embodiments, by a processor circuitry associated with the continuous analyte sensor. In other embodiments, the classifier model 1004 is pre-trained and provided to a processor circuitry associated with an already trained continuous analyte sensor.

[0294] The end-of-life status generation module 1006 may apply a threshold to the end-of-life likelihood values ​​generated by the classifier model 1004 to derive an end-of-life status for the continuous analyte sensor (e.g., an indication of whether the continuous analyte sensor is in an end-of-life state). If the end-of-life likelihood is greater than or equal to the threshold (i.e., 0.5 in probability space), the end-of-life status may be "1", indicating that the continuous analyte sensor is in an end-of-life state. If the calculated end-of-life likelihood is less than the threshold, the end-of-life status may be "0", indicating that the continuous analyte sensor is not in an end-of-life state.

[0295] In some examples, if the feature generation module 1002 and / or the classifier model 1004 have not generated an end-of-life likelihood value, the end-of-life state generation module 1006 may set the end-of-life state for the continuous analyte sensor to "-1", indicating "not calculated". The feature generation module 1002 and / or the classifier model 1004 may not have generated an end-of-life likelihood value, for example, if one or more of their components are initializing. In some examples, the end-of-life state generation module 1006 sets an end-of-life flag if the end-of-life likelihood indicates that the continuous analyte sensor is in an end-of-life state. The end-of-life flag may trigger various responsive actions by the processor circuitry in the aforementioned cases, for example, as described herein.

[0296] An additional end-of-life risk factor metric that may be considered in any of the configurations described herein is the number of times the continuous analyte sensor is dropped. For example, the continuous analyte sensor may include an accelerometer. The processing module may be in communication with the accelerometer to detect incidents in which the continuous analyte sensor is dropped. In some examples, the accelerometer may also be used to determine the force and / or location of the detected drop incident. In some examples, a large number of drops may correlate to a higher risk of the continuous analyte sensor experiencing a PSD and entering an end-of-life condition. Also, a higher force or a higher cumulative force across all drops may correlate to a higher risk of the continuous analyte sensor experiencing a PSD and entering an end-of-life condition. In some examples, the processor may write data describing the drop incidents to a memory associated with the continuous analyte sensor, which may be accessed later (e.g., during a sensor session) to determine the likelihood that the continuous analyte sensor is in an end-of-life condition.

[0297] 16 and 17, the number of detected drops, the location of the detected drops, and / or the force associated with the detected drops may be used as terms in a substitution function, such as a logistic regression function. Also, in some embodiments, the number of detected drops, the location of the detected drops, and / or the force associated with the detected drops may be provided to a dedicated substitution function, the results of which may then be weighted and summed with the results of other substitution functions to determine an end-of-life likelihood for the continuous analyte sensor, for example, as described herein. With reference to the embodiment of FIG. 18, the number of detected drops, the location of the detected drops, and / or the force associated with the detected drops may be provided to a feature generation module 1002, which may map this data to one or more features of a feature set provided to a classifier model 1004.

[0298] Another exemplary risk factor metric that may be considered in any of the configurations described herein is temperature exposure. For example, a continuous analyte sensor exposed to extreme temperatures (e.g., extremely high and / or extremely low temperatures) may be more likely to experience a PSD and enter an end-of-life condition. In some examples, the continuous analyte sensor may include a thermocouple or other temperature sensor in communication with a processor circuit. The processor circuit may process data received from the temperature sensor to detect very high and / or very low temperature events (e.g., events in which the continuous analyte sensor is exposed to temperatures above a high temperature threshold and / or below a low temperature threshold). The processor circuit may write an indication of the detected high and / or low temperature events to a memory associated with the continuous analyte sensor and may later retrieve the indication (e.g., during a sensor session) to determine an end-of-life likelihood of the continuous analyte sensor, for example, as described herein.

[0299] 16 and 17, the number of extreme temperature events, the severity of the extreme temperature events, the maximum temperature, the minimum temperature, and / or other data associated with the extreme temperature events may be applied as terms in a substitution function, such as a logistic regression function. Also, in some embodiments, the extreme temperature event data may be provided to a dedicated substitution function, the results of which may then be weighted and summed with the results of one or more other substitution functions to determine an end-of-life likelihood for the continuous analyte sensor, for example, as described herein. With reference to the embodiment of FIG. 18, the number of extreme temperature events, the severity of the extreme temperature events, the maximum temperature, the minimum temperature, and / or other data associated with the extreme temperature events may be provided to a feature generation module 1002, which may map this data to one or more features of a feature set provided to a classifier model 1004.

[0300] Other exemplary end-of-life risk factor metrics are based on the time elapsed measured from one or more milestones in the use of a continuous analyte sensor, such as, for example, the time since sensor use began (e.g., at the start of a sensor session), the time since the sensor was manufactured, whether the sensor's expiration date has passed, etc. For example, a continuous analyte sensor that has been in use for a longer period of time and / or is longer since manufacture may be at a higher risk of entering an end-of-life condition.

[0301] 16 and 17, the time since the start of sensor use, the time since sensor manufacture, and / or the expiration status of the sensor may be used as terms in a substitution function, such as a logistic regression function, and / or provided to a dedicated substitution function. The output of the substitution function may be used to generate an end-of-life likelihood for the continuous analyte sensor, as described herein. With reference to the example of FIG. 18, the time since the start of sensor use, the time since sensor manufacture, and / or the expiration status of the sensor may be provided to a feature generation module 1002, which may generate one or more features of a feature set that are provided to a classifier model 1004 based on the input.

[0302] Another exemplary end-of-life risk factor metric is the total number of glucose packets received from a continuous analyte sensor. For example, the continuous analyte sensor (or the sensor electronics associated therewith) may generate a raw data stream in the form of packets. A higher number of glucose packets received from the continuous analyte sensor may indicate a longer time that the continuous analyte sensor has been in use. A higher number of glucose packets received from the continuous analyte sensor may correlate with a higher risk that the continuous glucose sensor is in an end-of-life condition.

[0303] 16 and 17, the total number of glucose packets received may be used as a term in a substitution function, such as a logistic regression function, and / or provided to a dedicated substitution function. The output of the substitution function may be used to generate an end-of-life likelihood for the continuous analyte sensor, as described herein. With reference to the example of FIG. 18, the total number of glucose packets generated by the continuous analyte sensor may be provided to a feature generation module 1002, which may generate one or more features of a feature set that are provided to a classifier model 1004 based on the input.

[0304] Another exemplary end-of-life risk factor metric is the number of times a button or other input on a receiver or other component associated with a continuous analyte sensor is activated. For example, when a user activates a button on a receiver or other component associated with a continuous analyte sensor, it pushes the continuous analyte sensor, increasing the likelihood that the continuous analyte sensor will enter an end-of-life condition.

[0305] 16 and 17, the number of times a button or other input on a receiver or other component associated with a continuous analyte sensor is actuated may be used as a term in a substitution function, such as a logistic regression function, and / or provided to a dedicated substitution function. The output of the substitution function may be used to generate an end of life likelihood for the continuous analyte sensor, as described herein. With reference to the example of FIG. 18, the number of times a button or other input on a receiver or other component associated with a continuous analyte sensor is actuated may be provided to a feature generation module 1002, which may generate one or more features of a feature set that are provided to a classifier model 1004 based on the input.

[0306] Yet another exemplary end-of-life risk factor metric is the number of notifications provided to the user by the continuous analyte sensor. For example, notifications generated by the continuous analyte sensor may be provided to the user via a speaker, vibrator, or other output device of the continuous analyte sensor and / or the user device. Activation of the speaker, vibrator, or other output device pushes the continuous analyte sensor, making it more likely that the continuous analyte sensor will enter an end-of-life condition. In some examples, the magnitude or duration of the notification may also be used. For example, a notification having a longer duration and / or magnitude (e.g., volume, force of vibration, and / or the like) may make it more likely that the continuous analyte sensor will enter an end-of-life condition.

[0307] 16 and 17, the number of notifications provided to the user by the continuous analyte sensor, the duration of the notifications, and / or the magnitude of the notifications may be used as one or more terms in a substitution function, such as a logistic regression function, and / or may be provided to one or more dedicated substitution functions. The output of the one or more substitution functions may be used to generate an end of life likelihood for the continuous analyte sensor, as described herein. With reference to the example of FIG. 18, the number of notifications provided to the user by the continuous analyte sensor, the duration of the notifications, and / or the magnitude of the notifications may be provided to a feature generation module 1002, which may generate one or more features of a feature set that are provided to a classifier model 1004 based on the inputs.

[0308] Another exemplary end-of-life risk factor metric is the amount of time that a receiver display associated with a continuous analyte sensor has been active, for example, the amount of time that the receiver display has been active may correlate to a higher amount of time that the continuous analyte sensor has entered an end-of-life state.

[0309] 16 and 17, the amount of time that the receiver display associated with the continuous analyte sensor was active may be used as one or more terms in a substitution function, such as a logistic regression function, and / or may be provided to one or more dedicated substitution functions. The output of the one or more substitution functions may be used to generate an end of life likelihood for the continuous analyte sensor, as described herein. With reference to the example of FIG. 18, the amount of time that the receiver display associated with the continuous analyte sensor was active may be provided to a feature generation module 1002, which may generate one or more features of a feature set that are provided to a classifier model 1004 based on the input.

[0310] Another exemplary end-of-life risk factor metric is the number of times a continuous analyte sensor (and / or associated device) is connected to a power source, e.g., to recharge a battery associated with the continuous analyte sensor. This may include examples where the sensor electronics of the continuous analyte sensor include a battery and the battery is recharged. This may also include examples where the continuous analyte sensor is in communication with a user computing device and the battery of the user computing device is connected to a power source for charging. In some examples, a higher number of charging instances may correlate to a higher risk of the continuous analyte sensor entering an end-of-life condition.

[0311] 16 and 17, the number of times the continuous analyte sensor (and / or associated device) is connected to a power source may be used as one or more terms in a substitution function, such as a logistic regression function, and / or may be provided to one or more dedicated substitution functions. The output of the one or more substitution functions may be used to generate an end of life likelihood for the continuous analyte sensor, as described herein. With reference to the example of FIG. 18, the number of times the continuous analyte sensor (and / or associated device) is connected to a power source may be provided to a feature generation module 1002, which may generate one or more features of a feature set that are provided to a classifier model 1004 based on the input.

[0312] Another exemplary end-of-life risk factor metric is the amount of moisture to which a continuous analyte sensor has been exposed. This may include the amount of moisture to which the continuous analyte sensor has been exposed during manufacturing, storage, and / or use in a sensor session. In some examples, a processor circuit associated with the continuous analyte sensor may be in communication with the moisture sensor to indicate the level of moisture to which the continuous analyte sensor has been exposed. The processor circuit may store the amount of moisture in a memory for retrieval (e.g., during a sensor session) to assess an end-of-life condition.

[0313] 16 and 17, the amount of moisture to which the continuous analyte sensor was exposed may be used as one or more terms in a substitution function, such as a logistic regression function, and / or may be provided to one or more dedicated substitution functions. The output of the one or more substitution functions may be used to generate an end-of-life likelihood for the continuous analyte sensor, as described herein. With reference to the example of FIG. 18, the amount of moisture to which the continuous analyte sensor was exposed may be provided to a feature generation module 1002, which may generate one or more features of a feature set that are provided to a classifier model 1004 based on the input.

[0314] Yet another exemplary end of life risk factor metric is whether sensor maintenance tasks have been performed. Sensor maintenance tasks may include tasks to maintain the continuous analyte sensor, such as, for example, periodically cleaning the continuous analyte sensor and / or the sterile packaging in which the continuous analyte sensor is stored, sanitizing the continuous analyte sensor, etc. Failure to complete one or more sensor maintenance tasks may be correlated with an increased likelihood that the continuous analyte sensor will enter an end of life condition.

[0315] 16 and 17, the status of one or more maintenance activities may be used as one or more terms in a substitution function, such as a logistic regression function, and / or may be provided to one or more dedicated substitution functions. The output of the one or more substitution functions may be used to generate an end-of-life likelihood for the continuous analyte sensor, as described herein. With reference to the example of FIG. 18, the status of one or more maintenance activities may be provided to a feature generation module 1002, which may generate one or more features of a feature set that are provided to a classifier model 1004 based on the input.

[0316] In some examples, one or more of the end-of-life risk factor metrics described herein may be measured and / or used in conjunction with the contemporaneous temperature at the continuous analyte sensor. For example, the temperature of the continuous analyte sensor when another end-of-life risk factor metric is measured may provide an indication of whether the continuous analyte sensor is in contact with the host's skin or other tissue when an end-of-life risk factor metric such as a noise spike occurs. When an end-of-life risk factor metric is detected or measured, the processor circuitry may also measure or begin measuring the temperature of the continuous analyte sensor, for example, using a temperature sensor and / or the continuous analyte sensor itself. The processor circuitry may store the temperature in conjunction with the measured end-of-life risk factor metric.

[0317] 16 and 17, the continuous analyte sensor temperature may be used as one or more terms in a substitution function, such as a logistic regression function, along with other values ​​indicative of corresponding end-of-life risk factor metrics. The output of the one or more substitution functions may be used to generate an end-of-life likelihood for the continuous analyte sensor, as described herein. With reference to the example of FIG. 18, the continuous analyte sensor temperature may be provided to a feature generation module 1002, along with other values ​​indicative of corresponding end-of-life risk factor metrics. The feature generation module 1002 may generate one or more features of a feature set that are provided to a classifier model 1004 based on the input.

[0318] Various embodiments described herein utilize end of life risk factor metrics to determine an end of life status for a continuous analyte sensor. It will be appreciated that any of the embodiments described herein may utilize additional end of life risk factor metrics. With reference to the exemplary process flow of Figures 16 and 17, the additional end of life risk factor metrics may be processed in a manner similar to the end of life risk factor metrics 812, 814, 816, 842, 844, 846, 848 described herein. With reference to the exemplary process flow of Figure 18, the additional end of life risk factor metrics may be processed by the feature generation module 1002 and mapped to features that may be included in the feature set provided to the classifier model 1004.

[0319] In another variation of the subject matter described herein, the end-of-life determinations described herein may be performed on raw signals from a sensor prior to any post-processing that smooths the data or otherwise removes spikes or peaks. In other cases, the end-of-life determinations described herein may be performed on raw signals that are processed only upon the occurrence of one or more specified events that may affect the raw signals. That is, a limited amount of processing may be performed on the raw signals using a rules-based approach prior to performing the end-of-life determination.

[0320] In another variation of the subject matter described herein, the sample rate of data may be adjusted from its baseline value when a triggering event, such as the initiation of an end-of-life determination, is performed. For example, the sampling rate may be increased (or decreased) when an end-of-life determination is performed or when a particular risk factor metric or metrics are measured or otherwise assessed.

[0321] In some examples, the processor circuit is programmed to generate an end-of-life risk factor metric based on a potential downward drift of the raw signal stream over time, such a downward drift may indicate a decrease in sensitivity of the continuous analyte sensor, which may indicate an increased risk of the continuous analyte sensor entering an end-of-life condition.

[0322] In some examples, the processor circuit may detect and / or generate the downward drift risk factor metric by applying a pair of low pass filters to the raw data stream from the continuous analyte sensor. In some examples, the low pass filters are exponential low pass filters with different cutoff frequencies, one filter being faster than the other. An exemplary drift risk factor metric is defined as shown by the following exemplary equation [7]:

[0323]

number

[0324] In equation [7], FastFilterOutput is the output of the fast filter and SlowFilterOutput is the output of the slow filter. In some embodiments, the fast and slow, fast and slow low pass filters comprise two exponential filters having two different smoothing coefficient metrics, as shown by equations [8] and [9] below, where equation [8] represents a slow low pass filter and equation [9] represents a fast low pass filter.

[0325]

number

[0326] In equations [8] and [9], the signal (n) may be a raw data stream, e.g., α slowis a smoothing factor metric, where a value close to 0 provides little smoothing and a value of 1 provides very much smoothing. slow and α fast are both very close to 1, but α slow is α fast is closer to 1 than

[0327] At the first sample from the data stream, the filter uses the previous filter value. To account for this, in one particular embodiment, two exponential filters may be initialized using a dummy guess as shown by the following equation

[10] :

[0328]

number

[0329] In equation

[10] , Mo is the initial sensitivity of the sensor, 136 mg / dL is an exemplary average glucose value, and -25 mg / dL is the expected total glucose loss.

[0330] For each sample from the data stream, the percentage difference between the two filters may be calculated, for example, as shown by equation

[10] .

[0331]

number

[0332] The percentage difference between the slow and fast filters can indicate how much the trend is changing. If there is no underlying trend in the signal (i.e., desensitization), Filter slow (n) and Filt fast (n) should be similar in value, so the ratio will be approximately + / - 15%. If there is an underlying trend in the signal (i.e., desensitization), the ratio can become significantly negative (e.g., <-20%).

[0333] For each ratio value, the likelihood of desensitization can be calculated using the logistic function shown by equation

[11] , which provides a probabilistic view of significant desensitization. It is calculated as follows:

[0334]

number

[0335] In an embodiment of Equation

[11] , k may be equal to 25 and the midpoint may be -0.25. In various embodiments, Equation

[11] may be used as a transfer function for the downward drift risk factor metric, for example, in configurations similar to those shown in Figures 16 and 17. Also, in some embodiments, Equation

[11] may be executed by feature generation module 1002 in the configuration shown in Figure 18 to generate features for its Feature Set based on the detected desensitization, if any.

[0336] In some embodiments, the processor circuitry may generate an end of life risk factor metric based on significant trends in the data stream, both upward and downward in the data stream, and may calculate where significant positive ratios have a high likelihood, as shown in equation

[12] below.

[0337]

number

[0338] In equation

[12] , k may be equal to -25 and the midpoint may be 0.25. In various embodiments, equation

[12] may be used as a transfer function for a trend risk factor metric, for example in configurations similar to those shown in Figures 16 and 17. Also, in some embodiments, equation

[11] may be executed by feature generation module 1002 in the configuration shown in Figure 18 to generate features for the feature set based on the detected sensitivity reduction, if any. In some embodiments, the processor circuitry may be programmed to prompt the user to provide a baseline blood glucose value, for example from a finger stick glucometer. The baseline blood glucose value may be used in conjunction with the likelihood of equation

[11] or

[12] to generate an end-of-life risk factor metric that may be utilized in embodiments described herein, such as the embodiments of Figures 16-18.

[0339] In some examples, the processor circuitry may be programmed to generate an adaptive end-of-life risk factor metric based on sensor data generated by a continuous glucose sensor to account for glycemic control of a host or patient using the continuous analyte sensor. For example, if the host or patient experiences a hypoglycemic or hyperglycemic event, the processor circuitry may misidentify the event as an end-of-life risk factor metric, such as, for example, a downward drift in sensitivity. Adapting the end-of-life risk factor metric in this manner may help the processor circuitry distinguish between the occurrence of an actual condition indicative of end-of-life risk and a change in the sensor data based on glycemic changes in the host. This may reduce the dependency of the calculated end-of-life likelihood on the glycemic control of the host or patient.

[0340] To generate an adaptive end-of-life risk factor metric for a host glycemic condition, the processor circuitry can be programmed to implement a mathematical framework that measures changes in patient behavior and adapts the corresponding end-of-life risk factor metric accordingly. An exemplary function indicating an end-of-life risk factor metric based on sensor data or data derived from the sensor data, such as an estimated glucose value (EGV), is given by Equation

[13] :

[0341]

number

[0342] In the example of equation

[13] , y is the end-of-life risk factor metric, the signal is sensor data or data derived from sensor data, and the value weight represents the weight applied to the signal by the function f.

[0343] In some examples, an adaptive end-of-life risk factor metric may be generated by modifying the function given by Equation

[13] to take into account patient / host behavior. Patient / host behavior may be represented by h(egv), where h is a function of glucose level (egv) and captures specific patient behavior that may be reflected by hyperglycemia. In various examples, the function of Equation

[13] may be made to adapt to patient behavior by modifying the signal and / or by modifying one or more of the weights. An example in which a signal to mapping function is altered is provided by Equation

[14] below.

[0344]

number

[0345] In an embodiment of Equation

[14] , the function f may be nonlinear. As shown, a patient behavior function h is applied to the signal. The function h applied to the signal may reduce the impact of hyperglycemia on the resulting adaptive end-of-life risk factor metric. In some embodiments, the function h includes a set of nonlinear functions. For example, a capping function may be used in which the signal value is limited if the EGV value is greater than a data-driven threshold. For example, if the EGV is greater than the threshold, the signal value may be limited at the last signal value. Otherwise, the signal value is equal to the current signal value. In some embodiments, other functions may be used instead. In some cases, a capping threshold in the signal space (cappingThresholdSignal) is used to apply a nonlinear function to the input signal.

[0346] Also, in some embodiments, a correction factor metric may be added to the filter coefficients to adaptively change the weights of the function f given by equation

[13] . Equation

[15] shows an embodiment of incorporating patient behavior into the mapping function by modifying the weights of the mapping function.

[0347]

number

[0348] In the embodiment of equation

[15] , the weights vary as a function of the EGV.

[0349] FIG. 19 illustrates one example of a process flow 1900 that may be applied to sensor data, such as, for example, estimated glucose value (EGV), to reduce the impact of host / patient glycemic control. For example, process flow 1900 illustrates one exemplary way of implementing application of a function h to EGV to reduce the impact of host / patient glycemic control on end-of-life likelihoods generated by various embodiments described herein, including those described with respect to FIGS. 16-18. An input to process flow 1900 may be sensor data, such as, for example, an EGV value calculated by a processor circuit using a current CGM model. Process flow 1900 includes N processing layers that are applied sequentially. In the example of FIG. 19, N is equal to 3, but it will be understood that any suitable number of layers may be used.

[0350] 19, each layer (Layer 1, Layer 2, ... Layer N) comprises a filter and a pooling / downsampling module. This is illustrated by a breakout window 1902 showing an example filter 1904 and pooling / downsampling module 1906 associated with Layer 2. It will be appreciated that the other layers (Layer 1, ... Layer N) may be similarly arranged.

[0351] In some embodiments, filter 1904 may be a linear or nonlinear filter. In various embodiments, filter 1904 is a nonlinear order statistical filter. Filter 1904 may receive various filter parameters, such as a value alpha2, which may be a nonlinear function of sensor data (e.g., EGV), for example, as shown by the following exemplary equation

[16] :

[0352]

number

[0353] In an embodiment of Equation

[16] , beta can be between (0, 1). For example, when EGV < 250 → beta = 0, and when 200 < EGV < 350 → beta can be mapped linearly and / or non-linearly to (0, 1).

[0354] Pooling / Downsampling Module 1906 can apply downsampling and can take an average value, a median value, a minimum value, a maximum value, or a combination thereof. In an embodiment where Pooling / Downsampling Module 1906 applies downsampling, Module 1906 can receive a step size for downsampling as shown. The output of Process Flow 1900, e.g., the output of the Nth layer, can be used as an end-of-life risk factor metric in the configurations shown in FIGS. 16-18, for example. For example, as shown in FIG. 19, an adaptive end-of-life risk factor metric is determined to be more robust and less likely to result in incorrect end-of-life state results.

[0355] Display logic In some of the above embodiments, the display of sensor data (estimated analyte values such as estimated glucose values) is paused when the sensor is determined to have reached the end of its life. For example, FIG. 9 shows an exemplary implementation in which the end-of-life status of the sensor is determined in block 720 using a plurality of different risk factor metrics described above. The probability of end-of-life is output by block 720 and used in block 730 to provide an output based on this end-of-life status. FIGS. 16-18 show other examples for determining that a continuous analyte sensor is in an end-of-life state.

[0356] In some examples, the processor circuit may respond to a determination that the continuous analyte sensor has entered an end-of-life state by pausing the display of the sensor data (e.g., as all or part of the output in block 730). For example, when the continuous analyte sensor is in an end-of-life state, the EGV value represented by the sensor data may not be accurate. Thus, pausing the display when the continuous analyte sensor is in an end-of-life state may avoid displaying inaccurate data.

[0357] More generally, it would be advantageous to suspend or erase the display of sensor data whenever the sensor is in a state where it is likely to be providing inaccurate data, as well as when it is determined that the sensor is likely to be at the end of its life. For example, the sensor may provide inaccurate data at various stages of its life prior to reaching its end of life, as determined by any of the various methods described herein. For example, the sensor may provide an excessively noisy signal, resulting in inaccurate data, for reasons other than reaching its end of life. Conversely, the sensor may also provide data that is sufficiently accurate to be presented to the user even after the sensor has been classified as at the end of its life based on the various risk factor metrics described herein. Thus, it is desirable to suspend the display of sensor data only when it is specifically required to avoid displaying unreliable sensor data, and not simply when the sensor is classified as at some particular state (e.g., EOL state). In this manner, the useful life of the sensor may be extended by using the sensor whenever it can provide reliable data.

[0358] In some embodiments, instead of pausing the display of sensor data when various risk factor metrics collectively indicate that the sensor may be in an end-of-life stage, the display of sensor data may be paused based on a combination of end-of-life risk factor metrics including, for example, an end-of-life likelihood determined as described herein and / or a measure of noise, which may be any of the various noise risk metrics described herein, such as, for example, an absolute or relative noise magnitude.

[0359] FIG. 20 shows a graph of end-of-life likelihood versus noise measure on which a sample of sensor data from a clinical study is plotted. Data points are classified as reliable or unreliable when evaluated according to a predefined accuracy criterion. As shown, unreliable or outlier sensor readings have a wide variety of different combinations of PSD and noise measure values. However, as further shown in FIG. 20, a decision boundary may be identified where sensor readings having both end-of-life likelihood and noise measure values ​​above a certain threshold are generally classified as unreliable. As shown, some of these unreliable sensor readings have relatively low end-of-life likelihood values ​​and relatively high noise measure values, while other unreliable sensor readings have relatively high end-of-life likelihood values ​​and relatively low noise measure values. In other words, FIG. 20 shows that unreliable sensor readings may be identified as sensor readings that are located in a region of a two-dimensional space defined by the end-of-life likelihood and noise measure values. In the two-dimensional space shown in FIG. 20, these regions are located above the decision boundary.

[0360] In some embodiments, the display of sensor data may be suspended based not only on a combination of the end-of-life likelihood value and the noise measure value, but also on a combination of these values ​​with the values ​​of additional metrics, i.e., unreliable sensor readings may be identified as sensor readings that are located in a particular region of a multidimensional space defined by two or more features, two of which include the end-of-life likelihood value and the noise measure value.

[0361] Additional metrics that may be examined to identify unreliable sensor readings may include various non-linear combinations of the end-of-life likelihood value and the noise measure, including, by way of example, the product of the noise measure value and the end-of-life likelihood value, the square of the noise measure value, and the square of the end-of-life likelihood value. Yet another additional feature that may be examined includes the glucose level to which the sensor reading corresponds.

[0362] FIG. 21 illustrates one example of a process flow 2100 that may be executed by a processor circuit to suspend display of sensor data based on an end-of-life likelihood in combination with at least one other sensor property, such as a noise metric. In block 2102, the processor circuit may determine an end-of-life status of the continuous analyte sensor. The end-of-life status may be determined in any suitable manner, including, for example, those described herein in FIG. 7 and FIG. 16-18. In block 2104, the processor circuit may determine a noise metric of the continuous analyte sensor. The noise metric may be any noise metric described herein and may be determined in any suitable manner. In some examples, the end-of-life status determined in block 2102 and / or the noise metric determined in block 2104 may be determined for a sample from or derived from the data stream. In block 2106, the processor circuit may determine whether a current sample from the data stream is a reliable reading based on the end-of-life status and the noise metric. This may be determined, for example, by comparing a combination of the end of life status and the noise measure, as described herein, to a decision boundary, such as the decision boundary shown in Figure 20. Also, in some embodiments, other properties of the sensor may be taken into account, and the noise boundary may be more than two-dimensional, for example, as described herein.

[0363] If the processor circuit determines that the current sample represents a reliable reading, the processor circuit may continue to display the sensor data and / or pause the display of the sensor data at block 2108. After continuing to display the sensor data at block 2108 or when pausing the display of the sensor data, the processor circuit may return to block 2102 and perform its operations on the next sample from the data stream. If the processor circuit determines that the current sample is not a reliable reading at block 2106, the processor circuit may pause the display of the sensor data at block 2108. After pausing the display of the sensor data, the processor circuit may return to block 2102 and continue analyzing the continuous analyte sensor. In this manner, if the continuous analyte sensor produces a reliable reading after previously producing an unreliable reading, the processor circuit may pause the display of the sensor data at block 2108. In some examples, this configuration may reduce the risk of a premature end-of-life condition determination.

[0364] In some examples, instead of, or possibly in addition to, using the end-of-life likelihood to determine whether to pause the display of data, the likelihood of an early PSD may be used. The likelihood of an early PSD may be determined, for example, by examining one or more additional metrics related to characteristics of the sensor signal, the determined estimated glucose value (EGV), and / or environmental factor metrics, such as the temperature of the sensor. These characteristics may be used to classify the sensor value (e.g., as an outlier), which may then be used to determine the likelihood of an early PSD.

[0365] Exemplary examples of features related to the characteristics of the sensor signal that may be used to determine the likelihood of an early PSD include one or more values ​​associated with the raw data stream (e.g., a metric calculated based on the raw data stream, such as a moving average and / or accumulated value) and / or a rate of change of the raw data stream (e.g., a rate of change of a moving median). Similarly, features or characteristics of the EGV that may be used include EGV average, EGV variability (e.g., across a sensor session), or the total time that the EGV is above or below a particular threshold when a particular event (e.g., a hypoglycemic event) occurs, or EGV variance. For example, a hypoglycemic event may be determined to occur based on the residual between a percentile of the EGV and the instantaneous value, based on the variance from the raw EGV, etc. Such a determination regarding a hypoglycemic or other event may be assigned a confidence level indicative of the likelihood that such an event occurred.

[0366] Similar features or characteristics of EGV rate of change and / or EGV drift may also be used. Features related to EGV rate of change may be used to capture the behavior of the sensor signal before the onset of a hypoglycemic event and may be used to capture different statistics of the rate of change of the signal, such as the mean, slope, intercept, maximum, and minimum. Other characteristics of EGV rate of change that may be used to determine the likelihood of an early PSD may include the number of EGV samples with a particular high negative rate of change, or a high positive rate of change, or a drop in the signal before recovering from a hypoglycemic event. Features related to EGV drift may be used to capture the change in the average EGV level during the sensor session. When examining a particular feature associated with EGV drift, all glucose values ​​may be considered, or only glucose values ​​above or below a particular threshold may be considered.

[0367] Yet another illustrative example of features related to the characteristics of the sensor data used to determine the likelihood of an early PSD can look at the sensor signal after it has been filtered by a Kalman filter. Such features can include zero crossings of the residual signal (i.e., the difference between the raw sensor signal and the filtered (de-noised) signal), residual change, and / or residual bias. For example, metrics related to the residual signal (e.g., the number of times the residual signal changes sign within a given time frame and / or moving window, the rate of change of the residual signal, the accumulated value of the residual signal, the accumulated value of the sign of the residual signal during a moving window, etc.) can be used to characterize the noise (e.g., characterize different noise profiles) and / or to determine the likelihood of a PSD.

[0368] Another feature that may be used to determine the likelihood of an early PSD is the temperature of the sensor (e.g., mean, variability, and / or rate of change). In addition to the aforementioned features, supplemental information such as population information (e.g., from clinical data, regarding YSI distribution or rate of change, etc.) may also be used to characterize the population behavior and determine the expected distribution for some of the aforementioned features of the sensor signal being examined. Supplemental information such as user-specific data (e.g., previous values ​​or knowledge from earlier in the sensor session, such as EGV and / or rate of change information from days 2-6) may also be used in certain embodiments.

[0369] In the above-described embodiments, the display logic is used to suspend the display of sensor data based on sensor characteristics associated with features such as end-of-life likelihood, as well as sensor noise conditions and measurements. In other embodiments, the display logic may be used to suspend the display of sensor data whenever the sensor data is deemed unreliable for any reason, not just based on the aforementioned sensor characteristics. For example, in some cases, physiological characteristics of the patient may cause the sensor readings to be unreliable. One such example is "soak and recovery," which is a physiological condition surrounding the sensor that typically manifests as a suppression of the raw sensor signal for a limited period of time after the sensor is initially implanted. Thus, sensor data that is deemed unreliable due to soak and recovery conditions may also be prevented from being displayed to the user.

[0370] In some cases, at least some of the unreliable sensor data resulting from soak and recovery conditions may be replaced with predictive sensor data that is projected over a short period of time (e.g., 15-30 minutes) into the future. The predictive data may be obtained using patient-specific historical data (e.g., including knowledge of diet, activity levels, etc.) when soak and recovery conditions are detected, for example, based on measured deviations between observed and predicted glucose levels. The user may be prompted to provide one or more reference measurements (e.g., provided by a single-point blood glucose monitor) when the predicted sensor data is presented to further ensure that the data being displayed is accurate.

[0371] In some implementations, the likelihood of onset of soaking and recovery may be determined in real time based on the patient's insertion pain level. As previously mentioned, soaking and recovery are believed to be caused by trauma from the insertion of the implantable sensor. It is generally identified retrospectively. However, it may be possible to identify it in real time by correlating the user-reported insertion pain with the magnitude of the trauma, and thus the likelihood of a soaking and recovery state, and possibly its magnitude. The user may be prompted to rate the level of insertion pain on a display device, which may be compared to an aggregation of data from other patients (or previous sensor sessions of the same patient) related to the pain level and magnitude of soaking and recovery. In some cases, clusters of patients may be created based on the insertion pain response and additional data such as the patient's age, sex, and BMI, so that the patient's pain level may be compared to data from one or more clusters of patients with whom they most closely match.

[0372] Instead of pausing the display of sensor data during certain periods of soak and recovery states, some embodiments may compensate for unreliable sensor readings during soak and recovery by applying a temporary correction or adjustment to those sensor readings using a reference measurement (e.g., provided by a single-point blood glucose monitor), in which case the corrected or adjusted sensor data may be displayed to the user.

[0373] For example, in analyte monitoring systems, such as continuous glucose monitoring systems that undergo self-calibration or factory calibration, a user may provide baseline measurements multiple times during the first few hours after sensor insertion to correct for inaccurate readings during soak and recovery. In some conventional monitoring systems of this type, parameters used to calculate displayed sensor data (e.g., EGV) from the raw sensor signal are adjusted based on the baseline measurements to calculate and present to the user more accurate sensor readings. These parameters may include sensor sensitivity and baseline or background values. However, these adjustments to the parameters may not be applicable after the soak and recovery conditions dissipate and the suppressed raw sensor data signals recover. As a result, modifications to the parameters may result in overcompensation of the displayed sensor data after soak and recovery is complete, thereby providing unreliable data.

[0374] Thus, in some embodiments, during soak and recovery, deviations between the estimated analyte value and the reference measurement may be reduced or eliminated by applying a small, temporary compensation change to the estimated analyte value. This change may be applied for only a limited period of time (e.g., 30 minutes, 1-2 hours, etc.) after the reference measurement is received by the system. In this way, the estimated analyte value thus calibrated is only applicable for a relatively short period of time to avoid overcompensation of the soak and recovery. In some examples, the magnitude of the change applied may decrease continuously (or non-continuously) over the period over which it is applied, such that at the end of the period its magnitude is zero. In this way, the change made is limited to the time during which the soak and recovery condition is occurring and is not applied later in the sensor session. Moreover, if the user provides one or more additional reference measurements during the soak and recovery portion of the sensor session, an additional temporary compensation change may be applied to the estimated analyte value to reduce deviations from the reference measurement. This additional temporary compensation change may or may not overlap with the one that precedes it. That is, in some cases, a second compensation change may be applied only after a previous compensation change has been reduced (or set to zero). Alternatively, a second compensation change may be superimposed on a previous compensation change if that previous change has not yet been reduced to zero.

[0375] When a sensor session is first initiated, there is typically a warm-up period during which sensor data is not displayed to the user. Of course, it is desirable to shorten this period as much as possible to maximize the amount of time during a sensor session during which the sensor can provide reliable data. One way to help achieve this is by adjusting the way in which the raw sensor signal is filtered. The raw sensor signal can be processed by a Kalman filter to provide an estimate of the filtered signal by removing noise from the raw sensor signal. As previously mentioned, the sensor signal can be divided into a sensor sensitivity component (analyte component) and a baseline or background signal component (non-analyte component). The raw sensor signal processed by the Kalman filter can include both signal components. In one embodiment, a reduction in the error level can be achieved by filtering the raw sensor signal with a Kalman filter after subtracting an estimated background signal from the raw sensor signal. After the raw sensor signal has been filtered in this manner, the estimated background signal can be added back to the filtered sensor signal for further processing by other system components or modules, such as components that compensate for sensor-dependent time delays. Filtering the raw sensor signal in this manner to reduce error can reduce the transition time during the warm-up period required to generate reliable sensor data for display to the user, thereby further increasing the amount of time during a sensor session that the sensor can provide reliable data.

[0376] Exemplary Sensor System Configurations The embodiments of the present disclosure are described above and below with reference to flowchart diagrams of methods, devices, and computer program products. It will be understood that each block of the flowchart diagrams, and combinations of blocks in the flowchart diagrams, may be implemented by execution of computer program instructions. These computer program instructions may be loaded into a computer or other programmable data processing device, processor circuitry, etc. (e.g., controller, microcontroller, microprocessor, etc.) in a sensor electronics system to manufacture a machine, such that the instructions executed on the processor circuitry create instructions for implementing the functions specified in one or more blocks of the flowcharts. These computer program instructions may also be stored in a computer-readable memory that may instruct the computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable memory create a product including instructions that implement the functions specified in one or more blocks of the flowcharts. The computer program instructions may also be loaded onto a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to generate a computer-implemented process, such that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more blocks of the flowcharts presented herein.

[0377] In some embodiments, a sensor system for continuous measurement of an analyte (e.g., glucose) in a host is provided, the sensor system including a continuous analyte sensor configured to continuously measure a concentration of the analyte in the host, and sensor electronics physically connected to the continuous analyte sensor during sensor use. In one embodiment, the sensor electronics includes electronics configured to process the sensor data and process a data stream associated with the analyte concentration measured by the continuous analyte sensor to generate displayable sensor information including, for example, raw sensor data, transformed sensor data, and / or any other sensor data. The sensor electronics may include electronics configured to process the sensor data and process a data stream associated with the analyte concentration measured by the continuous analyte sensor to generate displayable sensor information including, for example, raw sensor data, transformed sensor data, and / or any other sensor data. The sensor electronics may include a processor and computer program instructions for implementing the processes discussed herein, including the functions specified in one or more blocks of the flowcharts presented herein.

[0378] In some embodiments, the receiver, which may also be referred to as a display device, communicates with the sensor electronics (e.g., via wired or wireless communication). The receiver may be an application-specific handheld device or a general-purpose device, such as, for example, a PC, a smartphone, or a tablet computer. In one embodiment, the receiver may be in data communication with the sensor electronics to receive sensor data, such as raw data and / or displayable data, and may include a processing module for processing and / or displaying the received sensor data. The receiver may also include an input module configured to receive inputs, such as calibration codes, reference analyte values, and any other information discussed herein, from a user, for example, via a keyboard or a touch-sensitive display screen, and may also be configured to receive information from external devices, such as insulin pumps and reference meters, via wired or wireless data communication. The inputs may be processed alone or in combination with information received from the sensor electronics module. The receiver's processing module may include a processor and computer program instructions for implementing any of the processes discussed herein, including the functions specified in one or more blocks of the flowcharts presented herein.

[0379] 22 illustrates an example of an environment 2200 that includes an analyte sensor system 2202. The analyte sensor system 2202 is coupled to a host 2201, which may be a human patient. In some examples, the host has a temporary or permanent diabetic condition or other health condition for which analyte monitoring is useful.

[0380] The analyte sensor system 2202 includes an analyte sensor 2204, which may be a continuous analyte sensor. In some embodiments, the analyte sensor 2204 is or includes a glucose sensor configured to measure a glucose concentration in the host 2201. The analyte sensor 2204 may be exposed to an analyte in any suitable manner at the host 2201. In some embodiments, the analyte sensor 2204 is fully implantable under the skin of the host 2201. In other embodiments, the analyte sensor 2204 is wearable on the host 2201's body (e.g., worn on the body but not under the skin). Also, in some embodiments, the analyte sensor 2204 is a transcutaneous device (e.g., having a sensor that resides at least partially under or in the skin of the host). It should be understood that the devices and methods described herein may be applied to any device capable of detecting a concentration of an analyte, such as glucose, and providing an output signal representative of the concentration of the analyte.

[0381] In the example of FIG. 22, the analyte sensor system 2202 also includes sensor electronics 2206. In some examples, the sensor electronics 2206 and the analyte sensor 2204 are provided in a single integrated housing (see FIG. 24). In other examples, the analyte sensor 2204 and the sensor electronics 2206 are provided as separate components or modules. For example, the analyte sensor system 2202 may include a disposable (e.g., single use) sensor attachment unit that may include the analyte sensor 2204, components for attaching the sensor 2204 to a host (e.g., an adhesive pad), and / or attachment structure configured to receive a sensor electronics unit that includes some or all of the sensor electronics 2206 shown in FIG. 23. The sensor electronics 2206 may be a reusable unit. The analyte sensor 2204 may provide a raw sensor signal indicative of the concentration of the analyte in the host 2201 using any known method, including invasive, minimally invasive, or non-invasive sensing techniques (e.g., optically stimulated fluorescence, microneedle, transdermal monitoring of glucose). The raw sensor signal may be converted to calibrated and / or filtered analyte concentration data used to provide a useful value of the analyte concentration (e.g., an estimated blood glucose concentration level) to a user, such as the host or a caretaker (e.g., a parent, relative, guardian, teacher, doctor, nurse, or any other individual interested in the health status of the host 2201).

[0382] In some examples, the analyte sensor 2204 is or includes a continuous glucose sensor. A continuous glucose sensor can be or include a subcutaneous, transcutaneous (e.g., transdermal), and / or intravascular device. In some embodiments, such a sensor or device can repeatedly (e.g., periodically or intermittently) analyze sensor data. A glucose sensor can use any glucose measurement method, such as enzymatic, chemical, physical, electrochemical, spectrophotometric, polarimetric, calorimetric, iontophoretic, radiometric, immunochemical, etc. In various examples, the analyte sensor system 2202 can be or include a continuous glucose monitor sensor available from DexCom™ (e.g., DexCom G5™ sensor, Dexcom G6™ sensor, DexCom G7™ sensor, or any variation thereof), Abbott™ (e.g., Libre™ sensor), or Medtronic™ (e.g., Enlite™ sensor).

[0383] The environment 2200 may also include various other external devices, including, for example, a medical device 2208. The medical device 2208 may be or may include a drug delivery device, such as an insulin pump or an insulin pen. In some examples, the medical device 2208 includes one or more sensors, such as another analyte sensor, a heart rate sensor, a respiration sensor, a motion sensor (e.g., an accelerometer), a posture sensor (e.g., a three-axis accelerometer), an acoustic sensor (e.g., to capture ambient or internal sounds), etc. The medical device 2208 may be wearable on a wearable item, such as, for example, a watch, glasses, contact lenses, a patch, a wristband, an ankle band, or another wearable item, or may be incorporated into a handheld device (e.g., a smartphone). In some examples, the medical device 2208 includes a multi-sensor patch that can detect, for example, one or more of an analyte level (e.g., glucose, lactate, insulin, or other substance), heart rate, respiration (e.g., using impedance), activity (e.g., using an accelerometer), posture (e.g., using an accelerometer), galvanic skin response, tissue fluid levels (e.g., using impedance or pressure).

[0384] In some examples, the analyte sensor system 2202 and the medical device 2208 are in communication with each other. The communication between the analyte sensor system 2202 and the medical device 2208 can be via any suitable wired connection and / or via wireless communication signals 2210. For example, the analyte sensor system 2202 (e.g., its sensor electronics 2206) can be configured to establish a communication connection with the medical device 2208 using a suitable short-range communication medium, such as, for example, a radio frequency medium (e.g., Bluetooth, Medical Implant Communication System (MICS), Wi-Fi, near field communication (NFC), radio frequency identification (RFID), Zigbee, Z-Wave, or other communication protocol), an optical medium (e.g., infrared), an acoustic medium (e.g., ultrasound), a cellular protocol-based medium (e.g., Code Division Multiple Access (CDMA) or Global System for Mobile (GSM)), and / or the like.

[0385] In some examples, the environment 2200 also includes other external devices, such as, for example, a wearable sensor 2230. The wearable sensor 2230 may include a sensor circuit (e.g., a sensor circuit configured to detect a glucose concentration or other analyte concentration) and a communication circuit (which may be, for example, an NFC circuit). In some examples, information from the wearable sensor 2230 may be obtained from the wearable sensor 2230 using, for example, a user computing device 2232, such as a smartphone, configured to communicate with the wearable sensor 2230 via the communication circuit of the wearable sensor when the user device 2232 is placed in proximity to the wearable sensor 2230. For example, swiping the user device 2232 over the sensor 2230 may obtain sensor data from the wearable sensor 2230 using NFC or other suitable wireless communication.

[0386] Using NFC communication may reduce power consumption by the wearable sensor 2230, reduce the size of the power source (e.g., battery or capacitor) in the wearable sensor 2230, or extend the usable life of the power source. In some examples, the wearable sensor 2230 may be wearable on the upper arm, as shown. In some examples, the wearable sensor 2230 may additionally or alternatively be on the patient's upper torso (e.g., over the heart or over the lungs), which may facilitate detection of, for example, heart rate, respiration, or posture. The wearable sensor 2236 may also be on the lower body (e.g., legs) or other parts of the body (e.g., abdomen).

[0387] In some examples, an array or network of sensors may be associated with the patient. For example, the analyte sensor system 2202 and / or one or more of the medical device 2208, an external device such as a wearable device 2220 such as a wristwatch, the additional wearable sensor 2230, etc. may communicate with each other via a short-range communication medium (e.g., Bluetooth, MICS, NFC, or any of the other options described above). The additional wearable sensor 2230 may be any of the examples described above with respect to the medical device 2208. The analyte sensor system 2202, the medical device 2208, and the additional sensor 2230 on the host 2201 are provided for purposes of illustration and description and are not necessarily drawn to scale.

[0388] The environment 2200 may also include one or more other external devices, such as a handheld smart device (e.g., a smartphone) 2212, a tablet 2214, a smart pen 2216 (e.g., an insulin delivery pen with processing and communication capabilities), a computer 2218, a wearable device 2220 such as a wristwatch, or a peripheral medical device 2222 (which may be a proprietary device such as a proprietary user device available from DexCom™), any of which may communicate with the analyte sensor system 2202 via a short-range communication medium as indicated by wireless communication signals 2210, and may also communicate with a server system (e.g., a remote data center) or remote terminal 2228 via a network 2224 to facilitate communication with a remote user (not shown), such as a technical support staff member or clinician.

[0389] The wearable device 2220 may include an activity sensor, a heart rate monitor (e.g., an optical or electrode-based sensor), a respiration sensor (e.g., an acoustic or electrode-based sensor), a location sensor (e.g., GPS), or other sensors.

[0390] In some examples, environment 2200 includes a server system 2226. Server system 2226 may include one or more computing devices, such as one or more server computing devices. In some examples, server system 2226 is used to collect analyte data from analyte sensor system 2202 and / or analyte or other data from multiple other devices, perform analysis on the collected data, generate or apply a universal or individualized model for glucose levels, and communicate such analysis, model, or information based thereon back to one or more devices in environment 2200. In some examples, server system 2226 collects inter-host and / or intra-host break-in data to generate one or more break-in characteristics, as described herein. In some examples, server system 2226 may implement it in a cloud environment.

[0391] The environment 2200 may also include a wireless access point (WAP) 2238 that is used to communicatively couple one or more of the analyte sensor system 2202, the network 2224, the server system 2226, the medical device 2208, or any of the peripheral devices mentioned above. For example, the WAP 2238 may provide Wi-Fi and / or cellular connectivity within the environment 2200. Other communication protocols, such as NFC or Bluetooth, may also be used between devices of the environment 2200.

[0392] In various examples, the processor circuitry described herein may be or may be a component of the sensor electronics 2206. In other examples, the processor circuitry described herein may be or may be a component of one of various other devices in communication with the analyte sensor system 2202, such as, for example, the medical device 2208, the user computing device 2232, the wearable device 2220, the tablet 2214, the handheld smart device 2212, the peripheral medical device 2222, the smart pen 2216, the computer 2218, the remote terminal 2228, and / or the server system 2226.

[0393] Figure 23 illustrates an example of a medical device system 2300 including the analyte sensor system 2202 of Figure 22. In the example of Figure 23, the analyte sensor system 2202 includes a sensor electronics 2206 and an exemplary sensor mounting unit 2390, although it should be understood that in some examples, the analyte sensor 2204 and the sensor electronics 2206 can be included in a common housing. Although an example of a division of components between the sensor mounting unit 2390 and the sensor electronics 2206 is shown, it should be understood that in some examples, additional components can be included in the sensor mounting unit 2390 or the sensor electronics 2206, and that some of the components (e.g., a battery or supercapacitor) shown in the sensor electronics 2206 can be alternatively or additionally (e.g., redundantly) provided in the sensor mounting unit 2390.

[0394] 23, the sensor mounting unit 2390 includes an analyte sensor 2204 and a battery 2392. In some embodiments, the sensor mounting unit 2390 may be replaceable and the sensor electronics 2206 may include debounce circuitry (e.g., a gate with hysteresis or delay) to, for example, avoid repeated power-up and power-down processes when the battery is repeatedly removed and replaced, and to avoid processing noisy signals associated with battery removal and replacement.

[0395] The sensor electronics 2206 may include electronic components configured to process sensor information, such as a raw sensor signal, to generate a corresponding analyte concentration value. The sensor electronics 2206 may include electronic circuitry associated with measuring, processing, storing, or communicating continuous analyte sensor data, including, for example, predictive algorithms associated with processing and calibrating the raw sensor signal. The sensor electronics 2206 may include hardware, firmware, and / or software that enables the measurement of analyte levels via the glucose sensor. The electronic components may be affixed to a printed circuit board (PCB) or the like and may take a variety of forms. For example, the electronic components may take the form of an integrated circuit (IC), such as an Application-Specific Integrated Circuit (ASIC), a microcontroller, and / or a processor.

[0396] In the example of FIG. 23 , the sensor electronics 2206 includes a measurement circuit 2302 (e.g., a potentiostat) coupled to the analyte sensor 2204 and configured to repeatedly obtain analyte sensor readings using the analyte sensor 2204. For example, the measurement circuit 2302 may continuously or repeatedly measure a raw sensor signal indicative of current flow in the analyte sensor 2204 between a working electrode and a reference electrode. The sensor electronics 2206 may include a gating circuit 2394 that may be used to gate a connection between the measurement circuit 2302 and the analyte sensor 2204. For example, the analyte sensor 2204 may accumulate charge over an accumulation period. After the accumulation period, the gating circuit 2394 is opened to allow the measurement circuit 2302 to measure the accumulated charge. Gating the analyte sensor 2204 can improve the performance of the sensor system 2202 by creating a greater signal-to-noise or interference ratio (e.g., because charge from the analyte reaction accumulates while interference sources such as the presence of acetaminophen near the glucose sensor do not accumulate or accumulate less than the charge from the analyte reaction).

[0397] The sensor electronics 2206 may also include a processor 2304. The processor 2304 is configured to retrieve instructions 2306 from a memory 2308 associated with the continuous analyte sensor 2204 and execute the instructions 2306 to control various operations in the analyte sensor system 2202. For example, the processor 2304 may be programmed to control the application of a bias potential to the analyte sensor 2204 via a potentiostat in the measurement circuit 2302, interpret a raw sensor signal from the analyte sensor 2204, and / or compensate for environmental factors. In some examples, the processor 2304 (e.g., along with other components of the sensor electronics 2206) may perform the operations of the processor circuitry described herein.

[0398] The processor 2304 can also store information in or retrieve information from the data storage memory 2310. In various embodiments, the data storage memory 2310 can be integrated with the memory 2308 or can be a separate memory circuit, such as a non-volatile memory circuit (e.g., flash RAM).

[0399] The sensor electronics 2206 may also include one or more sensors, such as a sensor 2312, which may be coupled to the processor 2304. The sensor 2312 may be a temperature sensor, an accelerometer, and / or another suitable sensor. The sensor 2312 may provide inputs used to generate various end-of-life risk factor metrics, as described herein. The sensor electronics 2206 may also include a power source, such as a capacitor or battery 2314, which may be integrated into the sensor electronics 2206, or may be removable, or may be part of a separate electronics unit. The battery 2314 (or other power storage component, e.g., a capacitor) may optionally be rechargeable via a wired or wireless (e.g., inductive or ultrasonic) recharging system 2316. The recharging system 2316 may collect energy or receive energy from an external or on-board source. In various embodiments, the recharging circuitry may include a circuit that harvests energy from a triboelectric charging circuit, a piezoelectric charging circuit, an RF charging circuit, an optical charging circuit, an ultrasonic charging circuit, a thermal charging circuit, a heat collection circuit, or a communication circuit. In some embodiments, the recharging circuitry may recharge a rechargeable battery using power provided from a replaceable battery (e.g., a battery provided with the base component).

[0400] The sensor electronics 2206 may also include one or more supercapacitors in the sensor electronics unit (as shown) or in the sensor mounting unit 2390. For example, the supercapacitor may allow for energy to be drawn from the battery 2314 in a very consistent manner to extend the life of the battery 2314. The battery 2314 may recharge the supercapacitor after the supercapacitor has provided energy to the communications circuitry or processor 2304, so that the supercapacitor is ready to provide energy during a subsequent period of high load. In some examples, the supercapacitor may be configured in parallel with the battery 2314. The device may be configured to preferentially draw energy from the supercapacitor as opposed to the battery 2314. In some examples, the supercapacitor may be configured to receive energy from a rechargeable battery for short-term storage and transfer energy to the rechargeable battery for long-term storage.

[0401] The supercapacitor can extend the operating life of the battery 2314 by reducing strain on the battery 2314 during periods of high load. In some embodiments, the supercapacitor reduces strain on the battery by at least 220% during high load. In some embodiments, the supercapacitor reduces strain on the battery by at least 20% during high load. In some embodiments, the supercapacitor reduces strain on the battery by at least 30% during high load. In some embodiments, the supercapacitor reduces strain on the battery by at least 50% during high load.

[0402] The sensor electronics 2206 may also include wireless communication circuitry 2318, which may include, for example, a wireless transceiver operably coupled to an antenna. The wireless communication circuitry 2318 may be operably coupled to the processor 2304 and may be configured to wirelessly communicate with one or more peripheral devices, such as an insulin pump or a smart insulin pen or other medical devices.

[0403] 23 embodiment, medical device system 2300 also includes optional external devices including, for example, peripheral device 2350. Peripheral device 2350 may be any suitable user computing device, such as, for example, a wearable device (e.g., an activity monitor) such as wearable device 2220. In other embodiments, peripheral device 2350 may be a handheld smart device (e.g., a smartphone or other device such as a proprietary handheld device available from Dexcom), tablet 2214, smart pen 2216, or dedicated computer 2218 as shown in FIG.

[0404] The peripheral device 2350 may include a UI 2352, a memory circuit 2354, a processor 2356, a wireless communication circuit 2358, a sensor 2360, or any combination thereof. The peripheral device 2350 does not necessarily include all of the components shown in FIG. 23. The peripheral device 2350 may also include a power source, such as a battery. In some embodiments, the processor 2356 (e.g., along with other components of the peripheral device 2350) may perform some or all of the operations described herein as being performed by a processor circuit.

[0405] The UI 2352 may be provided, for example, using any suitable input / output device of the peripheral device 2350, such as a touch screen interface, a microphone (e.g., for receiving voice commands), or a speaker, a vibrating circuit, or any combination thereof. The UI 2352 may receive information (e.g., instructions, glucose values) from a host or another user. The UI 2352 may also deliver information to the host or other user, for example, by displaying UI elements on the UI 2352. For example, the UI elements may show glucose or other analyte concentration values, glucose or other analyte trends, glucose or other analyte alerts, etc. Trends may be shown with UI elements such as arrows, graphs, charts, etc.

[0406] The processor 2356 may be configured to present information to or receive input from a user via the UI 2352. The processor 2356 may also be configured to store and retrieve information, such as communication information (e.g., pairing information and data center access information), user information, sensor data, trends, etc., in the memory circuit 2354. The wireless communication circuit 2358 may include a transceiver and antenna configured to communicate via a wireless protocol, such as any of the wireless protocols described herein. The sensors 2360 may include, for example, an accelerometer, a temperature sensor, a position sensor, a biosensor, or a blood glucose sensor, a blood pressure sensor, a heart rate sensor, a respiration sensor, or another physiological sensor.

[0407] The peripheral devices 2350 may be configured to receive and display sensor information that may be transmitted by the sensor electronics 2206 (e.g., in a customized data package transmitted to a display device based on respective preferences). The sensor information (e.g., blood glucose concentration level) or alerts or notifications (e.g., “high glucose level”, “low glucose level”, or “decline rate alert”) may be communicated via the UI 2352 (e.g., via a visual display, sound, or vibration). In some examples, the peripheral devices 2350 may be configured to display or otherwise communicate the sensor information communicated from the sensor electronics 2206 (e.g., in a data package transmitted to a respective display device). For example, the peripheral devices 2350 may transmit processed data (e.g., an estimated analyte value level that may be determined by processing raw sensor data), such that a device receiving the data may not need to further process the data to determine usable information (such as an estimated analyte value level). In other examples, the peripheral devices 2350 may process or interpret the received information (e.g., to issue an alert based on glucose values ​​or glucose trends). In various embodiments, the peripheral device 2350 may receive information directly from the sensor electronics 2206 or over a network (e.g., via a cellular or Wi-Fi network that receives information from the sensor electronics 2206 or from a device communicatively coupled to the sensor electronics 2206).

[0408] 23 example, medical device system 2300 includes optional medical device 2370. For example, medical device 2370 may be an external device used in addition to or instead of peripheral device 2350. Medical device 2370 may be or include any suitable type of medical or other computing device, including, for example, medical device 2208, peripheral medical device 2222, wearable device 2220, wearable sensor 2230, or wearable sensor 2236 shown in FIG. 22. Medical device 2370 may include UI 2372, memory circuitry 2374, processor 2376, wireless communication circuitry 2378, sensor 2380, therapy circuitry 2382, or any combination thereof. In some examples, processor 2376 (e.g., together with other components of medical device 2370) may perform some or all of the operations described herein as being performed by a processor circuitry.

[0409] Like the UI 2352, the UI 2372 may be provided using any suitable input / output device of the medical device 2370, such as a touch screen interface, a microphone or speaker, a vibration circuit, or any combination thereof. The UI 2372 may receive information from a host or another user (e.g., glucose values, alert settings, calibration codings). The UI 2372 may also deliver information to the host or other user, such as by displaying UI elements on the UI 2352. For example, the UI elements may show glucose or other analyte concentration values, glucose or other analyte trends, glucose or other analyte alerts, etc. Trends may be shown with UI elements such as arrows, graphs, charts, etc.

[0410] The processor 2376 may be configured to present information to or receive input from a user via the UI 2372. The processor 2376 may also be configured to store and retrieve information, such as communication information (e.g., pairing information and data center access information), user information, sensor data and trends, in the memory circuit 2374. The wireless communication circuit 2378 may include a transceiver and antenna configured to communicate via a wireless protocol, such as any of the wireless protocols described herein.

[0411] The sensor 2380 may include, for example, an accelerometer, a temperature sensor, a position sensor, a biosensor, or a blood glucose sensor, a blood pressure sensor, a heart rate sensor, a respiration sensor, or another physiological sensor. The medical device 2370 may include more than one sensor (or memory or other components), although only one sensor 2380 is shown in the example of FIG. 23. In various examples, the medical device 2370 may be a smart handheld glucose sensor (e.g., a blood glucose meter), a drug pump (e.g., an insulin pump), or other physiological sensor device, a therapy device, or a combination thereof.

[0412] In examples where the medical device 2370 is or includes an insulin pump, the pump and the analyte sensor system 2202 may communicate bidirectionally (e.g., the pump can request a change in the analyte transmission protocol, e.g., request data points or request data on a more frequent schedule), or the pump and the analyte sensor system 2202 may communicate using one-way communication (e.g., the pump can receive analyte concentration level information from the analyte sensor system). In one-way communication, the glucose value may be embedded in an advertisement message and encrypted with a pre-shared key. In two-way communication, the pump may request values ​​that the analyte sensor system 2202 may share or obtain and share in response to a request from the pump, and any or all of these communications may be encrypted using one or more pre-shared keys. The insulin pump may receive and track analyte (e.g., glucose) values ​​transmitted from the analyte sensor system 2202 using one-way communication to the pump for one or more of a variety of reasons. For example, an insulin pump can interrupt or activate insulin delivery based on glucose values ​​falling below or above a threshold value.

[0413] In some examples, the medical device system 2300 includes two or more peripheral devices and / or medical devices, each of which receives information directly or indirectly from the analyte sensor system 2202. Because different display devices have different user interfaces, the content of the data package (e.g., amount, format, and / or type of data displayed, alarms, etc.) may be customized (e.g., programmed differently by the manufacturer and / or end user) for each device. For example, referring now to the example of FIG. 22, multiple different peripheral devices may wirelessly communicate directly with the sensor electronics 2206 (e.g., on-skin sensor electronics 2206 physically connected to the continuous analyte sensor 2204, etc.) during a sensor session to enable multiple different types and / or levels of display and / or functionality associated with the displayable sensor information or to conserve battery power of the sensor system 2202, and one or more designated devices may communicate with the analyte sensor system 2202 to relay (i.e., share) information to other devices, either directly or via a server system (e.g., a networked data center) 2226.

[0414] 24 is a side view showing an example of an analyte sensor 2434 that may be implanted in a host. A housing 2402 may adhere to the skin of the host using an adhesive pad 2408. The adhesive pad 2408 may be formed from a stretchable material and may be removably attached to the skin using an adhesive. The sensor electronics may be positioned within the housing 2402. The sensor 2434 may extend from the housing 2402 under the skin of the host as shown.

[0415] It should be understood that all methods and processes disclosed herein may be used in any glucose monitoring system, continuous or intermittent. It should be further understood that the implementation and / or execution of all methods and processes may be performed by any suitable device or system, whether local or remote. Any combination of devices or systems may be used to implement the methods and processes.

[0416] All references cited herein are incorporated herein by reference in their entirety. To the extent that publications and patents or patent applications incorporated by reference conflict with the present disclosure contained herein, the present specification is intended to supersede and / or take precedence over any such conflicting material.

[0417] Unless otherwise defined, all terms (including technical and scientific terms) should be given their ordinary and customary meaning to those skilled in the art and should not be limited to any special or customized meaning unless expressly defined as such herein. It should be noted that the use of a particular term when describing a particular feature or aspect of the present disclosure should not be construed as implying that the term is redefined herein to be limited to include any specific characteristic of the feature or aspect of the present disclosure with which the term is associated. Terms and phrases used in this application, and variations thereof, should be construed as open ended, as opposed to limiting, unless expressly stated otherwise, particularly in the appended claims. For the foregoing example, the term "including" should be read to mean "including without limitation," "including, but not limited to," and the like. The term "comprising" as used herein is synonymous with "including," "containing," or "characterized by," and is inclusive or open ended and does not exclude additional, unrecited elements or method steps. The term "having" should be interpreted as "having at least." The term "including" should be interpreted as "including, but not limited to." The term "examples" is used to provide illustrative examples of the items in the discussion and is not an exhaustive or limiting list of such examples.Adjectives such as "known," "usual," "standard," and terms of similar import should not be construed as limiting the items described to a given period of time or to items available at a given time, but rather should be read to embrace known, conventional, or standard technology that may be available or known at any time now or in the future, and words such as "preferably," "preferred," "desired," or "desirable," and similar imports should not be understood to imply that a particular feature is critical, essential, or even essential to the structure or function of the invention, but rather are merely intended to highlight alternative or additional features that may or may not be utilized in a particular embodiment of the invention. Similarly, a group of items connected with the conjunction "and" should not be read as requiring that any and all of the items be present in the group, but rather should be read as "and / or" unless expressly stated otherwise. Similarly, a group of items connected with the conjunction "or" should not be read as requiring mutual exclusivity among the group, but rather should be read as "and / or" unless expressly stated otherwise.

[0418] When a range of values ​​is provided, it is understood that the upper and lower limits, and each intervening value between the upper and lower limits of that range, are included within an embodiment.

[0419] With respect to the use of substantially any plural and / or singular term in this specification, a person skilled in the art can substitute the plural for the singular and / or the singular for the plural as appropriate to the context and / or application. For clarity, various singular / plural permutations may be explicitly stated in this specification. The indefinite article "a" or "an" does not exclude the plural. A single processor or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain means are recited in mutually different dependent claims does not indicate that a combination of these means cannot be used to advantage. Reference signs in the claims shall not be construed as limiting the scope.

[0420] Those skilled in the art will further understand that where a specific number is intended in the recitation of a claim introduced, such intent will be explicitly recited in the claim, and that in the absence of such recitation, no such intent exists. For example, to aid in understanding, the following appended claims may include the use of the introductory phrases "at least one" and "one or more" to introduce the recitation of a claim. However, the use of such phrases should not be construed as suggesting that introducing a claim recitation with the indefinite article "a" or "an" limits any particular claim that includes such an introduced claim recitation to an embodiment that includes only one such recitation, even when the introductory phrase "one or more" or "at least one" and the indefinite article, e.g., "a" or "an," are included in the same claim (e.g., "a" and / or "an" should generally be construed to mean "at least one" or "one or more"). The same applies to the use of definite articles used to introduce the recitation of a claim. In addition, even if a specific number of recitations of an introduced claim is explicitly recited, a person skilled in the art will recognize that such recitation should typically be interpreted to mean at least the recited number (e.g., a mere recitation of "two recitations" without other qualifiers typically means at least two recitations, or more than two recitations). Furthermore, when a conventional expression similar to "at least one of A, B, and C, etc." is used, such a structure is generally intended in the sense that a person skilled in the art would understand the conventional expression (e.g., "a system having at least one of A, B, and C" includes, but is not limited to, a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, and C, etc.). When a conventional expression similar to "at least one of A, B, or C, etc." is used, such a structure is generally intended in the sense that one of ordinary skill in the art would understand the conventional expression (e.g., "a system having at least one of A, B, or C" includes, but is not limited to, systems having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, and C, etc.).Those skilled in the art will further appreciate that virtually any disjunctive word and / or phrase presenting two or more alternative terms, whether in the specification, claims, or drawings, should be understood to contemplate the possibility of including one of those terms, either of those terms, or both terms. For example, the phrase "A or B" is understood to include the possibilities of "A" or "B" or "A and B."

[0421] All numerical values ​​expressing amounts of raw materials, reaction conditions, and the like used herein should be understood to be modified in all instances by the term "about." Accordingly, unless otherwise indicated, the numerical parameters set forth herein are approximations that may vary depending upon the desired properties sought to be obtained. At the very least, and not as an attempt to limit the application of the doctrine of equivalents to the scope of any claims in any application claiming priority to this application, each numerical parameter should be construed in light of the number of significant digits and ordinary rounding approaches.

[0422] Moreover, although the above has been described in some detail by way of illustration and example for purposes of clarity and understanding, it will be apparent to those skilled in the art that certain changes and modifications may be implemented. The functions, steps, and / or actions of the method claims according to the embodiments of the invention described herein need not be performed in any particular order. Furthermore, although elements of the invention may be described or claimed in the singular, the plural is contemplated unless limitation to the singular is expressly stated. Therefore, the specification and examples should not be construed as limiting the scope of the disclosure to the specific embodiments and examples described herein, but rather should be construed to cover all modifications and alternatives that come within the true scope and spirit of the disclosure. [Explanation of symbols]

[0423] 2204 Analyte Sensor 2206 Sensor Electronics 2208 Medical Devices 2224 Network 2302 Measurement circuit 2304 Processor 2306 Command 2308 Memory 2310 Data storage device 2312 Sensor 2314 battery 2316 Recharging System 2318 Wireless communication circuit 2350 Peripheral Devices 2352 User Interface 2354 Memory 2356 processor 2358 Wireless communication circuit 2360 Sensor 2370 Medical Devices 2372 User Interface 2374 Memory Circuit 2376 processor 2378 Wireless communication circuit 2380 Sensor 2382 Treatment Circuit 2390 Sensor Mounting Unit 2392 battery 2394 Gate Circuit

Claims

1. 1. A method for determining an end of life status of a continuous analyte sensor, comprising: receiving sensor data from an analyte sensor; evaluating a plurality of risk factor metrics associated with an end of life condition of an analyte sensor, at least a portion of the plurality of risk factor metrics being based on the sensor data, the plurality of risk factor metrics including: downward drift of sensor sensitivity over time, amount of asymmetric non-stationary noise, and duration of noise; determining the end of life status of the analyte sensor based at least in part on the evaluating; and and providing an output related to the end of life status of the analyte sensor.

2. 2. The method of claim 1, further comprising replacing a first risk factor metric of the plurality of risk factor metrics with a first end-of-life risk factor value, the first end-of-life risk factor value indicating a likelihood that the first risk factor metric indicates that the continuous analyte sensor is in an end-of-life condition.

3. 3. The method of claim 2, further comprising replacing a second risk factor metric of the plurality of risk factor metrics with a second end-of-life risk factor value, wherein the replacing of the first risk factor metric is performed using a first replacement function and the replacing of the second risk factor metric is performed using a second replacement function different from the first replacement function.

4. The method of claim 3 , wherein at least one of the first substitution function or the second substitution function is a logistic regression function.

5. assigning a first weight to the first end-of-life risk factor value to determine a first waiting end-of-life risk factor value; assigning a second weight to the second end-of-life risk factor value to determine a second waiting end-of-life risk factor value; determining a weighted average using the first end-of-life risk factor value and the second end-of-life risk factor value; The method of claim 3 further comprising:

6. The determining step comprises: determining that the weighted average meets a threshold; and 6. The method of claim 5, further comprising determining that the continuous analyte sensor is at an end of life condition based on the determining that the weighted average meets the threshold.

7. The method of claim 6 , further comprising initiating a termination of sensor usage.

8. The method of claim 1 , wherein the plurality of risk factor metrics further comprises a rate of change of downward drift of the sensor sensitivity.

9. 10. The method of claim 8, further comprising replacing the plurality of risk factor metrics with an end-of-life likelihood using a common replacement function.

10. The method of claim 9 , wherein the substitution function is a logistic regression function.

11. The method of claim 1 , further comprising determining the downward drift of the sensor sensitivity over time, at least in part, by examining a ratio of short-term sensitivity to long-term sensitivity.

12. downsampling the sensor data to generate a downsampled signal; periodically determining a slope between selected data points in the downsampled signal to obtain a series of calculated slopes; periodically comparing the currently calculated gradient with a previously calculated gradient; 2. The method of claim 1, further comprising determining the downward drift of sensor sensitivity over time, at least in part, by: determining that the sensor sensitivity is decreasing if a ratio of the currently calculated slope to the previously calculated slope decreases over time.

13. filtering the sensor data with a slow low pass filter having a lower cutoff frequency than the fast low pass filter to obtain slow filtered sensor data; filtering the sensor data with the fast low pass filter to obtain fast filtered sensor data; 2. The method of claim 1, further comprising determining at least in part a downward drift of sensor sensitivity over time by examining a ratio of a difference between the fast filtered sensor data and the slow filtered sensor data to the slow filtered sensor data, wherein the more negative the ratio, the more likely a downward drift of sensor sensitivity over time.

14. 2. The method of claim 1, further comprising applying a classifier model to generate an end-of-life likelihood for the continuous analyte sensor based at least in part on the plurality of risk factor metrics, and wherein the end-of-life status of the analyte sensor is based at least in part on the end-of-life likelihood.

15. applying a plurality of sequential processing layers to the sensor data; the applying a first processing layer of the plurality of sequential processing layers includes applying a first filter to the sensor data to generate a first filter output; and downsampling the first filter output to generate a first layer output; 2. The method of claim 1, wherein the applying a second processing layer of the plurality of sequential processing layers comprises applying a second filter to the first layer output to generate a second filter output and downsampling the second filter output to generate a second layer output, wherein at least one of the plurality of risk factor metrics is based at least in part on the second layer output.

16. determining at least in part the downward drift of the sensor sensitivity over time by examining a rate of change of the sensor sensitivity, wherein the examining the rate of change of the sensor sensitivity comprises: filtering the sensor data with a low pass filter to obtain filtered sensor data; determining, for a first moving window, a first gradient between a local maximum of the filtered sensor data within the first moving window and a local minimum of the filtered sensor data within the first moving window; determining, for a second moving window after the first moving window, a second gradient between a local maximum of the filtered sensor data in the second moving window and a local minimum of the filtered sensor data in the second moving window; determining a rate of change of a sensor sensitivity based on the first slope and the second slope.

17. 17. The method of claim 16, wherein determining a rate of change of the sensor sensitivity comprises determining that a downward drift of the sensor sensitivity is increasing if the second slope is greater than the first slope.

18. determining a noise metric describing the sensor data; determining that the sensor data does not produce a reliable output based at least in part on the end of life status of the analyte sensor and the noise metric; suspending display of at least one value based on the sensor data in response to determining that the sensor data does not produce a reliable output; and The method of claim 1 further comprising:

19. receiving second sensor data from the analyte sensor after pausing the display of at least one value based on the sensor data; evaluating a second plurality of risk factor metrics, at least a portion of the second plurality of risk factor metrics based at least in part on the second sensor data; determining a second end of life status of the analyte sensor based at least in part on the evaluating; and determining a second noise metric describing the second sensor data; determining that the second sensor data produces a reliable output based at least in part on the second end of life status of the analyte sensor and the second noise metric; displaying at least one value based on the second sensor data in response to determining that the second sensor data produces a reliable output; and 20. The method of claim 18, further comprising:

20. 1. A continuous analyte sensor system comprising a processor circuit, the processor circuit comprising: receiving sensor data from an analyte sensor; evaluating a plurality of risk factor metrics associated with an end of life condition of an analyte sensor, at least a portion of the plurality of risk factor metrics being based on the sensor data, the plurality of risk factor metrics including: downward drift of sensor sensitivity over time, amount of asymmetric non-stationary noise, and duration of noise; determining an end of life status of the analyte sensor based at least in part on said evaluating; and and providing an output related to the end of life status of the analyte sensor.

Images