Method and system for mitigating sensor error propagation

Abstract

Medical devices and related systems and methods are provided. A method of using a transformation model to provide notifications related to a physiological condition involves identifying an error metric associated with an input variable associated with the transformation model, determining a reference output of the transformation model by providing a reference value of the input variable to the transformation model, generating an adjusted value of the input variable based on the reference value using the error metric, determining a simulated output of the transformation model by providing the adjusted value of the input variable to the transformation model, and updating the transformation model to reduce a weight associated with the input variable when a difference between the simulated output and the reference output is greater than a threshold value.

Description

Technical Field

[0001] The embodiments of the subject matter described herein relate generally to medical devices, and more specifically, the embodiments of the subject matter relate to mitigating the effects of measurement errors to improve the monitoring and management of patients through the use of sensing devices. Background Technology

[0002] Managing a diabetes treatment kit involves monitoring and regulating a patient's blood glucose levels in a largely continuous manner. Intermittently sensed glucose data samples are more commonly used for continuous glucose monitoring (CGM) purposes than continuous sampling and monitoring of the user's blood glucose levels. The sensed glucose measurements can then be used to calculate bolus doses, determine infusion pump operating commands, and provide advice or other insights to patients, healthcare providers, and / or others regarding the management of the patient's condition.

[0003] Many CGM sensors measure glucose in the interstitial fluid (ISF). To achieve the desired accuracy and reliability and mitigate the effects of noise and other stray signals, sensor data is typically calibrated based on a known good glucose level, usually obtained using a glucometer that measures blood glucose in capillaries via a so-called "finger prick measurement." However, performing such calibration measurements increases patient burden and perceived complexity, and can be inconvenient, uncomfortable, or otherwise undesirable to patients. Furthermore, based on the time it takes for glucose to diffuse from the capillaries into the interstitial space (where it is measured by the CGM sensor), ISF glucose measurement lags behind blood glucose measurement, requiring signal processing (e.g., filtering) or other techniques to compensate for this physiological lag. Additionally, various factors can cause transient variations in sensor output, which can affect calibration accuracy. Deterioration of sensor performance over time or manufacturing variations can further exacerbate these problems. Therefore, it is desirable to incorporate variations and measurement errors into consideration to reduce patient burden and improve the overall user experience without compromising accuracy or reliability. Summary of the Invention

[0004] This document provides medical devices, related systems, and methods of operation. In one embodiment, a method is provided for providing notification related to a physiological condition using an instance of a sensing element capable of providing an electrical signal influenced by a physiological condition in a patient's body. The method involves: identifying an error metric associated with an input variable, the input variable being associated with a transformation model that provides an output influenced by the value of the input variable and a weight associated with the input variable; determining a reference output of the transformation model by providing a reference value of the input variable to the transformation model; generating an adjusted value for the input variable based on the reference value using the error metric; determining an analog output of the transformation model by providing the adjusted value of the input variable to the transformation model; and updating the transformation model using reduced weights associated with the input variable when the difference between the analog output and the reference output is greater than a threshold. When one or more subsequent values ​​of the input variable, derived from one or more electrical signals output by an instance of the sensing element, are input to the updated transformation model with the reduced weights, a notification is generated at least in part based on the output of the updated transformation model.

[0005] In another embodiment, a method is provided for monitoring a patient's blood glucose levels using a glucose sensing element that provides an electrical signal influenced by the patient's glucose levels. The method involves: identifying an error metric associated with a variable influenced by the electrical signal provided by the glucose sensing element; using the error metric to determine an adjustment value for the variable; using a reference value for the variable to determine a reference output of a conversion model, wherein the reference output of the conversion model is influenced by the reference value and a weight associated with the variable; and using the adjustment value for the variable to determine a simulated output of the conversion model, wherein the simulated output of the conversion model is influenced by a simulated value and the weight. When the difference between the simulated output and the reference output exceeds a threshold, the method continues: updating the conversion model to reduce the difference; and providing the updated conversion model to a device associated with the patient, wherein when one or more subsequent values ​​of the variable derived from one or more subsequent electrical signals output by the glucose sensing element in response to the patient's glucose levels are input to the updated conversion model, the device generates a notification at least in part based on the output of the updated conversion model.

[0006] In another embodiment, a system is provided that includes a database and a server coupled to the database. The database stores historical measurement data corresponding to instances of sensing elements affected by physiological conditions. The server identifies, at least in part, an error metric associated with a variable input to a conversion model based on the historical measurement data; identifies a reference value for the input variable using the historical measurement data; determines an adjustment value for the input variable based on the reference value using the error metric; determines a reference output of the conversion model by providing the reference value of the input variable to the conversion model; determines an analog output of the conversion model by providing the adjustment value of the input variable to the conversion model; and updates the conversion model to reduce the influence of the input variable when the difference between the analog output and the reference output is greater than a threshold. When one or more subsequent values ​​of the input variable, derived from one or more electrical signals output by an instance of the sensing element, are input to the updated conversion model, a user notification is generated, at least in part, based on the output of the updated conversion model.

[0007] The above-described invention is provided to introduce, in a concise form, a series of concepts further described below in the detailed embodiments. This invention is not intended to identify the principal or essential features of the claimed subject matter, nor is it intended to help determine the scope of the claimed subject matter. Attached Figure Description

[0008] A more thorough understanding of the subject matter can be derived when considered in conjunction with the following figures, with reference to the detailed description and claims, wherein similar reference numerals throughout the figures refer to similar elements that are shown concisely and clearly and are not necessarily drawn to scale.

[0009] Figure 1 An exemplary implementation of the patient monitoring system is shown;

[0010] Figure 2 For application Figure 1 A block diagram of an exemplary embodiment of the sensor arrangement of a patient monitoring system;

[0011] Figure 3 For application to combination Figure 1 Patient monitoring system or Figure 2 A block diagram of an exemplary implementation of a system for analyzing the usage of a sensing arrangement;

[0012] Figure 4 For application in one or more exemplary embodiments Figure 3 A flowchart illustrating an exemplary error mitigation process used in a situation analysis system; and

[0013] Figure 5 For applicability in one or more exemplary embodiments Figure 3 The situation analysis system combines Figure 4 A block diagram of the insight model implemented together with the error mitigation process. Detailed Implementation

[0014] The following detailed description is illustrative in nature and is not intended to limit the scope of this subject matter or embodiments of this application, or the application and use of such embodiments. As used herein, the word “exemplary” means “serving as an example, instance, or illustration.” Any specific implementation described herein as an example is not necessarily to be construed as being more preferred or advantageous than other specific implementations. Furthermore, one is not to be bound by any express or implied theory presented in the foregoing technical fields, background art, summary of the invention, or the following detailed description.

[0015] Exemplary embodiments of the subject matter described herein generally relate to sensing elements and associated sensing arrangements or devices that provide outputs indicative of one or more features or conditions sensed, measured, detected, or otherwise quantified by the sensing elements and / or affected by such features or conditions. While the subject matter described herein is not necessarily limited to any particular type of sensing application, exemplary embodiments are described primarily in the context of sensing elements, such as interstitial glucose sensing elements that generate or otherwise provide electrical signals indicative of physiological conditions in the body of a human user or patient and / or affected by such physiological conditions.

[0016] For illustrative purposes, exemplary embodiments of this subject matter are described herein in conjunction with medical devices such as portable electronic medical devices. While many different applications are possible, the following description is primarily based on the context of a continuous glucose monitoring (CGM) device or similar patient monitoring device or system. However, the subject matter can be implemented in an equivalent manner in the context of other medical devices such as, for example, infusion devices (or infusion pumps) as part of an infusion system arrangement, injection pens (e.g., smart injection pens), etc. For the sake of brevity, conventional techniques related to glucose sensing, blood glucose meters, sensor calibration, infusion system operation, and / or other aspects of system functionality (and individual operating components of the system) may not be described in detail herein. It should be noted that the subject matter described herein can be used generally in the context of overall diabetes management or in other physiological conditions independent of or without the use of infusion devices or other medical devices (e.g., when using oral medications), and the subject matter described herein is not limited to any particular type of medication. In this respect, the subject matter is not limited to medical applications and can be implemented in any device or application that includes or incorporates sensing elements.

[0017] As described in more detail below, the topics described herein help to mitigate or otherwise reduce measurement errors, noise, and / or other variations in the output or performance of various conversion models that may be applied to the measurement output by sensing elements, in order to gain insights or otherwise provide feedback related to the monitored condition. In this regard, a conversion model can be any kind of equation or function derived for computing or otherwise transforming a set of input variables into different output variables or representations. For example, an estimation model can be obtained through machine learning or other artificial intelligence techniques to obtain an equation or function that computes an estimate of the model's output (e.g., an estimated glucose value) based on a number of input variables derived from electrical signals output by sensing elements using different weighting factors or other relative relationships between the individual input variables. As described below, a simulated output of the conversion model is generated using errors associated with one or more input variables in the conversion model, representing the potential or probable effect of the errors relative to a given input variable. Based on the magnitude of the deviation of the simulated output relative to some reference output caused by the error associated with a specific input variable, the relative weights of the input variables relative to the output of the transformation model can be reduced or otherwise adjusted to reduce the deviation between the simulated output and the reference output, thereby obtaining an updated version of the transformation model with less noise or a lower error rate relative to the input variables.

[0018] The error mitigation process described in this paper can be performed relative to lower-level transformation models, higher-level transformation models, or any combination or sequence thereof that may be employed in a given system. For example, in a system with multi-level transformation models (e.g., a lower-level glucose estimation model followed by a higher-level classification model that utilizes the lower-level glucose estimation model to generate features or other feedback or insights), the error mitigation process can take into account the propagation effects of errors associated with different lower-level input variables (e.g., input variables in the lower-level glucose estimation model) on higher-level outputs (e.g., the output of the higher-level classification model) by adjusting weights or changing the influence of those lower-level input variables or lower-level model outputs on the outputs of higher-level models. Therefore, the outputs of higher-level models are more flexible in response to measurement errors or other noise, variability or errors associated with lower-level variables, thus providing more representative insights across different operating environments, conditions, or scenarios.

[0019] Patient monitoring overview

[0020] Figure 1An exemplary embodiment of a patient monitoring system 100 is shown, which includes a medical device 102 communicatively coupled to a sensing element 104 inserted into or otherwise worn by a patient to acquire measurement data indicative of physiological conditions in the patient's body, such as sensed glucose levels. In the illustrated embodiment, the medical device 102 is communicatively coupled to a client device 106 via a communication network 110, wherein the client device 106 is communicatively coupled to a remote device 114 via another communication network 112. In such embodiments, the client device 106 may act as an intermediary for uploading or otherwise providing measurement data from the medical device 102 to the remote device 114. However, it should be understood that, for illustrative purposes, Figure 1 A simplified representation of the patient monitoring system 100 is shown and is not intended to limit the subject matter described herein in any way. For example, some implementations of the patient monitoring system 100 may support direct communication between the medical device 102 and the remote device 114 via the communication network 112. Additionally, actual implementations of the patient monitoring system 100 may include any number of instances of the medical device 102 and / or the client device 106 to support simultaneous monitoring of any number of patients.

[0021] Sensing element 104 generally represents a component of a patient monitoring system 100 configured to generate, produce, or otherwise output one or more electrical signals indicating a physiological condition sensed, measured, or otherwise quantified by sensing element 104 (e.g., sensing element 204). In this respect, the patient's physiological condition influences the characteristics of the electrical signals output by sensing element 104 such that the characteristics of the output signals correspond to or are otherwise related to the physiological condition to which sensing element 104 is sensitive. In an exemplary embodiment, sensing element 104 is implemented as an interstitial glucose sensing element inserted at a location in the patient's body, generating an output electrical signal having a current (or voltage) associated with or otherwise influenced by the interstitial fluid glucose level sensed or otherwise measured by sensing element 104 and related to that level.

[0022] Medical device 102 generally refers to a component of patient monitoring system 100 communicatively coupled to the output of sensing element 104 to receive or otherwise acquire measurement data samples from sensing element 104, to store or otherwise maintain the measurement data samples, and to upload or otherwise transmit the measurement data to server 114 via client device 106. In one or more embodiments, medical device 102 is implemented as a stand-alone sensing or monitoring device, such as, for example, a continuous glucose monitor (CGM), an interstitial glucose sensing arrangement, or a similar device. In this regard, it should be noted that, although Figure 1While the medical device 102 and sensing element 104 are depicted as separate components, in practice, the medical device 102 and sensing element 104 may be integrated or otherwise combined to provide a single device that can be worn by a patient. However, in other embodiments, the medical device 102 may be implemented as an infusion device configured to deliver fluids such as insulin into the patient's body.

[0023] In an exemplary embodiment, medical device 102 includes a controller 122, a data storage element 124 (or memory), a communication interface 126, and a user interface 128. User interface 128 generally refers to input user interface elements and / or output user interface elements associated with medical device 102. Controller 122 generally refers to hardware, circuitry, logic, firmware, and / or one or more other components of medical device 102 coupled to sensing element 104 to receive electrical signals output from sensing element 104 and to perform or otherwise support the various other tasks, operations, functions, and / or processes described herein. Depending on the embodiment, controller 122 may be implemented or carried out using a general-purpose processor, microprocessor, controller, microcontroller, state machine, content-addressable memory, application-specific integrated circuit, field-programmable gate array, any suitable programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. In some embodiments, controller 122 includes an analog-to-digital converter (ADC) or other similar sampling arrangement that samples or otherwise converts the output electrical signals received from sensing element 104 into corresponding digital measurement data values. In other embodiments, the sensing element 104 may be combined with one or more ADCs and output one or more digital measurements.

[0024] Communication interface 126 typically represents the hardware, circuitry, logic, firmware, and / or other components of medical device 102 coupled to controller 122 to output data and / or information from medical device 102 to / from client device 106 to medical device 102. For example, communication interface 126 may include or otherwise couple to one or more transceiver modules capable of supporting wireless communication between medical device 102 and client device 106. In an exemplary embodiment, communication interface 126 is implemented as a Bluetooth transceiver or adapter configured to support Bluetooth Low Energy (BLE) communication.

[0025] In an exemplary embodiment, client device 106 is implemented as a mobile phone, smartphone, tablet computer, or other similar mobile electronic device. However, in other embodiments, client device 106 may be implemented as any kind of electronic device capable of communicating with medical device 102 via network 110, such as a laptop or notebook computer, desktop computer, etc. In an exemplary embodiment, network 110 is implemented as a Bluetooth network, ZigBee network, or other suitable personal area network. Nevertheless, in other embodiments, network 110 may be implemented as a wireless ad hoc network, wireless local area network (WLAN), or local area network (LAN). In an exemplary embodiment, client device 106 includes or is coupled to a display device capable of graphically presenting data and / or information related to the patient's physiological condition, such as a monitor, screen, or other conventional electronic display. Client device 106 also includes a user input device (such as a keyboard, mouse, touchscreen, etc.) capable of receiving input data and / or other information from the user of client device 106 or otherwise associated with such user input device.

[0026] In some implementations, a user (such as a patient, the patient's doctor, or other healthcare provider) manipulates client device 106 to execute client application 108, which supports communication with medical device 102 via network 110. In this regard, client application 108 supports establishing a communication session with medical device 102 on network 110 and receiving data and / or information from medical device 102 via the communication session. Medical device 102 may similarly execute or otherwise implement a corresponding application or process that supports establishing a communication session with client application 108. Client application 108 generally represents a software module or other feature generated or otherwise implemented by client device 106 to support the processes described herein. Therefore, client device 106 typically includes a processing system and a data storage element (or memory) capable of storing programming instructions for execution by the processing system, which, when read and executed, cause the processing system to create, generate, or otherwise facilitate client application 108 and execute or otherwise support the processes, tasks, operations, and / or functions described herein. Depending on the implementation, the processing system may be implemented using any suitable processing system and / or apparatus configured to support the operation of the processing system described herein, such as, for example, one or more processors, central processing units (CPUs), graphics processing units (GPUs), controllers, microprocessors, microcontrollers, processing cores, and / or other hardware computing resources. Similarly, data storage elements or memory may be implemented as random access memory (RAM), read-only memory (ROM), flash memory, magnetic or optical mass storage, or any other suitable non-transitory short-term or long-term data storage or other computer-readable media and / or any suitable combination thereof.

[0027] In one or more embodiments, client device 106 and medical device 102 establish an association (or pair) with each other via network 110 to support the subsequent establishment of a peer-to-peer communication session between medical device 102 and client device 106 via network 110. For example, according to one embodiment, network 110 is implemented as a Bluetooth network, wherein medical device 102 and client device 106 pair by performing a discovery procedure or other suitable pairing procedure (e.g., by acquiring network identification information stored in each other). Pairing information acquired during the discovery process enables either medical device 102 or client device 106 to initiate the establishment of a secure communication session via network 110.

[0028] In one or more exemplary embodiments, client application 108 is also configured to store or otherwise maintain the network address and / or other identification information of remote device 114 on second network 112. In this respect, second network 112 may be physically and / or logically different from network 110, such as, for example, the Internet, a cellular network, a wide area network (WAN), etc. Remote device 114 generally refers to a server or other computing device configured to receive and analyze or otherwise monitor measurement data, event log data, and other information that may be obtained for a patient associated with medical device 102. In an exemplary embodiment, remote device 114 is coupled to database 116, which is configured to store or otherwise maintain data associated with individual patients. In practice, remote device 114 may reside in a location physically different from and / or separate from medical device 102 and client device 106, such as, for example, at a facility owned and / or operated or otherwise attached to the manufacturer of medical device 102. For illustrative purposes, but not limitingly, remote device 114 may alternatively be referred to herein as a server.

[0029] It should be noted that in some implementations, some or all of the functions and processing intelligence of the remote computing device 114 may reside on the medical device 102 and / or other components or computing devices compatible with the patient monitoring system 100. In other words, the patient monitoring system 100 does not necessarily rely on a network-based or cloud-based server deployment (such as...). Figure 1 (As shown in the illustration), although such a deployment may be the most efficient and economical implementation. This disclosure considers these and other alternative arrangements. To this end, some embodiments of system 100 may include additional devices and components that serve as data sources, data processing units, and / or proposed transmission mechanisms. For example, system 100 may include, but is not limited to, any or all of the following elements: computer equipment or systems; patient monitors; healthcare provider systems; data communication devices; etc.

[0030] In an exemplary embodiment, remote device 114 receives patient-specific measurement data values ​​(e.g., sensor glucose measurements, acceleration measurements, etc.) acquired using sensing element 104 from client device 106, and remote device 114 stores or otherwise maintains historical measurement data (e.g., using one or more unique patient identifiers) in association with the patient in database 116. Additionally, remote device 114 may also receive meal data or other event log data that may be input by the patient or otherwise provided (e.g., via client application 108) from or via client device 106, and store or otherwise maintain patient-related historical meal data and other historical event or activity data in database 116. In this regard, meal data may include, for example, a time or timestamp associated with a specific meal event, meal type or other information indicating the content or nutritional characteristics of the meal, and an indication of the size associated with the meal. In an exemplary embodiment, remote device 114 may also receive historical fluid delivery data corresponding to basal or bolus doses of fluid delivered to the patient by infusion devices, syringes, etc. For example, client application 108 may communicate with the infusion device to obtain the insulin delivery dose and corresponding timestamp from the infusion device, and then upload the insulin delivery data to remote device 114 for storage associated with a specific patient. Remote device 114 may also receive geographic location data and potentially other contextual data associated with devices 102, 106 from client device 106 and / or client application 108, and store or otherwise maintain historical operational context data associated with a specific patient. In this regard, one or more of devices 102, 106 may include a Global Positioning System (GPS) receiver, or similar modules, components, or circuitry capable of outputting or otherwise providing data characterizing the respective geographic locations of devices 102, 106 in real time.

[0031] Figure 2 The applicable scheme according to one or more embodiments is shown. Figure 1 An exemplary embodiment of the sensing arrangement 200 of the medical device 102 in the patient monitoring system 100. In this regard, Figure 2 An embodiment in which sensing element 104 is integrated into medical device 102 is illustrated. The illustrated sensing device 200 includes, but is not limited to, a controller 202, sensing element 204 (e.g., sensing element 104), output interface 208, and data storage element (or memory) 208. The controller 202 is coupled to sensing element 204, output interface 208, and memory 206, and the controller 202 is appropriately configured to support the operations, tasks, and / or processes described herein.

[0032] Sensing element 204 generally refers to a component of sensing device 200 configured to generate, produce, or otherwise output one or more electrical signals indicating a condition sensed, measured, or otherwise quantified by sensing device 200. In this regard, the user's physiological condition will affect the characteristics of the electrical signals output by sensing element 204, such that the characteristics of the output signals correspond to or are otherwise related to the physiological condition to which sensing element 204 is sensitive. Sensing element 204 may be implemented as a glucose sensing element that generates an output electrical signal having an associated current (or voltage) related to the interstitial fluid glucose level sensed or otherwise measured in the patient's body by sensing arrangement 200.

[0033] Still referencing Figure 2 Controller 202 generally refers to the hardware, circuitry, logic, firmware, and / or other components of sensing device 200 coupled to sensing element 204 to receive electrical signals output from sensing element 204 and perform various additional tasks, operations, functions, and / or processes described herein. For example, controller 202 may filter, analyze, or otherwise process the electrical signals received from sensing element 204 to obtain measurements for conversion into calibrated measurements of interstitial fluid glucose levels. Additionally, in one or more embodiments, controller 202 may implement or otherwise perform a calibration application that uses calibration data associated with sensing element 204 stored or otherwise maintained in memory 206 to calculate or otherwise determine calibrated measurement parameters based on the measurements, as detailed below. The calibrated measurement parameters can then be used to obtain calibrated measurements of the patient's interstitial fluid glucose levels.

[0034] Depending on the implementation, controller 202 may be implemented or realized using a general-purpose processor, microprocessor, controller, microcontroller, state machine, content-addressable memory, application-specific integrated circuit, field-programmable gate array, any suitable programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. In this regard, the steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be directly embodied in hardware, firmware, software modules executed by controller 202, or any real combination thereof. In an exemplary embodiment, controller 202 includes or otherwise accesses a data storage element or memory 206. Memory 206 may be implemented using any kind of RAM, ROM, flash memory, registers, hard disk, removable disk, magnetic or optical mass storage device, short-term or long-term storage medium, or any other non-transitory computer-readable medium storing programming instructions, code, or other data for execution by controller 202. When the computer-executable programming instructions are read and executed by controller 202, controller 202 performs the tasks, operations, functions, and processes detailed below.

[0035] In some embodiments, controller 202 includes an analog-to-digital converter (ADC) or other similar sampling arrangement that samples or otherwise converts the output electrical signal received from sensing element 204 into one or more corresponding digital measurement data values ​​related to the interstitial fluid glucose level sensed by sensing element 204. For example, in various embodiments, controller 202 may sample, capture, or otherwise analyze the output electrical signal to obtain one or more output measurements affected by the glucose concentration in the patient's interstitial fluid compartment. These output measurements may include one or more currents output by sensing element 204, electrochemical impedance spectroscopy (EIS) values ​​(for one or more frequencies) or other measurements indicating the characteristic impedance associated with sensing element 204, the voltage difference between the electrodes or terminals of sensing element 204 (also referred to herein as the reverse electrode voltage (Vctr)), and / or others. For example, controller 202 may include hardware and firmware co-configured to collect current measurements corresponding to the current through sensing element 204, while simultaneously calculating the reverse electrode voltage (Vctr) and performing electrochemical impedance spectroscopy at various time intervals and at multiple frequencies relative to the current and voltage. In other embodiments, the sensing element 204 may be combined with one or more ADCs and output one or more digital measurements based on analog electrical signals from the sensing element 204.

[0036] Output interface 208 typically represents hardware, circuitry, logic, firmware, and / or other components of sensing arrangement 200 coupled to controller 202 for outputting data and / or information to / from sensing arrangement 200 (e.g., to / from client device 106 and / or remote device 114). In this regard, in one or more exemplary embodiments, output interface 208 is implemented as a communication interface configured to support communication to / from sensing device 200. In such embodiments, communication interface 208 may include or otherwise coupled to one or more transceiver modules capable of supporting wireless communication between medical device 200 and other electronic devices (e.g., client device 106). Alternatively, communication interface 208 may be implemented as a port adapted to receive or otherwise coupled to a wireless adapter including one or more transceiver modules and / or other components supporting the operation of sensing device 200 as described herein. In other embodiments, communication interface 208 may be configured to support wired communication to / from sensing device 200. In other embodiments, the output interface 208 may include or be otherwise implemented as an output user interface element for providing notifications or other information to a user, such as a display element (e.g., a light-emitting diode), a display device (e.g., a liquid crystal display), a speaker or other audio output device, a haptic feedback device, etc. In such embodiments, the output user interface 208 may be integrated with the sensing arrangement 200 (e.g., within a common housing) or implemented separately.

[0037] It should be understood that this article is for illustrative purposes and is not intended to limit the subject matter in any way. Figure 2 This is a simplified representation of the sensing device 200. In this respect, although... Figure 2 Various elements located within sensing device 200 are shown, but one or more elements of sensing device 200 may be different from or otherwise separated from other elements of sensing device 200. For example, sensing element 204 may be separate from and / or physically different from controller 202 and / or communication interface 208. Furthermore, the features and / or functions described herein, as implemented by controller 202, may alternatively be implemented in other devices within the patient monitoring system.

[0038] Figure 3An exemplary block diagram of a situation analysis system 300, which may be implemented by or within a patient monitoring system 100, is shown, providing suggestions, feedback, or other insights regarding a patient's physiological condition. In this regard, the situation analysis system 300 includes a sensing element 302 (e.g., sensing elements 104, 204) having an output coupled to a digitizer 304 configured to convert, transform, or otherwise transform one or more output electrical signals from the sensing element 302 into one or more corresponding digital measurement parameter values. The output of the digitizer 304 is provided to an estimation module 306 configured to receive one or more digital measurement parameter values ​​from the digitizer 304 and convert, transform, or otherwise transform them into estimated measurements of the patient's physiological condition. The output of estimation module 306 is provided to insight engine 308, which analyzes the estimated measurements and may analyze other data and / or information to generate insights, summaries, overviews, recommendations, or other feedback on one or more aspects of the patient's physiological condition. This feedback can then be provided to the patient, healthcare provider, or other users via output interface 310 (e.g., user interface 128 or output interface 208). For example, insight engine 308 may generate user notifications with feedback on the impact of the patient's physiological condition on eating, injections, exercise, sleep, or any other activity (e.g., "Your lunch had a negative impact on your glucose levels," "Your body is responding well after a good night's sleep and exercise," etc.).

[0039] In an exemplary embodiment, one or more of components 304, 306, and 308 may be implemented by or at sensing devices 102, 200, while other components 304, 306, or 308 may be implemented by other devices 106, 114 within the patient monitoring system 100. Similarly, output interface 310 may be associated with any of devices 102, 106, and 114, but may or may not implement one or more of components 304, 306, and 308 itself. For example, in one embodiment, digitization and estimation components 304, 306 are implemented by or at sensing devices 102, 200 (e.g., controllers 122, 202) to output or otherwise provide estimated measurements to client device 106 and / or remote device 114, wherein insight engine 308 is implemented by one of client device 106 or remote device 114. Insights, suggestions, or other notifications generated by the insights engine 308 implemented by either client device 106 or remote device 114 can then be presented by client application 108 at client device 106 or sensing devices 102, 200 (e.g., via user interface 128, 208). However, in some embodiments, the situation analysis system 300 may be implemented entirely by or at sensing devices 102, 200. Accordingly, it should be understood that the subject matter described herein is not intended to be limited to any particular system or architecture for implementing the situation analysis system 300. Furthermore, while the situation analysis system 300 and related topics described herein are not limited to glucose or diabetes, for illustrative purposes, the description of the situation analysis system 300 may be primarily contextualized as providing suggestions, feedback, or other insights regarding the blood glucose status of diabetic patients based on interstitial glucose measurements.

[0040] Still referencing Figure 3 and combined Figure 1 and Figure 2In one or more exemplary embodiments, in the context of interstitial glucose sensing elements 104, 204, 302 used in a diabetic patient, digitizer 304 typically refers to an ADC and other electronic components coupled to interstitial glucose sensing elements 104, 204, 302 to sample, capture, or otherwise process analog output electrical signals generated or otherwise provided by sensing elements 104, 204, 302 in response to interstitial fluid glucose concentration, and to convert those analog electrical signals affected by the patient's interstitial fluid glucose level into corresponding digital measurement parameter values. For example, digitizer 304 may sample the current flowing through interstitial glucose sensing elements 104, 204, 302 and the voltage at one or more electrodes or terminals of sensing elements 104, 204, 302, and then analyze or otherwise process the sampled digital values ​​to generate measurements of current through sensing elements 104, 204, 302 (or isig), electrode voltage (Vctr), and one or more EIS values.

[0041] The estimation module 306 typically represents software, firmware, hardware, and / or other electronic devices that receive digital measurement parameter values ​​from the digitizer 304 and convert them into estimated glucose measurements for the patient. In this regard, the estimation module 306 may implement or otherwise support algorithms for calculating or otherwise determining the estimated glucose measurement based on one or more of the measured current (isig), reverse electrode voltage (Vctr), and measured EIS values. In one embodiment, the estimation module 306 utilizes a model derived through machine learning or other artificial intelligence techniques. For example, referencing… Figure 1 The remote server 114 or other computing devices in the patient monitoring system 100 may utilize neural networks or other machine learning or artificial intelligence techniques to determine, based on relationships between a set of reference blood glucose measurements and combinations of measurement parameter values ​​that are storable or otherwise maintained in the database 116 (e.g., as historical patient data), which combination of digital measurement parameter values ​​from the digitizer 304 is associated with or predicts interstitial fluid glucose concentration. The remote server 114 can then determine, based on those associated measurement parameter values, a corresponding equation, function, or model for calculating the estimated glucose measurement by optimizing the weighting factors assigned to the measurement parameter values ​​and / or the relationships between the measurement parameter values, to minimize the cost function corresponding to the cumulative difference between the model-predicted glucose measurement and the reference blood glucose measurement. The remote server 114 may store or otherwise maintain data defining the glucose estimation model in the database 116 and transmit or otherwise provide this glucose estimation model data to other devices 102, 106 for implementation of the glucose estimation model at devices 102, 106.

[0042] Insight Engine 308 typically refers to software, firmware, hardware, and / or other electronics that receives estimated glucose measurements from a patient and analyzes these estimated glucose measurements in relation to other patient data (e.g., insulin delivery data, dietary data, exercise data, sleep data, geolocation data, event log data, and / or other historical or contextual patient data) to generate one or more recommendations, summaries, analyses, and / or other insights about the patient's glycemic status. In this regard, Insight Engine 308 may employ one or more insight models that characterize the patient's glycemic status based on estimated glucose measurements provided by glucose estimation model 306. Similarly, insight models may be derived using machine learning or other artificial intelligence techniques. For example, neural networks or other machine learning or artificial intelligence techniques may be used to determine which combination of estimated glucose measurements output by glucose estimation model 306 and other contextual patient data (e.g., dietary data, exercise data, sleep data, and / or other activity or event log data) is relevant to or predictive of a patient's specific glycemic status or state. In this regard, in some implementations, the insight model can classify a patient's glycemic status into one of several potential categories in real time based on current or recently estimated glucose measurements and historical, recent, concurrent, and / or contemporaneous contextual data associated with the patient. Based on the set of estimated glucose measurements and the relationships between them and their associated or corresponding patient outcomes and / or contextual data, a remote server 114 can identify which contextual data variables are associated with a specific glycemic state or outcome by incorporating the glucose estimation model output. Then, by optimizing the weighting factors assigned to these variables, based on the glucose estimation model output and the associated contextual data, a corresponding equation, function, or model is determined to calculate or otherwise determine the characteristics of the patient's glycemic status, minimizing the cost function corresponding to the cumulative difference between the model-based predicted results and the observed patient outcomes.

[0043] Reduced sensor measurement error

[0044] Figure 4 An exemplary embodiment of an error mitigation process 400 suitable for integration with sensing devices or patient monitoring systems to mitigate the effects or impacts of various sources of measurement error is shown. In this regard, variations in manufacturing, materials, or components, electromagnetic interference, and / or other factors can introduce noise or error into the measurement signal, which may then propagate through the measurement, thereby reducing the accuracy or reliability of subsequent analysis of the measurement. For example, refer to... Figures 1 to 3Manufacturing variations in sensing elements 104, 204, and 302 may affect the relationship between the analog output electrical signals generated or otherwise provided by each sensing element 104, 204, and 302 and a specific interstitial fluid glucose concentration. This could subsequently introduce noise or error (e.g., by digitizer 304) into one or more output measurement parameters determined based on those electrical signals. Measurement errors in the measurement parameters can then introduce errors into the glucose measurements determined based on them, which could subsequently affect the accuracy or reliability of the analysis and corresponding recommendations, summaries, insights, and / or other aspects of the glucose measurements.

[0045] While manufacturing variations can typically be reduced, doing so can be costly and fail to deliver the desired performance improvements due to other external or environmental sources of measurement error. Accordingly, the error mitigation process 400 and related topics described herein reduce or otherwise mitigate the effects of measurement errors without increasing manufacturing or hardware costs associated with sensing elements 104, 204, 302 or other electronics of sensing devices 102, 200. As detailed below, the error mitigation process 400 identifies one or more input variables of a transformation model, while the output variable used in the transformation model is susceptible to error. The error mitigation process 400 then reduces the influence of the input variables on the transformation model and dynamically updates or otherwise re-establishes the transformation model by re-optimizing the weighting factors assigned to other input variables, while keeping the reduced weights assigned to the identified variables constant. Therefore, the output of the updated transformation model is unaffected by errors relative to the identified variables of interest, thereby improving the accuracy or reliability of the transformation model. Additionally, in some cases, the importance of reducing or eliminating input variables to the conversion model may propagate back to the design or manufacture of sensing elements 104, 204, 302 or other hardware prior to the conversion model (e.g., by relaxing manufacturing tolerances or requirements) to further reduce costs. For example, if the glucose estimation model utilized by estimation module 306 does not require any EIS values, the digital converter 304 can be simplified by removing any hardware, firmware, software, and / or other components that might otherwise be used solely to determine EIS values. While this may primarily be discussed in this paper... Figures 1 to 3 The error mitigation process 400 is described in the context of monitoring blood glucose levels in diabetic patients. However, it should be understood that the error mitigation process 400 is not limited to glucose sensing, glucose measurement, CGM, diabetes management, etc., and in practice, the error mitigation process 400 can be implemented in an equivalent manner in any type of sensing element or monitoring system.

[0046] The various tasks performed in conjunction with the error mitigation process 400 can be executed by hardware, firmware, software, or any combination thereof via processing circuitry. For illustrative purposes, the following description may refer to the above-mentioned combinations. Figures 1 to 3 The mentioned components. It should be understood that the error mitigation process 400 may include any number of additional or alternative tasks, which need not be performed in the order shown and / or these tasks may be performed simultaneously, and / or the error mitigation process 400 may be incorporated into a more comprehensive procedure or process with other functions not described in detail herein. Furthermore, as long as the intended overall functionality remains intact, Figure 4 One or more of the tasks shown and described in the context can be omitted from the practical implementation of the error mitigation process 400.

[0047] refer to Figure 4 And continue to refer to Figures 1 to 3 The error mitigation process 400 illustrated is initialized or otherwise initiated by identifying or otherwise acquiring error profiles associated with the input variables of interest to be analyzed relative to the transformation model of interest (or a combination or sequence of transformation models), which is the subject of this error mitigation process (task 402). In this regard, the transformation model used to analyze a particular case is identified, selected, or otherwise determined to achieve error immunity or mitigation, wherein the individual input variables to the selected transformation model (or input variables to the transformation model preceding the selected transformation model) are also identified, selected, or otherwise determined for further analysis. When the selected transformation model is downstream of other transformation models (e.g., the glucose estimation model implemented by estimation module 306 or the transformation model implemented by digitizer 304), the error mitigation process 400 can be performed relative to the input variables to the upstream model. In one embodiment, an administrator can command, signal, or otherwise instruct remote server 114 to initiate or otherwise perform the error mitigation process 400 relative to a specific insight model implemented by a previously developed insight engine 308. However, in other implementations, the error mitigation process 400 may be automatically executed as a follow-up process after the development of a new transformation model or the updating of an existing transformation model, to provide less noise or higher error immunity, or otherwise improve the robustness of the model before deployment. After identifying the transformation model, the remote server 114 may selectively or otherwise identify an input variable of the transformation model for analysis.

[0048] In one or more exemplary embodiments, remote server 114 calculates or otherwise determines an error profile for a selected variable of interest based on data maintained in database 116. In this regard, the error profile includes one or more error measures indicating measurement error, noise, or other fluctuations associated with the input variable. Depending on the embodiment, the error profile may include one or more standard deviations of the standard deviations associated with the input variable, a probability distribution associated with the input variable, a normal distribution of the input variable, and / or others. For example, when the input variable of interest is the measured current (isig), remote server 114 may calculate the error profile based on the relationship between an estimated sensor glucose measurement calculated from the measured current and corresponding reference blood glucose measurement data maintained in database 116. In yet another embodiment, the error profile may be calculated or otherwise determined during the manufacture or processing of sensing elements 104, 204, 302. For example, instances of sensing elements 104, 204, 302 may be exposed to the same reference glucose concentration, wherein different electrical signals output by different instances of sensing elements 104, 204, 302 for the same reference glucose concentration are analyzed to determine an error profile for a given type, configuration, manufacture, and / or model of sensing elements 104, 204, 302.

[0049] After identifying the error profile of the input variable of interest, error mitigation process 400 extracts or otherwise obtains a reference value for the input variable of interest and uses the error profile to generate an adjusted value for the input variable based on the reference value (tasks 404, 406). In this respect, the error profile is used to alter, jitter, or otherwise introduce perturbations to the reference value to produce an adjusted value from the reference value, where the adjusted value simulates the potential measurement error relative to the input variable. For example, when the input variable of interest is the measured current (isig), remote server 114 may retrieve historical measured current (isig) values ​​previously collected consistently for different patients from database 116 and calculate or otherwise determine an adjusted current value (e.g., using Monte Carlo simulation based on the historical distribution of the measured current) that deviates from the historical value, reflecting the approximate or possible range of potential deviations taken into account in the associated error profile.

[0050] Still referencing Figure 4The error mitigation process 400 uses reference values ​​of the input variables to calculate or otherwise determine a reference output for the transformation model to continue (task 408). In this regard, historical or current measurements of the input variables can be input or otherwise provided to the transformation model, while no other concurrent or related historical or actual values ​​of other input variables are currently analyzed. The transformation model then calculates or otherwise determines the output based on the input combination of reference values ​​of the input variables and concurrent or simultaneous values ​​of other input variables. For example, a reference value of the measured current (isig) can be input or otherwise provided to the glucose estimation model used by the estimation module 306 along with simultaneous values ​​of other measurement parameters output by the digitizer 304 to obtain a reference set of estimated measurements for the glucose estimation model. In an embodiment of the error mitigation process 400 performed by an insight model downstream of a glucose estimation model, a reference set of estimated measurements may be input or otherwise provided to the insight model, which is then used to calculate or otherwise determine reference output data for the insight model based on the reference estimated measurements and corresponding historical or actual values ​​of other input variables to the insight model (e.g., concurrent or periodic dietary data or other event log data).

[0051] Similarly, error mitigation process 400 uses simulated values ​​of input variables to calculate or otherwise determine the simulated output of the transformation model to continue (task 410). In this respect, adjusted values ​​of input variables are input to or otherwise provided to the transformation model, while the values ​​of other input variables not currently analyzed to the transformation model are the same as those used to determine the reference output. For example, adjusted values ​​of current may be input to or otherwise provided to the glucose estimation model used by estimation module 306 along with simultaneous values ​​of other measurement parameters output by digitizer 304 to obtain a simulated set of measurements for the glucose estimation model that represents or otherwise demonstrates the possible impact of measurement error relative to the measured current (isig) input variable. Furthermore, in an embodiment of error mitigation process 400 performed with respect to an insight model downstream of the glucose estimation model, the simulated set of measurements may be input to or otherwise provided to the insight model, which is then used to calculate or otherwise determine the simulated output data of the insight model based on the simulated measurements and the same values ​​of other input variables to the insight model, which are used to generate reference output data for the insight model. Therefore, the simulated output data represents, or otherwise demonstrates, the potential impact of measurement error relative to the lower level on the higher level output of the insight model of the measured current (isig) input variable.

[0052] Still referencing Figure 4The error mitigation process 400 continues by comparing or otherwise analyzing the relationship between the reference model output and the simulated model output to identify or otherwise determine when the difference between the model output sets exceeds a performance threshold (Task 412). When the deviation between the model output sets exceeds the performance threshold, the error mitigation process 400 calculates or otherwise determines a reduction in the weights or effects applied to the input variables in one or more transformation models, and then updates the transformation model based on the reduced weights of the input variables by re-optimizing the weights of the other input variables (Tasks 414, 416). Thereafter, the error mitigation process 400 repeats the following steps: using reference values ​​of the input variables with the reduced weights assigned to the input variables to determine the reference output of the updated transformation model; using adjusted values ​​of the input variables with the reduced weights assigned to the input variables to determine the simulated output of the updated transformation model; and comparing the simulated output with the reference output (Tasks 408, 410, 412). Thus, the error mitigation process 400 can progressively or iteratively adjust (e.g., increase or decrease) the weights of the input variables (e.g., by using predetermined or analytically determined variable increment values) until the transformed model (or a combination or sequence thereof) exhibits the desired level of immunity relative to the noise or error in the input variables.

[0053] For example, when the conversion model of interest is a glucose estimation model to be used by estimation module 306, remote server 114 may calculate or otherwise determine one or more statistics characterizing the difference between reference glucose values ​​and simulated glucose values, and then verify or otherwise confirm that the values ​​of those statistics are less than their respective performance thresholds (e.g., task 412). For example, remote server 114 may calculate the percentage of probability of correct or false recognition based on the differences between each pair of contemporaneous reference and simulated glucose values, and when the correct or false recognition metric is greater than an acceptable threshold, remote server 114 may calculate or otherwise determine the amount of weight reduction for the measured current (isig) input variable to the glucose estimation model. In this regard, the amount of weight reduction may be calculated or otherwise determined based on the magnitude of the difference between one or more statistics and their performance thresholds. For example, if the calculated value of the correct or false recognition metric is 50% greater than the performance threshold, remote server 114 may calculate or otherwise determine an updated weight for the measured current (isig) input variable that is 50% of the previous weight (e.g., task 414). Subsequently, the remote server 114 may calculate or otherwise determine an updated version of the glucose estimation model that maintains the weights of the measured current (isig) input variable at a determined reduced weight value, while modifying the weights of other input variables in the glucose estimation model to use the reduced weights to achieve the desired accuracy or reliability of the glucose estimation model.

[0054] When the model of interest is an insight model to be used by the insight engine 308, the remote server 114 may similarly calculate or otherwise determine one or more statistics characterizing the difference between the reference output and the simulated output of the insight model. For example, when the insight model is a classifier or classification model, the remote server 114 may calculate or otherwise determine the percentage or ratio of consistency (or conversely, the ratio of inconsistency) between the reference output and the simulated output. In this regard, when the consistency ratio is less than a threshold consistency ratio (or the inconsistency ratio is greater than a threshold inconsistency ratio), the remote server 114 determines that the difference between the reference output and the simulated output of the insight model exceeds a desired performance threshold. Similarly, the remote server 114 may calculate or otherwise determine the weights of the input variables based on reduced bias, and then dynamically determine an updated version of the insight model and / or a downstream glucose estimation model used with the insight model using the reduced weights. In this regard, the updated insight model can weight glucose measurements provided by the estimation module 306 relative to other input variables of the insight model (e.g., dietary data, event log numbers, and / or others) compared to previous iterations of the insight model (e.g., by reducing or increasing the impact of glucose measurements relative to other activity or contextual data). In such embodiments, the weights of lower-level input variables to lower-level models, such as the glucose estimation model, can be progressively and iteratively reduced until the individual input variables and corresponding weights for the combination or sequence of the lower-level glucose estimation model and the higher-level insight model are achieved, in order to achieve the desired measurement error immunity for the higher-level insight model.

[0055] After implementing one or more updated transformation models that achieve the desired level of error mitigation relative to specific input variables of interest, the error mitigation process 400 sends, provides, or otherwise deploys the updated transformation models for future use (task 418). In one or more embodiments, after determining updated glucose estimation models and / or insight models representing improved error immunity, remote server 114 may automatically push or otherwise provide these models to other devices 102, 106 for use by or at those devices. For example, after determining an updated glucose estimation model that reduces the weight of specific measurement parameter values ​​output by digital converter 304, remote server 114 may automatically (directly or via intermediate client device 106) push or otherwise send the updated estimation module to sensing devices 102, 200 for implementation of the updated glucose estimation model by estimation module 306 at sensing devices 102, 200. In this respect, previous glucose estimation model data (e.g., input variable weighting factors, etc.) stored in memories 124, 206 can be overwritten with updated glucose estimation model data (e.g., updated input variable weighting factors with the amount of reduction in the input variable of interest) for subsequent reference by controllers 122, 202 when implementing estimation module 306 for digital measurement parameter values ​​derived from the output electrical signals of sensing elements 104, 204, 302. Similarly, after determining the updated insight model, remote server 114 can automatically push or otherwise send the updated insight module to client device 106 for implementation of the updated insight model by insight engine 308 at client device 106. In this respect, insight engine 308 can be implemented as a feature or component of client application 108 for providing insights, suggestions, etc., to the patient on the patient's client device 106.

[0056] Combination Figures 1 to 3 refer to Figure 4In one or more embodiments, an error mitigation process 400 is performed relative to the glucose estimation model adopted by estimation module 306. In this regard, for input variables of the glucose estimation model, such as the measured current (isig), an error profile or error metric associated with the measured current is calculated, derived, or otherwise identified based on historical measured current data maintained in database 116, to generate an adjusted set of measured current data based on the historical measured current dataset maintained in database 116. This historical measured current data is used as a reference measured current value, which is input or otherwise provided to the glucose estimation model to obtain a reference set of estimated sensor glucose measurements based on the reference measured current value and possibly based on one or more other input variables (e.g., the value of the historical reverse electrode voltage of the same period). This adjusted set of measured current values ​​is similarly input or otherwise provided to the glucose estimation model to obtain a simulated set of estimated sensor glucose measurements based on the adjusted measured current values, while simultaneously using the same concurrent values ​​for other input variables to the glucose estimation model. Therefore, this simulated set of estimated sensor glucose measurements represents the relative impact of noise or error in the measured current input to the glucose estimation model. When the cumulative difference between the simulated set of estimated sensor glucose measurements and the reference set of estimated sensor glucose measurements exceeds an acceptable threshold, the glucose estimation model is updated or modified to reduce the difference by decreasing the weights associated with the measured current input variable. For example, error mitigation process 400 may iteratively reduce or otherwise adjust the weight factors associated with the measured current input variable and iteratively update the glucose estimation model (e.g., tasks 408, 410, 412, 414, 416) until the cumulative difference between the simulated set of estimated sensor glucose measurements and the reference set of estimated sensor glucose measurements is less than an acceptable threshold. The resulting updated glucose estimation model, which has higher noise immunity relative to the measured current, may be automatically pushed to or otherwise deployed to instances of sensing devices 102, 200 by remote server 114 for implementation by estimation module 306 in place of the previously exhibited glucose estimation model sensitive to unacceptable noise.

[0057] Figure 5An exemplary block diagram of a patient insight model 500, which may be implemented or otherwise employed by the insight engine 308, is shown. The insight model 500 includes a plurality of input variables 502 that are calculated or otherwise determined based on estimated sensor glucose measurements output by the estimation module 306. For example, the calculated input variables 502 may include the average difference between the estimated sensor glucose measurements and a target glucose value, a percentage of a time range calculated based on the percentage or duration of sensor glucose measurements within the target range of glucose values, the area under the curve (iAUC) calculated based on the estimated sensor glucose measurements, etc. The insight model 500 then applies a normalization model 504 to each calculated input variable 502 to convert the individual calculated input variables into normalized values ​​between 0 and 1. Then, before inputting or otherwise providing the weighted, normalized input variables to the insight formula 508, the normalized input variables are reduced or otherwise adjusted by various weighting factors 506, and the insight formula calculates or otherwise determines the value of the insight measure based on the weighted, normalized input variables. The calculated insight measure can then be used to score or otherwise categorize aspects of the patient's glycemic status.

[0058] refer to Figures 1 to 3 refer to Figure 5In one or more embodiments, an error mitigation process 400 may be performed relative to the insight model 500 employed by the insight engine 308 to adjust the weighting factors 506 of various variables associated with a particular computed input variable 502. In this regard, error profiles or error measures associated with sensor glucose measurements are calculated, derived, or otherwise identified based on historical sensor glucose measurement data maintained in database 116 for use in generating an adjusted set of sensor glucose measurement data based on the historical sensor glucose measurement dataset maintained in database 116. These historical sensor glucose measurements can be used to compute one or more reference values ​​for the computed input variable 502 of the insight model 500, and, for example, by computing one or more reference values ​​for other input variables 502 based on historical sensor glucose measurements, and by applying various normalization models 504 and variable weights 506 to the respective reference computed input variable values ​​before putting the weighted and normalized reference computed input variable values ​​into the insight formula 508 to obtain reference output values, the insight model can then be used to compute or otherwise determine one or more reference output values ​​for the insight formula 508. Similarly, the sensor glucose adjustment measurement can be used to calculate one or more adjustment values ​​for the calculated input variable 502 of insight model 500, and, for example, by calculating one or more reference values ​​for other input variables 502 based on historical sensor glucose measurements, and by applying respective normalization models 504 and variable weights 506 to the respective calculated input variable values ​​before putting the weighted and normalized input variable values ​​into insight formula 508 to obtain simulated output values, the insight model can then be used to calculate or otherwise determine one or more simulated output values ​​of insight formula 508. In this respect, the simulated output of insight formula 508 represents the sensitivity of each calculated input variable of interest 502 to the effects or influence of noise, error, or other fluctuations in the estimated sensor glucose measurement provided by estimation model 306.

[0059] When the cumulative difference between the simulated output value of insight formula 508 and the reference output value of insight formula 508 exceeds an acceptable threshold, one or more aspects of insight model 500 may be adjusted or otherwise modified to reduce the difference. For example, the weight factors 506 associated with each calculated input variable of interest may be reduced or otherwise adjusted to reduce the weight factors associated with the calculated input variable of interest 502. Insight model 500 may then be updated to reflect the reduced weight factors, for example, by altering the equation or function of insight formula 508 and / or modifying the weights 506 associated with other calculated input variables through artificial intelligence or other machine learning techniques to obtain an updated version of insight model 500. For example, given a reference input signal and a modulated input signal, a model may be trained to reduce the difference between a reference output generated based on the reference input signal and an simulated output generated based on the modulated input signal. In this way, the machine learning algorithm can learn to reduce the weights assigned to the input variables most prone to error, thereby reducing the deviation or error between the simulated model output and the reference model output. Similarly, the weights 506 associated with the computed variable of interest 502 can be iteratively adjusted, with other weights 506 and / or insight formulas 508 being updated iteratively in a corresponding manner (e.g., tasks 408, 410, 412, 414, 416) until the cumulative difference between the simulated insight output and the reference insight output is less than an acceptable threshold. This acceptable threshold can be determined based on an error amount or difference value that would cause the output derived by the insight model to change (e.g., the error amount would cause the insight model output to change from "Your diet is good" to "Your diet is poor"). The resulting updated insight model, which has higher noise immunity relative to the computed variable of interest, can be automatically pushed by remote server 114 to or otherwise deployed to instances of sensing devices 102, 200, or other devices 106 so that it can be implemented by insight engine 308 instead of the previously configured insight model 500, which exhibited sensitivity to unacceptable noise.

[0060] In other embodiments, additionally or alternatively, an error mitigation process 400 may be performed relative to the insight model 500 to adjust or otherwise modify the normalization model 504 associated with each calculated input variable 502, either alone or in conjunction with adjusting the weighting factors 506 associated with each calculated input variable 502. Furthermore, although the implementation of the error mitigation process 400 for each variable has been described above, in practice, the error mitigation process 400 may be implemented simultaneously with respect to all calculated input variables 502 in an equivalent manner. For example, one or more adjusted values ​​for each calculated input variable 502 of the insight model 500 may be calculated using sensor glucose adjusted measurements before applying the individual normalization models 504 and the weights 506 of the variables to the adjusted input variable values ​​before feeding the weighted and normalized input variable values ​​into the insight formula 508 to obtain the simulated output values. In such embodiments, the simulated output represents the overall sensitivity of the insight model 500 to noise, errors, or other fluctuations in the estimated sensor glucose measurements provided by the estimation model 306 within the calculated input variables 502. In such implementations, the error mitigation process 400 iteratively adjusts one or more of the weighting factor 506, insight formula 508, and / or normalization model 504 to obtain an updated configuration of the insight model 500 that is less susceptible to sensor glucose measurement errors or noise within the range of the calculated input variables of the insight model 500.

[0061] Still referencing Figures 1 to 5 It should be noted that in some implementations, an error mitigation process 400 may be performed relative to a glucose estimation model implemented simultaneously by estimation model 306 and insight model 500. For example, adjusted measurement current data that can be input to estimation model 306 to obtain simulated sensor glucose measurements may be obtained using the error associated with the measurement current (isig) input variable to estimation model 306. This simulated sensor glucose measurement can then be input to insight model 500 and used to obtain an adjustment value for one or more calculated input variables 502, which is then used to calculate or otherwise derive a simulated output of insight model 500 representing the sensitivity of insight model 500 to noise, errors, or other fluctuations associated with lower-level measurements or input variables to lower-level estimation model 306. In this case, the error mitigation process 400 can be executed by the server 114 to iteratively adjust the glucose estimation model and insight model 500 to be implemented at the estimation model 306, which work together, until a configuration is obtained in which the glucose estimation model and insight model 500 implement the output of the insight engine 308 or insight model 500, which is able to exhibit the desired level of immunity to noise, error or other fluctuations relative to the measured current.

[0062] For the sake of brevity, conventional techniques related to glucose sensing and / or monitoring, sampling, filtering, calibration, and other functional aspects of this subject matter will not be described in detail here. Additionally, certain terms may be used herein for reference only and are not intended to be limiting. For example, terms such as “first,” “second,” and other such numerical terms relating to structures do not imply order or sequence unless the context clearly indicates otherwise. The description above may also refer to elements, nodes, or features that are “connected” or “coupled” together. As used herein, unless otherwise explicitly stated, “coupled” means that one element / node / structure is directly or indirectly combined (or directly or indirectly connected) with other elements / nodes / structures, and is not necessarily mechanically connected.

[0063] While at least one exemplary embodiment has been presented in the foregoing detailed description, it should be understood that numerous variations exist. It should also be understood that the one or more exemplary embodiments described herein are not intended to limit the scope, applicability, or configuration of the claimed subject matter in any way. For example, the subject matter described herein is not necessarily limited to the infusion devices and related systems described herein. Furthermore, the foregoing detailed description will provide those skilled in the art with a convenient roadmap for implementing the one or more described embodiments. It should be understood that various changes to the function and arrangement of elements can be made without departing from the scope defined by the claims, including equivalents known and foreseeable at the time of filing this patent application. Therefore, the details of the foregoing exemplary embodiments or other limiting aspects should not be construed as part of the claims unless expressly intended to the contrary.

Claims

1. A system comprising: One or more processors; and One or more processor-readable media storing instructions that, when executed by the one or more processors, cause the following operations to be performed: Identify an error metric associated with an input variable that is associated with a transformation model that provides an output glucose value that is influenced by the value of the input variable and the weights applied to the input variable; The reference output glucose value of the conversion model is determined by providing the reference input value of the input variable to the conversion model; The error metric is used to generate an adjustment value for the input variable based on the reference input value; The simulated output glucose value of the conversion model is determined by providing the adjusted values ​​of the input variables to the conversion model; When the difference between the simulated output glucose value and the reference output glucose value is greater than a threshold, the conversion model is updated using the reduced weights applied to the input variable, thereby obtaining an updated conversion model that reduces the error associated with the input variable. The updated conversion model provides an output glucose value based on one or more subsequent values ​​of the input variable derived from one or more electrical signals output by an instance of a sensing element capable of providing an electrical signal influenced by the patient's glucose level. as well as Provide notifications related to the patient's glucose levels, wherein the notifications are generated at least in part based on the output glucose values ​​of the updated conversion model.

2. The system of claim 1, wherein the conversion model includes an estimation model for providing estimated glucose values, wherein: Determining the reference output glucose value includes using a reference set of input values ​​to determine the glucose measurement value; Determining the simulated output glucose value includes using the adjustment value to determine a set of simulated glucose measurements; and Updating the transformation model involves updating the estimation model by applying the reduced weights to the input variables in the estimation model to reduce the difference between the simulated set of glucose measurements and the reference set of glucose measurements.

3. The system according to claim 2, wherein: The sensing element includes an interstitial glucose sensing arrangement; The estimation model includes a sensor glucose estimation model for providing estimated glucose values; The reference set of glucose measurements includes a reference set of estimated glucose measurements determined by the sensor glucose estimation model based on the reference input values ​​and the weights applied to the input variables; and The simulated set of glucose measurements includes a simulated set of estimated glucose measurements determined by the sensor glucose estimation model based on the adjustment value and the weights applied to the input variables.

4. The system of claim 2, wherein the transformation model includes an insight model for generating the notification based on an estimate of the patient's glucose level provided by an estimation model having weights assigned to the input variables, based on an aspect of the glucose level in the patient's body.

5. The system of claim 4, wherein updating the conversion model comprises: The estimated model is updated by using the reduced weights assigned to the input variables, thereby obtaining an updated estimated model; as well as The insight model used to generate the notification is updated based on the estimated glucose value provided by the updated estimation model with the reduced weights applied to the input variables.

6. The system of claim 1, wherein updating the conversion model comprises: Reduce the weights applied to the input variables. The updated reference output is determined at least in part based on the reduced weights, and The updated analog output is determined at least in part based on the reduced weight applied to the adjustment value, such that the difference between the analog output and the reference output is less than the threshold.

7. The system of claim 3, wherein the input variables include output current, electrode voltage, or electrochemical impedance spectroscopy (EIS) values ​​determined based on electrical signals provided by the interstitial glucose sensing arrangement.

8. A processor-implemented method, the processor-implemented method comprising: Identify an error metric associated with an input variable that is associated with a transformation model that provides an output glucose value that is influenced by the value of the input variable and the weights applied to the input variable; The reference output glucose value of the conversion model is determined by providing the reference input value of the input variable to the conversion model; The error metric is used to generate an adjustment value for the input variable based on the reference input value; The simulated output glucose value of the conversion model is determined by providing the adjusted values ​​of the input variables to the conversion model; When the difference between the simulated output glucose value and the reference output glucose value is greater than a threshold, the conversion model is updated using the reduced weights applied to the input variable, thereby obtaining an updated conversion model that reduces the error associated with the input variable. The updated conversion model provides an output glucose value based on one or more subsequent values ​​of the input variable derived from one or more electrical signals output by an instance of a sensing element capable of providing an electrical signal influenced by the patient's glucose level. as well as Provide notifications related to the patient's glucose levels, wherein the notifications are generated at least in part based on the output glucose values ​​of the updated conversion model.

9. The method of claim 8, wherein the conversion model comprises an estimation model for providing estimated glucose values, wherein: Determining the reference output glucose value includes using a reference set of input values ​​to determine the glucose measurement value; Determining the simulated output glucose value includes using the adjustment value to determine a set of simulated glucose measurements; and Updating the transformation model involves updating the estimation model by applying the reduced weights to the input variables in the estimation model to reduce the difference between the simulated set of glucose measurements and the reference set of glucose measurements.

10. The method according to claim 9, wherein: The sensing element includes an interstitial glucose sensing arrangement; The estimation model includes a sensor glucose estimation model for providing estimated glucose values; The reference set of glucose measurements includes a reference set of estimated glucose measurements determined by the sensor glucose estimation model based on the reference input values ​​and the weights applied to the input variables; and The simulated set of glucose measurements includes a simulated set of estimated glucose measurements determined by the sensor glucose estimation model based on the adjustment value and the weights applied to the input variables.

11. The method of claim 9, wherein the transformation model comprises an insight model for generating the notification based on an estimate of the patient's glucose level provided by an estimation model having weights assigned to the input variables, based on an aspect of the glucose level in the patient's body.

12. The method of claim 11, wherein updating the transformation model comprises: The estimated model is updated by using the reduced weights assigned to the input variables, thereby obtaining an updated estimated model; as well as The insight model used to generate the notification is updated based on the estimated glucose value provided by the updated estimation model with the reduced weights applied to the input variables.

13. The method of claim 8, wherein updating the transformation model comprises: Reduce the weights applied to the input variables. The updated reference output is determined at least in part based on the reduced weights, and The updated analog output is determined at least in part based on the reduced weight applied to the adjustment value, such that the difference between the analog output and the reference output is less than the threshold.

14. The method of claim 10, wherein the input variables include output current, electrode voltage, or electrochemical impedance spectroscopy (EIS) values ​​determined based on electrical signals provided by the interstitial glucose sensing arrangement.

Images