Method and electronic device for artifact detection

Abstract

The application discloses a method and an electronic device for artifact detection. The method comprises: when an analyte fluctuation event of a target object in a non-motion state is detected, acquiring physical motion characteristics and physiological signal characteristics of the target object; determining an artifact confidence degree according to the physical motion characteristics and the physiological signal characteristics; determining whether the analyte fluctuation event is an artifact based on the artifact confidence degree, and suppressing a warning event corresponding to the analyte fluctuation event in the case that the analyte fluctuation event is an artifact. The application solves the technical problem that related technologies are difficult to identify false analyte value anomaly events caused by physical compression sensors, and false alarms are easily triggered.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and more specifically, to a method and electronic device for artifact detection. Background Technology

[0002] Continuous monitoring devices (such as CGM and continuous insulin monitors) continuously monitor the concentration of analytes in interstitial fluid through subcutaneous sensors. In scenarios such as sleep at night, users often experience unconscious body movements (such as turning over) or localized pressure, which can cause artifacts in the analyte signal. However, the signal processing algorithms used in these technologies struggle to effectively distinguish between signal anomalies caused by physical pressure on the sensor and genuine physiological fluctuations (such as hypoglycemia or hypoinsulinemia), easily triggering false alarms.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides a method and electronic device for artifact detection, which at least solves the technical problem that related technologies have difficulty in identifying false abnormal analysis values ​​caused by physical pressure sensors, leading to a high risk of triggering false alarms.

[0005] According to one aspect of the embodiments of this application, a method for artifact detection is provided, comprising: when an analyte fluctuation event is detected in a non-motion state of a target object, acquiring physical motion characteristics and physiological signal characteristics of the target object, wherein the physical motion characteristics are used to characterize the body motion state of the target object, and the physiological signal characteristics are used to characterize the dynamic trend of the analyte value of the target object changing over time; determining an artifact confidence level based on the physical motion characteristics and physiological signal characteristics, wherein the artifact confidence level is used to quantify the probability that the analyte fluctuation event is a false analyte value reduction event caused by the body motion behavior of the target object; determining whether the analyte fluctuation event is an artifact based on the artifact confidence level, wherein, in the case that the analyte fluctuation event is an artifact, suppressing a warning event corresponding to the analyte fluctuation event.

[0006] According to another aspect of the embodiments of this application, an artifact detection apparatus is also provided, comprising: a detection module, configured to acquire physical motion characteristics and physiological signal characteristics of the target object when an analyte fluctuation event is detected in a non-motion state of the target object, wherein the physical motion characteristics are used to characterize the body motion state of the target object, and the physiological signal characteristics are used to characterize the dynamic trend of the analyte value of the target object changing over time; a determination module, configured to determine an artifact confidence level based on the physical motion characteristics and physiological signal characteristics, wherein the artifact confidence level is used to quantify the probability that the analyte fluctuation event is a false analyte value reduction event caused by the body motion behavior of the target object; and an execution module, configured to determine whether the analyte fluctuation event is an artifact based on the artifact confidence level, wherein, in the case that the analyte fluctuation event is an artifact, a warning event corresponding to the analyte fluctuation event is suppressed.

[0007] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the above-described method for artifact detection.

[0008] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, the non-volatile storage medium including a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-described artifact detection method by running the computer program.

[0009] According to another aspect of the embodiments of this application, a computer program product is also provided, including computer instructions that, when executed by a processor, implement the above-described method for artifact detection.

[0010] In this embodiment, an artifact confidence assessment method based on the fusion of physical motion characteristics and physiological signal characteristics of analyte fluctuations is adopted. By quantifying the likelihood of false reductions in analyte fluctuation events caused by body movement, and suppressing corresponding warning events based on the confidence assessment results, the method achieves the goal of accurately distinguishing between physical pressure interference and real physiological abnormalities in non-motor states such as sleep. This reduces the false alarm rate and ensures the accuracy of monitoring data, thereby solving the technical problem that related technologies have difficulty identifying false abnormal analyte values ​​caused by physical pressure sensors, which easily triggers false alarms. Attached Figure Description

[0011] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0012] Figure 1This is a hardware structure block diagram of a computer terminal for an artifact detection method according to an embodiment of this application;

[0013] Figure 2 This is a flowchart of a method for artifact detection according to an embodiment of this application;

[0014] Figure 3 This is a schematic diagram of a time window for an artifact detection method according to an embodiment of this application;

[0015] Figure 4 This is a system architecture diagram of an artifact detection method according to an embodiment of this application;

[0016] Figure 5 This is a schematic diagram of an artifact detection device according to an embodiment of this application. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] To better understand the embodiments of this application, the technical terms involved in the embodiments of this application are explained below:

[0020] Physical motion characteristics: refers to data features captured by motion sensors such as accelerometers that reflect the physical state of the target object's body surface, including but not limited to silent background activity, transient impact amplitude, high-frequency low-amplitude tremor frequency and body position angle, etc. In the embodiments of this application, it is used to characterize whether the user has physical behaviors such as turning over, pressing or slight tremors that may cause abnormal sensor signals in a non-motion state (such as during sleep).

[0021] Physiological signal characteristics: refers to dynamic trend data captured by continuous monitoring sensors that reflect the changes in the value of analytes in the target object over time, including but not limited to concentration values, decline slope, V-shaped reversal pattern (rapid drop and rebound), and recovery rate. In the embodiments of this application, it is used to quantify the actual change pattern of analyte concentration and combine it with physical motion characteristics to determine whether the signal drop is caused by physiological reasons (such as true hypoglycemia) or physical reasons (such as compression).

[0022] Artifact Confidence: A quantitative value between 0 and 1, used to indicate the degree of probability that the currently detected analyte fluctuation event is a false decrease event caused by the bodily movement behavior of the target object (such as compression). In the embodiments of this application, for example, it can be calculated by weighted fusion of multimodal features, and differentiated processing strategies (such as alarm suppression, data labeling, or triggering calibration) are performed according to its numerical range.

[0023] Continuous monitoring devices (including continuous glucose monitors and continuous insulin monitors) use sensors implanted under the skin to track the concentration of analytes in the interstitial fluid of target tissues in real time and transmit the acquired concentration data and corresponding timestamps to a smart terminal. However, in a non-movement state (taking nighttime sleep as an example), unconscious physical movements of the user (such as turning over) or local pressure (such as squeezing the sensor location when lying on one's side) can easily cause artifact interference in the analyte monitoring signal.

[0024] The signal processing algorithms used in related technologies suffer from the following significant technical bottlenecks in practical applications:

[0025] (1) High false alarm rate of pressure artifacts: When the sensor is subjected to physical pressure, local microvascular blood flow is obstructed or tissue fluid flow is restricted, resulting in a temporary decrease in the concentration of analytes in the interstitial fluid. It is difficult to effectively distinguish between this "signal drop caused by physical pressure" and the real "physiological low concentration crisis" (such as hypoglycemia, hypoinsulin and other critical conditions), which is very likely to trigger false alarms at night, seriously disturbing the user's normal sleep or causing alarm fatigue, causing the user to ignore the real critical alarm.

[0026] (2) Difficulty in identifying silent body movement: Unlike strenuous exercise such as running, compression during sleep is often accompanied by stillness or small high-frequency tremors. The filtering methods based on the intensity threshold of the related technologies are mainly designed for high-energy strenuous exercise and cannot effectively identify such low-energy "silent compression" state, thus failing to accurately remove signal interference caused by body movement.

[0027] (3) Lack of multidimensional feature fusion mechanism: Related technologies usually rely on a single dimension (such as monitoring only the rate of change of analyte or only the acceleration energy) for judgment. They lack a discrimination model that deeply couples "physiological signal trend" with "physical motion characteristics", which makes it impossible to accurately quantify "confidence of artifact occurrence" and difficult to retain real low-concentration events while removing artifacts.

[0028] (4) Lack of a quantitative mechanism for the confidence of artifacts: Related technologies often use hard threshold judgment (such as judging motion when the acceleration amplitude exceeds a certain value), which cannot give the probability of the current signal being squeezed and interfered with, resulting in rigid processing in critical situations, either missing or false alarms, and lacking a probabilistic assessment of the possibility of artifacts occurring.

[0029] To address the aforementioned technical problems, this application provides corresponding solutions, which are detailed below.

[0030] The artifact detection method embodiments provided in this application can be executed on mobile terminals, computer terminals, or similar computing devices. Figure 1 A hardware block diagram of a computer terminal for implementing a method for artifact detection is shown. Figure 1 As shown, the computer terminal 10 may include one or more processors (shown as 102a, 102b, ..., 102n in the figure) (the processor may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission module 106 for communication functions connected via wired and / or wireless networks. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0031] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0032] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the artifact detection method in this embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby implementing the aforementioned artifact detection method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0033] The transmission module 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network interface controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission module 106 may be a radio frequency (RF) module, used for wireless communication with the Internet.

[0034] The display can be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10.

[0035] It should be noted here that, in some optional embodiments, the above... Figure 1 The computer terminal shown may include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that... Figure 1 This is only one instance of a specific particular instance, and is intended to illustrate the types of components that may exist in the aforementioned computer terminal.

[0036] In the above operating environment, this application provides a method embodiment for artifact detection. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0037] Figure 2 This is a flowchart of a method for artifact detection according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following steps:

[0038] Step S202: When an analyte fluctuation event is detected in the target object in a non-motion state, the physical motion characteristics and physiological signal characteristics of the target object are acquired. The physical motion characteristics are used to characterize the body motion state of the target object, and the physiological signal characteristics are used to characterize the dynamic trend of the analyte value of the target object changing over time.

[0039] In step S202 above, the analyte refers to a target chemical substance or biomarker that is continuously and in real time collected by a sensor implanted under the skin or on the body surface in a continuous monitoring device (also known as a continuous analyte monitoring device). This includes, but is not limited to, biomolecules such as blood glucose, insulin, lactate, ketone bodies, and cortisol. It should be noted that a continuous monitoring device refers to a medical or health management system that uses implantable or non-implantable sensors to continuously monitor specific biomarkers (i.e., analytes) in the body of a target individual in real time or near real time, and transmits the monitoring data to a processing terminal. This includes, but is not limited to, continuous glucose monitors (CGM), continuous insulin monitors (CIM), and continuous lactate monitors (CLM).

[0040] It should be noted that the non-motion state refers to the physiological state in which the target object's overall limb activity intensity is below a preset vigorous exercise threshold, and no high-energy physical activity that causes significant spatial displacement is being performed. In some embodiments of this application, the non-motion state can be characterized by a low level of macroscopic acceleration energy (such as the overall root mean square value), allowing the system to eliminate high noise interference generated by vigorous movements such as running and jumping, thereby focusing on capturing the "silent" physical movement characteristics caused by local static compression, minute high-frequency tremors, or changes in posture. The non-motion state includes, but is not limited to, the sleep state (such as lying on one's side or prone, which are postures that easily generate local compression), the resting state (such as sitting or lying quietly when awake), and the sedentary state (such as maintaining a sitting posture for a long time without significant limb movement).

[0041] An analyte fluctuation event refers to a significant change in the concentration of an analyte in the interstitial fluid of a target object during continuous monitoring. This can manifest as a sharp decrease in concentration or entry into a preset low-risk range. In some embodiments of this application, an analyte fluctuation event is determined as follows: if the analyte value of the target object is less than a preset threshold, an analyte fluctuation event is determined to exist; or, if the rate of decrease in the analyte concentration value of the target object is greater than a first preset rate, an analyte fluctuation event is determined to exist.

[0042] Specifically, taking blood glucose monitoring as an example, when the CGM value is detected to enter the "hypoglycemic risk zone" (e.g., <3.9mmol / L) or shows a sharp downward trend (e.g., the downward slope is >0.1mmol / L / min), an analyte fluctuation event is identified.

[0043] In some embodiments of this application, when an analyte fluctuation event is detected in the target object while it is asleep, the triggering of an alert event for the analyte fluctuation event is paused. It should be noted that an alert event refers to a notification message sent by the system to the user terminal (such as a smartwatch or mobile app) to indicate an abnormal physiological state, including but not limited to sound alarms, vibration alerts, or pop-up notifications. For example, the system sets an alarm suspension timer (e.g., with a maximum of 60 seconds), during which a hypoglycemia alert event that is about to be triggered is temporarily suspended (i.e., not sent to the user).

[0044] Physical motion characteristics refer to a multidimensional set of data captured by accelerometers that reflects the physical state and posture changes of a target object's body surface. These data are used to characterize the body motion state of the target object during non-motion periods. For example, they may include, but are not limited to, silent background features reflecting the overall amount of activity (such as root mean square value), transient impact features reflecting instantaneous movements (such as acceleration spikes when turning over), morphological features reflecting minute vibrations (such as variance or zero-crossing rate of high-frequency low-amplitude tremors), and angular features reflecting body posture (such as lateral decubitus angle).

[0045] Physiological signal characteristics refer to dynamic data sequences acquired through continuous monitoring sensors that reflect the evolution of analyte concentrations within a target object over time. These sequences are used to characterize the dynamic trends of analyte concentrations and determine the physiological rationality of fluctuations. Examples include, but are not limited to, the slope of concentration changes, the morphology of fluctuations (such as whether a "V-shaped" rapid drop and rebound occurs), and the recovery rate after a fluctuation.

[0046] In some embodiments of this application, the physical motion characteristics and physiological signal characteristics of the target object can be obtained in the following ways: obtaining acceleration data and analyte data corresponding to the fluctuation event of the analyte within a first preset time window, wherein the first preset time window includes historical data and prospective data corresponding to the occurrence time of the fluctuation event of the analyte; determining the physical motion characteristics of the target object based on the acceleration data; and determining the physiological signal characteristics based on the analyte data.

[0047] It should be noted that the first preset time window refers to a specific time period locked by the system for artifact identification after the analyte fluctuation event is detected. This time period is based on the occurrence time (or pre-trigger time) of the fluctuation event, retrospectively looking back for a certain duration (historical data) and extending forward for a certain duration (look-ahead data). It should also be noted that since signal artifacts caused by sleep compression often have specific temporal characteristics (such as the impact of turning over precedes the signal drop, or the drop is accompanied by a rapid recovery), it is impossible to accurately judge them based on instantaneous data alone. By using a window that includes historical and look-ahead data, the system can capture the complete chain of "body movement-compression-signal change-recovery", thus providing sufficient temporal dimension information for calculating the confidence of artifacts.

[0048] Acceleration data refers to the raw triaxial acceleration signal sequence acquired by an accelerometer integrated into a continuous monitoring device or a wearable terminal associated with it, while analyte data refers to the raw signal sequence or a pre-calibrated concentration value sequence reflecting the concentration of the analyte in the target interstitial fluid, acquired by the electrochemical or optical elements of the continuous monitoring sensor.

[0049] Specifically, when the system activates the artifact detection module via slope pre-triggering or numerical threshold determination, it immediately reads the current time (T) from the non-volatile memory. now () as a time segment centered on, for example, Figure 3 As shown, the system is set to a backtrack time of 10 minutes (T). past The look-ahead time is 2 minutes (T). future A fixed 12-minute window is formed, and all sampling point data within the window are synchronously read from the storage buffer, including raw triaxial acceleration values ​​(such as accelerometer sequence A) and raw analyte signal values ​​(such as raw CGM sequence G) or concentration values ​​after preliminary filtering.

[0050] Furthermore, determining the physical motion characteristics of the target object based on acceleration data can be achieved in some embodiments of this application by: determining a first physical motion characteristic based on acceleration data, wherein the first physical motion characteristic is used to characterize the instantaneous intensity of the physical motion behavior that causes local compression of the analyte sensor; determining a second physical motion characteristic based on acceleration data, wherein the second physical motion characteristic is used to characterize the energy intensity or frequency density of the high-frequency oscillation component in the acceleration signal that reflects local compression; determining a third physical motion characteristic based on acceleration data, wherein the third physical motion characteristic is used to characterize the risk of local compression caused by the body motion posture of the target object; and using the first physical motion characteristic, the second physical motion characteristic, and the third physical motion characteristic together as the physical motion characteristics.

[0051] It should be noted that (1) the first physical motion characteristic refers to the index quantified by acceleration data, which reflects the intensity of instantaneous impact or pressure on the sensor caused by violent body movements (such as turning over or hitting the bed) of the target object in a non-motion state (such as during sleep). For example, the system first performs bandpass filtering on the acceleration data to remove noise, and then searches for the local maximum value of the acceleration amplitude within a preset time window (such as within 10 seconds before the analysis object falls); the system records the maximum acceleration amplitude as a direct measure of the transient impact intensity, and at the same time calculates the root mean square value (RMS) of the acceleration signal within a short time window (such as 1 second) before and after the peak value as the impact energy index; if the maximum amplitude exceeds the preset threshold, it is determined that there is a significant transient impact, and the amplitude is normalized and used as the first physical motion characteristic value.

[0052] (2) The second physical motion feature refers to the energy intensity or frequency distribution characteristics of the local high-frequency micro-vibrations or tremors generated by the sensor after being compressed, which are extracted from the acceleration data. For example, the system first performs a fast Fourier transform (FFT) on the acceleration signal to calculate the proportion of its energy in a specific high-frequency band (such as 10Hz to 50Hz, the specific range can be adjusted according to the sensor characteristics) to the total energy, which is used as the energy intensity index of the high-frequency oscillation component; the original acceleration signal is zero-mean processed, and the number of times the signal crosses the zero line (zero crossing rate) is counted. The higher the zero crossing rate, the higher the signal oscillation frequency is usually; the normalized frequency band energy proportion and the zero crossing rate are weighted and fused to obtain the second physical motion feature.

[0053] (3) The third physical motion characteristic refers to the risk index calculated from the triaxial acceleration data, which reflects the current body motion posture of the target object (such as the side-lying angle and pitch angle) and the possibility of pressure on the sensor wearing area. For example, the system uses data from the triaxial accelerometer under static or quasi-static conditions (i.e., ignoring linear acceleration and mainly considering gravitational acceleration) to calculate the pitch angle and roll angle of the device through the arctangent function or quaternion algorithm; based on the user's sensor wearing position (such as the left upper arm), a local coordinate system is established, and the projection components of the gravity vector on each axis of the coordinate system are calculated. Taking the sleeping posture in the sleep state as an example, when the user is lying on their side and the sensor is side-facing, the roll angle is close to 90 degrees or 270 degrees. At this time, the component of gravity in the sensor axis is the largest. Based on this, the system calculates a pressure risk index between 0 and 1, which directly characterizes the possibility that the current body position will cause pressure on the sensor.

[0054] After obtaining the first physical motion feature, the second physical motion feature, and the third physical motion feature, the feature vectors of the above three dimensions are fused to form a complete physical motion feature set.

[0055] In some embodiments of this application, the following steps may also be performed: obtaining the activity level of the target object within a second preset time window, wherein the activity level is used to quantify the overall energy level of the target object's body movement; and triggering the determination of the physical movement characteristics of the target object when the activity level is less than a preset activity level threshold.

[0056] Specifically, the system first performs noise reduction on the triaxial acceleration data within the second preset time window, and then calculates the root mean square (RMS) values ​​of the X, Y, and Z triaxial acceleration data, or calculates the RMS value of the triaxial composite acceleration vector. The obtained RMS value is used as the activity level. This value intuitively reflects the fluctuation amplitude of the acceleration signal. The lower the value, the more still the body is.

[0057] Furthermore, the system pre-sets a fixed activity threshold (e.g., RMS < 0.05g). When the calculated activity is less than this threshold, a "silent mode" flag is immediately set, and the execution permission of the physical motion feature extraction module is unlocked. If the activity is greater than or equal to this threshold, the fine feature (e.g., second physical motion feature) extraction step is skipped, and the time period is directly marked as "high motion interference" or no analysis is performed.

[0058] The above embodiments use activity level as an indicator to screen out silent sleep scenarios suitable for detailed physical feature analysis, thereby avoiding invalid calculations and misjudgments of vigorous exercise data.

[0059] To facilitate understanding of the process of determining the above physical motion characteristics, the following explanation is provided in conjunction with some specific embodiments. Taking blood glucose monitoring as an example, for a sleep scenario, this application defines a set of features specifically used to identify compressive behaviors, which are different from the energy characteristics of conventional motion, including but not limited to:

[0060] (1) Silent background characteristics: Calculate the total activity level (such as the root mean square value of RMS) within the second preset time window. If the activity level is extremely low (close to stillness, i.e. less than the preset activity level threshold), it meets the background conditions for the occurrence of pressure.

[0061] (2) Transient impact characteristics (i.e., first physical motion characteristics): Within a preset time interval (e.g., 5 to 10 seconds) before the CGM signal drops, detect instantaneous spikes (rollover shock waves) or rapid oscillation components (micro high-frequency vibrations) in the acceleration data. Among them, the rollover shock wave is manifested as an instantaneous spike in the acceleration amplitude, and the high-frequency vibration is manifested as a rapid oscillation of the acceleration signal.

[0062] (3) Morphological characteristics (i.e., second physical motion characteristics): Calculate the variance and zero-crossing rate of the acceleration signal. It should be noted that the signal under pressure usually manifests as high-frequency, low-amplitude tremors (such as sensor vibration under pressure), which is significantly different from the steady breathing waveform during sleep.

[0063] (4) Postural change characteristics (i.e., third physical motion characteristics): The user's posture angle (such as lateral angle and pitch angle) is estimated by the three-axis data of the accelerometer. Among them, the lateral position (especially the one directly pressing on the side of the sensor) is a high-risk factor for compression artifacts.

[0064] In some embodiments of this application, physiological signal characteristics can be determined in the following ways: determining a first physiological signal characteristic based on analyte data, wherein the first physiological signal characteristic is used to characterize the degree of abnormality of analyte fluctuations; determining a second physiological signal characteristic based on analyte data, wherein the second physiological signal characteristic is used to characterize the intensity of physiological signal rebound after the local pressure on the analyte sensor is relieved; and using the first physiological signal characteristic and the second physiological signal characteristic together as physiological signal characteristics.

[0065] It should be noted that the first physiological signal characteristic refers to an indicator quantified from analyte data, reflecting the degree of abnormality in the geometric shape (such as waveform, slope, curvature, etc.) of the analyte concentration over time. True physiological fluctuations usually follow certain biodynamic laws (such as slow decrease or increase), while pressure-induced artifacts often manifest as a sharp drop or "V-shaped" reversal that violates physiological laws.

[0066] The second physiological signal characteristic refers to the intensity index quantified by analyte data, reflecting the false rebound (i.e., rapid increase in value) of the analyte concentration signal after the local compression is relieved. When the sensor is relieved of pressure, the restoration of local blood flow may cause a temporary increase in interstitial fluid glucose concentration, forming a "false rebound." This characteristic can be used to capture this specific phenomenon. If a significant rebound intensity is detected, and this rebound immediately follows the physical compression signal, the confidence that the current fluctuation is an artifact is greatly increased.

[0067] Specifically, the system first performs noise reduction and smoothing on the analyte data (optionally), then calculates the local slope and curvature of the signal curve, and detects the presence of a "V-shaped" or "U-shaped" inversion structure, i.e., a rapid signal decline followed by a rapid rebound. For example, the system calculates the average absolute value of the slope during the drop phase and the average absolute value of the slope during the rebound phase. If both exceed a preset threshold (e.g., 0.5 mmol / L / min), it is determined to be morphologically abnormal. At the same time, the system calculates the total time span of the signal during the drop and rebound process; the shorter the time, the higher the degree of morphological abnormality. The normalized slope product and the reciprocal of the time are used as the first physiological signal feature value.

[0068] Furthermore, after detecting that the analyte signal has fallen to its lowest point, the system searches backward for the maximum value within the next preset time window (e.g., 1-3 minutes); calculates the difference between the maximum value and the lowest point as the rebound intensity; at the same time, it calculates the time required to rise from the lowest point back to the maximum value as the rebound speed; and weights and fuses the normalized rebound amplitude and rebound speed to obtain the second physiological signal characteristic.

[0069] Step S204: Determine the artifact confidence level based on physical motion characteristics and physiological signal characteristics. The artifact confidence level is used to quantify the likelihood that the analyte fluctuation event is a false analyte value reduction event caused by the body movement behavior of the target object.

[0070] In step S204 above, the artifact confidence can be determined in the following way: obtain the body motion impact characteristics (i.e., the first physical motion characteristics), high-frequency low-amplitude tremor characteristics (i.e., the second physical motion characteristics), body position risk characteristics (i.e., the third physical motion characteristics) from the physical motion characteristics of the target object, and the morphological characteristics (i.e., the first physiological signal characteristics) and recovery characteristics (i.e., the second physiological signal characteristics) from the physiological signal characteristics of the target object; determine the weighting coefficients of each feature; multiply each normalized feature value obtained by its corresponding weighting coefficient, and add all the product results to obtain the probability score of the current analyte fluctuation event being an artifact as the artifact confidence.

[0071] To facilitate understanding of the process of determining the confidence level of artifacts described above, the following explanation is provided in conjunction with some specific embodiments.

[0072] For example, a scoring function based on multi-feature fusion can be constructed to calculate the probability that the current event is an artifact. (i.e., artifact confidence), the formula for which it is calculated is as follows:

[0073]

[0074] in, To assess the impact characteristics, detect whether there is a rollover or impact acceleration peak before the CGM falls (normalized to 0~1). The high-frequency, low-amplitude jitter characteristics are calculated using the variance or zero-crossing rate of the acceleration signal. The morphological characteristics of CGM signals, such as whether they exhibit a "V-shaped" rapid drop and rebound; For positional risk characteristics, the first value is used when the patient is in a lateral decubitus position with compression, and the second value is used for other positions. The first value is greater than the second value. The recovery characteristic is whether the body recovers quickly after a fall (a false recovery caused by the restoration of blood flow after the pressure is relieved).

[0075] As an example, its weights can be set as follows:

[0076] , , , , .

[0077] It should be noted that (1) if the CGM drops sharply and the acceleration signal exhibits high-frequency, low-amplitude characteristics, then Significant increase. That is, when the system detects a sharp drop in analyte concentration (CGM) (indicating possible hypoglycemia), if it also detects "high-frequency, low-amplitude" oscillation characteristics in the acceleration signal, the system will determine that the tremor component in the current physical motion characteristic (i.e., the second physical motion characteristic) is significantly enhanced. This corresponds to the fine vibration generated when the sensor is subjected to local physical pressure (such as limb pressure sensor). This vibration is superimposed on the blood glucose signal to form an artifact. The increase in high-frequency, low-amplitude tremor characteristics directly increases the confidence that the event is a pressure artifact.

[0078] (2) If the CGM shows a "V-shaped" reversal (falling very quickly and rising very quickly), and it follows closely after the body movement, then Significantly increased. That is, when the analyte data exhibits a typical "V-shaped" geometric shape (i.e., the first physiological signal characteristic), the system will check whether the time point of this shape occurs closely after a physical movement event (such as turning over or the pressure being relieved). If such a close temporal causal relationship exists, the system determines that the recovery component in the current physiological signal characteristic is significantly enhanced. This is because after the physical pressure is relieved, local blood flow recovers rapidly, leading to a false rapid increase in interstitial fluid glucose concentration, forming a "false rebound".

[0079] (3) If CGM decreases but acceleration indicates the user is in a state of deep stillness and without pressure, then The system will reduce the risk of hypoglycemia (determined to be a true hypoglycemia). That is, when the analyte data shows a downward trend, if the system detects an acceleration signal indicating that the user is in a highly static state (low silent background characteristics), and the body position characteristics (i.e., the third physical motion characteristics) indicate that the user is not in a high-risk position that may cause compression (such as not being on the side of compression), then the system determines that the current physical motion characteristics cannot support the situation of physical compression. This means that the system tends to interpret the signal fluctuation as a true physiological hypoglycemic event, and thus retains the alarm or prompt.

[0080] In some embodiments of this application, the artifact confidence level can also be determined in the following ways: when it is determined that there is an analyte fluctuation event, a warning event corresponding to the analyte fluctuation event is triggered with a preset delay; within the preset delay, the target physical motion characteristics and target physiological signal characteristics of the target object are acquired, wherein the target physical motion characteristics include features used to characterize the instantaneous intensity of the target object's physical motion behavior, and the target physiological signal characteristics include features used to characterize the abnormality of the analyte fluctuation; when the target physical motion characteristics and target physiological signal characteristics meet preset conditions, a preset first artifact confidence level is used as the artifact confidence level, wherein the first artifact confidence level is used to indicate the suppression of the warning event.

[0081] Specifically, to avoid delays or false triggers of hypoglycemia alarms due to feature extraction and scoring calculations (for example, if V-shaped inversion compression artifacts are only identified after the alarm is issued, they lose their inhibitory significance), in some embodiments of this application, a mechanism combining slope pre-triggering and alarm suspension can be introduced, taking blood glucose monitoring as an example:

[0082] (1) Slope pre-trigger: The system continuously monitors the decline slope of CGM. Once the slope exceeds the threshold (e.g., 0.1 mmol / L / min), the artifact identification module is activated before the CGM value reaches the hypoglycemic alarm threshold, thus gaining a time window for feature extraction.

[0083] (2) Alarm Suspension: When the slope is pre-triggered or the CGM value enters the risk range, the system temporarily suspends (delays the issuance) the hypoglycemia alarm that was about to be triggered. The upper limit of the suspension time can be, for example, Second.

[0084] (3) Fast feature priority: During the suspension period, the most easily obtained transient impact features are extracted first. (i.e., target physical motion characteristics) and early CGM morphological characteristics (i.e., target physiological signal characteristics, such as whether rapid turning occurs), without waiting for a complete prospective window (i.e., the sub-window corresponding to prospective data), if a typical compression pattern is identified (e.g., rapid CGM recovery after a rolling shock wave, i.e., a preset condition), then a high-confidence artifact (i.e., the first artifact confidence level, such as...) is directly output during the suspension period. This will permanently suppress the alarm.

[0085] (4) Release after timeout: If the judgment is not completed by the end of the suspension period, a temporary confidence level will be output based on the existing partial characteristics. If necessary, an alarm will be released to ensure that the real hypoglycemia is not suppressed for a long time.

[0086] The above mechanism ensures a balance between real-time alarm response (hang-up time ≤ 60 seconds) and accurate artifact recognition. Especially for V-shaped inversion compression scenarios, it can complete the inhibition decision before blood glucose reaches the alarm threshold.

[0087] Step S206: Determine whether the analyte fluctuation event is an artifact based on the artifact confidence level, wherein, if the analyte fluctuation event is an artifact, suppress the warning event corresponding to the analyte fluctuation event.

[0088] In step S206 above, the warning event refers to the notification signal sent by the system to the user terminal to indicate abnormal analyte concentration (such as hypoglycemia) and to suggest that the user take intervention measures.

[0089] In some embodiments of this application, whether an analyte fluctuation event is an artifact can be determined in the following ways: the artifact confidence level is compared with a first threshold and a second threshold respectively to obtain a comparison result, wherein the first threshold is less than the second threshold; if the comparison result indicates that the artifact confidence level is less than the first threshold, the analyte fluctuation event is determined to be a real event; if the comparison result indicates that the artifact confidence level is greater than or equal to the second threshold, the analyte fluctuation event is determined to be an artifact event; if the comparison result indicates that the artifact confidence level is not less than the first threshold and less than the second threshold, whether the analyte fluctuation event is an artifact is determined based on the value range of the artifact confidence level.

[0090] Specifically, taking blood glucose monitoring as an example, (1) when the calculated artifact confidence is less than the first threshold (e.g., 0.2), the system determines that the current analyte fluctuation event is a real hypoglycemic event. In this case, the system immediately performs an emergency alarm operation and sends a high-intensity warning signal to the user terminal, including sound alarm, vibration feedback or prominent interface pop-up, to prompt the user to take immediate intervention measures (e.g., sugar intake) to ensure user safety.

[0091] (2) When the confidence level of the artifact is in the range of 0.2 to 0.4 (inclusive of 0.2, exclusive of 0.4), the system determines that the current event tends to be a real hypoglycemia, but there is still some uncertainty. In this case, the system issues a "caution" level prompt, which is a mild reminder of non-emergency nature, such as displaying a yellow warning icon on the user interface or sending a low-intensity vibration, suggesting that the user pay attention to the current blood glucose trend and recheck in time, without triggering a mandatory emergency intervention alarm.

[0092] (3) When the confidence level of the artifact is in the range of 0.4 to 0.6 (inclusive of 0.4, exclusive of 0.6), the system determines that the nature of the current event is uncertain and cannot clearly distinguish whether it is a real hypoglycemia or an artifact caused by physical compression. In this case, the system triggers a "recommend fingertip blood calibration" notification, prompting the user to use a traditional fingertip blood glucose meter to measure and confirm the current blood glucose level, and wait for further confirmation of subsequent data points, thereby improving the accuracy of the judgment through cross-validation of multi-source data.

[0093] (4) When the confidence level of the artifact is in the range of 0.6 to 0.9 (inclusive of 0.6, exclusive of 0.9), the system determines that the current event is likely an artifact. In this case, the system suppresses the hypoglycemia alarm that should have been triggered to avoid unnecessary interference to the user. At the same time, the system marks the data segment of the analyte as "suspicious compression interference" on the application side. Although the data is marked, the system still retains the display function of the data for the user to view, but the data segment is not included in the core clinical report (such as the daily average blood glucose statistics) to balance data integrity and alarm accuracy.

[0094] (5) When the confidence level of the artifact is greater than or equal to the second threshold (e.g., 0.9), the system determines that the current event is a high-confidence artifact and confirms that it is a false signal caused by physical compression. In this case, the system completely blocks any form of alarm for the event. At the data level, the data segment of the analyte is marked as "invalid (compression artifact)" and removed in subsequent data processing. It is not included in daily blood glucose statistics, average glucose level profiles and any clinical decision support reports, thereby ensuring the purity and reliability of long-term blood glucose monitoring data.

[0095] In some embodiments of this application, real analyte fluctuation events (i.e., real events) are determined, including but not limited to, through the following methods:

[0096] (1) When the calculated confidence level of the artifact is lower than the set first threshold, the event is determined to be a real analyte fluctuation event. It should be noted that a low confidence level means that the current signal lacks characteristic evidence of physical compression (such as lack of rolling impact, high-frequency tremors, V-shaped rebound, etc.), so it tends to believe that the signal change is caused by physiological reasons, such as excessive insulin or a real decrease in blood glucose caused by not eating.

[0097] (2) The accelerometer data shows that the target object is in a silent background state (extremely low activity, close to stillness), and no typical compression posture (such as lying on one side of the sensor) or transient impact (such as rolling over and hitting the sensor) is detected, indicating that the signal drop is not caused by physical compression; moreover, the CGM signal shows a continuous decline or a continuous low level after the drop, without the "V-shaped rapid rebound" phenomenon commonly seen after the compression is relieved. In this situation, an emergency alarm can be triggered to prompt the user to intervene (such as ingesting sugar), rather than suppressing the alarm.

[0098] (3) If the initial artifact confidence level is in the uncertain range (such as the range of 0.2 to 0.6 mentioned above), the observation window can be extended or the user can be prompted to perform finger-prick blood calibration: if subsequent data confirm that the signal remains low and there is no characteristic rebound of compression; or if the finger-prick blood measurement confirms hypoglycemia, then it is finally confirmed as a real analyte fluctuation event.

[0099] In order to intelligently select the optimal artifact identification strategy based on real-time conditions, in some embodiments of this application, the following steps may also be performed: when the physiological signal characteristics indicate that the rate of decrease of the analyte value of the target object is greater than the second preset rate and there is an instantaneous spike in the acceleration signal in the physical motion characteristics, the second artifact confidence is used as the artifact confidence, wherein the second artifact confidence is used to indicate that the analyte fluctuation event is an artifact; or, when the artifact confidence is within the target value range, the observation window used to determine the physical motion characteristics and physiological signal characteristics is extended, and the third artifact confidence is re-determined based on the data within the observation window.

[0100] Specifically, for example, when the CGM decrease rate is >0.2 mmol / L / min (i.e., the second preset rate) and a significant shock wave occurs in the acceleration (rolling or impact, i.e., the acceleration signal has a momentary spike), the system directly increases the acceleration. Output quickly when the value reaches saturation. (i.e., the second artifact confidence level) to achieve second-level response and avoid unnecessary delays.

[0101] when When the calculation result falls within the uncertainty range of 0.3 to 0.7 (i.e., the target value range), the system automatically extends the observation window (e.g., waits for another 3 CGM sampling points, approximately 3 minutes) and recalculates the score. Furthermore, if the two scores are consistent, corresponding measures are taken; if they are inconsistent, the "uncertain" state is maintained and a calibration prompt is displayed.

[0102] In addition, when the device battery level is lower than the preset level (e.g., 15%) or the user is in a known period of no stress (e.g., sitting for long periods during the day), the system can temporarily disable some high-overhead features of the artifact detection module (e.g., high-frequency jitter analysis) and rely solely on the original CGM alarm to extend the device's battery life.

[0103] To achieve dynamic adjustment based on power consumption and computing resources, the following steps can also be performed: obtain the power consumption value of the continuous analyte monitoring device; if the power consumption value is greater than a first power consumption value, determine the fourth artifact confidence level using a first number of physical motion features and physiological signal features; if the power consumption value is less than a second power consumption value, determine the fifth artifact confidence level using a second number of physical motion features and physiological signal features, wherein the second power consumption value is less than the first power consumption value, the second number is less than the first number, and the threshold of the fifth artifact confidence level is greater than the threshold of the fourth artifact confidence level.

[0104] Specifically, when the power is sufficient, full feature calculation and low threshold (e.g., P≥0.6) are used to pursue high accuracy; when the power is insufficient, high-overhead features (e.g., high-frequency jitter analysis) are automatically turned off. Since the reduction of feature dimensions will reduce the reliability of confidence scoring, the judgment threshold (e.g., P≥0.8) needs to be increased accordingly to control the risk of misjudgment and achieve a balance between power consumption and performance.

[0105] In some embodiments of this application, the following steps may also be performed: acquiring historical artifact events corresponding to the target object and determining the artifact feature vector of the historical artifact events; determining the compression mode of the target object based on the artifact feature vector, wherein the compression mode is used to reflect the compression habit of the target object on the analyte sensor in a non-motion state; and reducing the detection trigger threshold corresponding to the physical motion feature during the compression time period indicated by the compression mode, wherein the detection trigger threshold is used to reflect the sensitivity to compression detection.

[0106] It should be noted that the artifact feature vector refers to the combination of key features extracted from historical artifact events. It typically includes a multi-dimensional data array comprising dimensions such as timestamp, sleep position angle at the time, body motion impact intensity, background activity level, and the descent slope of the analyte signal. This data is used to quantitatively describe the environmental state and physiological signal manifestations during the compression event. For example, the feature vector (timestamp, position angle, body motion impact intensity, CGM descent slope, etc.) of each artifact event can be stored in non-volatile memory to form a user-specific "compression artifact profile."

[0107] Compression patterns refer to the regular habits of physical compression of a target object during sleep, identified by analyzing the feature vectors of historical artifacts. These patterns may include specific time periods (such as 2:00-3:00 AM), specific body positions (such as lying on the left side), or specific sequences of movements.

[0108] The detection trigger threshold is a critical value used to determine whether physical motion features are sufficient to trigger the artifact identification module or increase the artifact confidence score. It is used to adjust the system’s sensitivity to physical compression signals. Lowering this threshold means that the system is more sensitive to weak body movements or small signal changes, thus enabling it to capture artifact features earlier in the early stages of compression or when the compression signal is weak.

[0109] Specifically, the system first retrieves all historical events marked as high-confidence artifacts from non-volatile memory within a past period (e.g., the past month). For each historical event, the system extracts its specific time of occurrence (e.g., 2:30 AM), the body position on which the sensor was worn at the time (e.g., lying on the left side, determined by the attitude angle estimated by the accelerometer), and the total activity level (RMS value) at that time. Lightweight clustering algorithms (e.g., density-based spatial clustering) are then used to analyze these feature vectors to identify frequently occurring "time-position" combinations. For example, if a user habitually lies on their left side between 2:00 AM and 3:00 AM to compress the sensor, the system automatically lowers the sensor during similar time periods. and Lowering the trigger threshold (e.g., by 20%) increases sensitivity to pressure.

[0110] In addition, if the original CGM signal noise is high (e.g., the sensor is nearing its failure period), it can be appropriately increased. The threshold for judging artifacts (e.g., increasing the artifact judgment threshold from 0.6 to 0.8) is set to avoid misjudging the noise of the sensor itself as compression artifacts.

[0111] Before detecting analyte fluctuation events in a non-motion state of the target object, the following steps can be performed: obtain the target state of the target object; if the target state matches the compression mode, obtain candidate physical motion features and candidate physiological signal features of the target object within the prediction time window, and determine the confidence of candidate artifacts based on the candidate physical motion features and candidate physiological signal features, wherein the length of the prediction time window is less than the time window corresponding to the analyte fluctuation event.

[0112] It should be noted that the target state refers to the current spatiotemporal and physiological environmental state of the target object, including but not limited to the current timestamp, current sleeping position (such as side-lying, supine, etc., posture angle estimated by accelerometer), current background activity level (such as resting or slight movement), and current device battery status. The prediction time window refers to the time interval during which the system starts artifact detection in advance after the target state successfully matches the historical "compression pattern". The start time of this window is earlier than the detection time based on slope pre-triggering or numerical limit exceeding, and its total length is less than the complete time window corresponding to the standard analyte fluctuation event.

[0113] Candidate physical motion features refer to preliminary physical indicators extracted within the prediction time window to characterize the body motion state of the target object. It should be noted that candidate physical motion features may include one or more of the first, second, and third physical motion features mentioned above, without limitation. Candidate physiological signal features refer to preliminary physiological indicators extracted within the prediction time window to characterize the dynamic trend of analyte concentration. It should be noted that candidate physiological signal features may include one or more of the first and second physiological signal features mentioned above, without limitation.

[0114] Specifically, the system monitors the current user status in real time. If a high match with historical artifact patterns is detected (e.g., same time period, same body position, similar background activity), it enters a high-sensitivity artifact detection state in advance, and begins calculation before the CGM value drops significantly. And appropriately reduce the time window for triggering the judgment.

[0115] Furthermore, each actual artifact event (confirmed by the user or verified by subsequent data) can be fed back to the model to dynamically adjust the matching threshold, thereby achieving self-learning and adaptive prediction capabilities.

[0116] To identify sensor detachment events, the following steps can also be performed: In the case of abnormal physiological signal characteristics, determine a fourth physical motion characteristic within a first abnormal time window, wherein the first abnormal time window is located before the occurrence of the abnormality, and the fourth physical motion characteristic is used to characterize the physical movement behavior that causes the analyte sensor to detach from the body surface; determine a third physiological signal characteristic within a second abnormal time window, wherein the second abnormal time window is located after the occurrence of the abnormality, and the third physiological signal characteristic is used to characterize the recovery state of the target object's physiological signals after the occurrence of the abnormality; and determine whether an analyte sensor detachment event exists based on the fourth physical motion characteristic and the third physiological signal characteristic.

[0117] It should be noted that anomaly refers to the signal state in which the analyte value data output by the analyte sensor deviates from the normal physiological monitoring range or the expected working state of the device. In some embodiments of this application, anomaly may be manifested as an abnormal drop in the analyte value (e.g., a sharp drop below the preset hypoglycemia risk threshold or the lower limit of the normal fluctuation range in a short period of time) or a continuous drop below the invalid detection threshold (e.g., the signal value is continuously below the effective detection lower limit such as 0.1 mmol / L, or presents an invalid identifier defined by the system).

[0118] The first abnormal time window refers to a preset time period before the moment when the analytical value drops abnormally or remains invalid. It is used to capture the physical causes that cause the sensor to detach from the body surface and detect violent physical actions such as "precursor to detachment" or "moment of detachment", such as large fluctuations in acceleration signals caused by turning over or rubbing.

[0119] The fourth physical motion feature refers to the feature vector extracted based on the acceleration data within the first abnormal time window. It is used to characterize the intensity and form of the physical motion behavior that causes the analyte sensor to detach from the body surface. Physical compression usually manifests as high-frequency low-amplitude tremors or static pressure changes, while detachment corresponds to large displacement or impact.

[0120] The second abnormal time window refers to a preset time period after the occurrence of the abnormality. It is used to observe the signal evolution trend after the abnormality occurs, assess the signal persistence after the sensor loses contact, and verify whether the signal abnormality is caused by the interruption of physical connection. For example, if the sensor falls off, the signal will no longer reflect physiological changes, but will remain in the invalid range or the extremely low value range.

[0121] The third physiological signal feature refers to the feature extracted based on the analyte values ​​within the second abnormal time window. It is used to characterize the recovery status or maintenance status of the target object's physiological signal after the occurrence of the abnormality. For example, if the signal is continuously below the effective detection threshold (such as continuously invalid or extremely low value) and there is no sign of recovery, this feature indicates that the signal has not recovered from the abnormality.

[0122] Specifically, when the system detects an abnormal drop or persistent invalid value in the CGM signal (i.e., abnormal physiological signal characteristics), it triggers a backtracking mechanism to obtain acceleration data within the first abnormal time window before the anomaly occurs. By analyzing the acceleration waveform during this period, it extracts the fourth physical motion characteristic: whether there is a "significant instantaneous impact with a large amplitude" (or a significant change in the target object's body posture (such as rolling from supine to side-lying, or violent limb swinging) within a short period of time). For example, if the acceleration signal shows a large spike a few seconds before the anomaly occurs, it indicates that the user has experienced violent body movement (such as rolling over and rubbing against the sensor), which is consistent with the physical cause characteristics of sensor detachment.

[0123] Furthermore, after confirming the occurrence of an anomaly, the system enters a second anomaly time window to continuously monitor and analyze the physical data; observes the evolution of the signal after the anomaly occurs, and extracts the third physiological signal characteristic, namely whether the signal is "persistently below the effective detection threshold" and "has no upward trend". For example, if the sensor detaches, the signal will no longer change with blood glucose, but will continue to show an invalid value or an extremely low fixed value (such as <2.0 mmol / L) with no signs of recovery.

[0124] Furthermore, if the third physiological signal characteristic indicates that no recovery is observed within a preset time period (e.g., recovery slope ≤ 0), and the fourth physical motion characteristic indicates that abnormal displacement / impact is detected, it is determined that the sensor has fallen off, triggering the "Please check if the sensor is attached" reminder, and the artifact-related alarm suppression is not executed.

[0125] Alternatively, if the third physiological signal characteristic indicates that the physiological signal shows a V-shaped recovery (e.g., recovery slope > 0.1 mmol / L / min), then continue to enter the artifact confidence scoring model for normal discrimination; if the physiological signal does not recover and the acceleration is not abnormal, then further processing is carried out as true hypoglycemia or sensor failure.

[0126] Through steps S202 to S206, an artifact confidence assessment method based on the fusion of physical motion characteristics and physiological signal characteristics of analyte fluctuations is adopted. By quantifying the likelihood of false reductions in analyte fluctuation events caused by body movement, and suppressing corresponding warning events based on the confidence level, the method achieves the goal of accurately distinguishing between physical pressure interference and real physiological abnormalities in non-motor states such as sleep. This reduces the false alarm rate at night and ensures the accuracy of monitoring data, thereby solving the technical problem that related technologies are unable to identify false abnormal analyte values ​​caused by physical pressure sensors, which easily triggers false alarms.

[0127] Figure 4 This is an artifact detection system according to an embodiment of this application, such as... Figure 4 As shown, the system is integrated into a continuous analyte monitoring device, including a sensor 402, an accelerometer 404, an artifact identification module 406, and a memory 408, wherein:

[0128] Sensor 402 is used to continuously collect analyte values ​​(such as glucose) in the subcutaneous interstitial fluid of a target object. When the target object is detected to be in a sleep state and the analyte value shows fluctuations that meet preset conditions (such as being below a specific threshold or having a rate of decrease exceeding a specific threshold), sensor 402 outputs an analyte data sequence containing the original analyte value and a timestamp.

[0129] Accelerometer 404 is used to synchronously acquire three-dimensional acceleration data of the target object to characterize the body motion state of the target object. When the artifact detection module 406 is triggered, accelerometer 404 provides the original acceleration signal within the corresponding time window to memory 408 or artifact detection module 406 for extraction of physical motion features reflecting the body motion state (such as rolling impact, high-frequency tremor, body position angle);

[0130] The memory 408 is used to store raw analyte data, raw accelerometer data, historical artifact event archives, and instructions and parameters required for executing the artifact identification module 406. Specifically, when the artifact identification module 406 determines that an analyte fluctuation event exists, the memory 408 reads the analyte data sequence and the corresponding acceleration data sequence within the historical and look-ahead time windows centered on the current moment from the cache and transmits them to the artifact identification module 406. In addition, the memory 408 is also used to store the artifact confidence score results after judgment, the final judgment conclusion (such as true hypoglycemia, tendency artifacts, etc.), and processing action instructions (such as whether to suppress alarms, data marking status, etc.).

[0131] The artifact detection module 406, as the core processing unit of the system, performs, but is not limited to, the following specific actions:

[0132] (1) Feature extraction: Based on the analysis data sequence and acceleration data sequence obtained from the memory 408, extract the physical motion features (including silent background features, transient impact features, high frequency low amplitude tremor features and body position risk features) and physiological signal features (including morphological abnormality features and rebound intensity features) of the target object.

[0133] (2) Confidence calculation: Based on the extracted physical motion features and physiological signal features, the artifact confidence is calculated through a multi-feature fusion scoring model (such as linear weighting or classification probability mapping). This numerical value quantifies the probability that the current analyte fluctuation event is a false analyte reduction event caused by the body movement behavior of the target object.

[0134] (3) Artifact detection and handling: Based on the calculated artifact confidence level, determine whether the analyte fluctuation event is an artifact by comparing it with the preset confidence level threshold. If it is determined to be an artifact (e.g., artifact confidence level is greater than 0.6), a suppression command is generated to notify the system to suppress the warning event (such as sound alarm or vibration reminder) corresponding to the analyte fluctuation event, and the data segment is marked as "suspicious pressure interference" or "invalid (pressure artifact)" in the memory 408; if it is determined to be a real hypoglycemia (e.g., artifact confidence level is less than 0.2), an emergency alarm is triggered or the user is prompted to intervene.

[0135] (4) Adaptive optimization: Furthermore, the artifact identification module 406 also obtains historical artifact events and their corresponding artifact feature vectors from the memory 408, and determines the compression pattern of the target object (such as compression habits in a specific time period or a specific body position) by analyzing the historical artifact feature vectors; within the compression time period indicated by the identified compression pattern, the detection trigger threshold corresponding to the physical motion feature is automatically reduced to improve the sensitivity to compression signals.

[0136] It should be noted that, Figure 4 The system shown is used to execute Figure 2 The artifact detection method shown is therefore Figure 2 The explanations and descriptions in the artifact detection methods also apply to Figure 4 The system shown will not be described in detail here.

[0137] This application's embodiments effectively identify false hypoglycemia caused by sleep compression through multi-dimensional feature fusion and confidence scoring, reducing the number of false alarms at night and significantly improving user experience. It not only eliminates artifacts but also retains genuine hypoglycemic events, ensuring the accuracy of medical data and the reliability of clinical decisions. Furthermore, through multi-mode strategies, dynamic threshold management, and historical learning mechanisms, the system can automatically adjust according to user habits, device status, and environmental changes, achieving an optimal balance between accuracy and resource consumption. The introduction of an artifact occurrence prediction model enables the system to react before or at the onset of compression, further improving timing response capabilities and user experience. In addition, through a "slope pre-trigger + alarm suspension" mechanism, it ensures that the identification of compression artifacts such as V-shaped inversions is completed before the hypoglycemic alarm is issued or during the alarm suspension period, avoiding the defects of post-event identification and truly achieving effective alarm suppression.

[0138] Figure 5 This is a structural diagram of an artifact detection apparatus according to an embodiment of this application, as shown below. Figure 5 As shown, the device includes:

[0139] The detection module 502 is used to acquire the physical motion characteristics and physiological signal characteristics of the target object when the target object is detected to have analyte fluctuation events in a non-motion state. The physical motion characteristics are used to characterize the body motion state of the target object, and the physiological signal characteristics are used to characterize the dynamic trend of the analyte value of the target object changing over time.

[0140] The determination module 504 is used to determine the artifact confidence level based on physical motion characteristics and physiological signal characteristics. The artifact confidence level is used to quantify the degree to which the analyte fluctuation event is a false analyte value reduction event caused by the body motion behavior of the target object.

[0141] The execution module 506 is used to determine whether an analyte fluctuation event is an artifact based on the artifact confidence level, wherein, if the analyte fluctuation event is an artifact, the warning event corresponding to the analyte fluctuation event is suppressed.

[0142] It should be noted that, Figure 5 The device shown for artifact detection is used to perform Figure 2 The artifact detection method shown is therefore Figure 2 The explanations and descriptions in the artifact detection methods also apply to Figure 5 The artifact detection device shown will not be described in detail here.

[0143] This application also provides an electronic device, which includes a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the steps of implementing the artifact detection method in various embodiments of this application.

[0144] This application also provides a non-volatile storage medium including a stored computer program, wherein the device containing the non-volatile storage medium executes the steps of the artifact detection method in various embodiments of this application by running the computer program.

[0145] This application also provides a computer program product, including computer instructions that, when executed by a processor, implement the steps of the artifact detection method in various embodiments of this application.

[0146] This application also provides a computer program that, when executed by a processor, implements the steps of the artifact detection method in various embodiments of this application.

[0147] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0148] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0149] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0150] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0151] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0152] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0153] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for artifact detection, characterized in that, include: When an analyte fluctuation event is detected in a non-moving state of the target object, the physical motion characteristics and physiological signal characteristics of the target object are acquired. The physical motion characteristics are used to characterize the body motion state of the target object, and the physiological signal characteristics are used to characterize the dynamic trend of the analyte value of the target object changing over time. The artifact confidence level is determined based on the physical motion characteristics and the physiological signal characteristics, wherein the artifact confidence level is used to quantify the probability that the analyte fluctuation event is a false analyte value reduction event caused by the body movement behavior of the target object; Based on the artifact confidence level, it is determined whether the analyte fluctuation event is an artifact, wherein, if the analyte fluctuation event is an artifact, the warning event corresponding to the analyte fluctuation event is suppressed.

2. The method according to claim 1, characterized in that, Acquiring the physical motion characteristics and physiological signal characteristics of the target object includes: Acquire acceleration data and analyte data corresponding to the fluctuation event of the analyte within a preset time window, wherein the preset time window includes historical data and prospective data corresponding to the occurrence time of the fluctuation event of the analyte; The physical motion characteristics of the target object are determined based on the acceleration data; The physiological signal characteristics are determined based on the analyte data.

3. The method according to claim 2, characterized in that, Determining the physical motion characteristics of the target object based on the acceleration data includes: A first physical motion characteristic is determined based on the acceleration data, wherein the first physical motion characteristic is used to characterize the instantaneous intensity of the physical motion behavior that causes the analyte sensor to be subjected to local pressure; A second physical motion characteristic is determined based on the acceleration data, wherein the second physical motion characteristic is used to characterize the energy intensity or frequency density of the high-frequency oscillation component in the acceleration signal that reflects the local compression. A third physical motion feature is determined based on the acceleration data, wherein the third physical motion feature is used to characterize the risk of the target object's body motion posture causing the local compression; The first physical motion feature, the second physical motion feature, and the third physical motion feature are collectively referred to as the physical motion feature.

4. The method according to claim 2, characterized in that, Determining the physiological signal characteristics based on the analyte data includes: A first physiological signal feature is determined based on the analyte data, wherein the first physiological signal feature is used to characterize the degree of abnormality in analyte fluctuations; A second physiological signal feature is determined based on the analyte data, wherein the second physiological signal feature is used to characterize the physiological signal rebound intensity after the local pressure on the analyte sensor is relieved; The first physiological signal feature and the second physiological signal feature are used together as the physiological signal feature.

5. The method according to claim 1, characterized in that, Determining artifact confidence based on the physical motion characteristics and the physiological signal characteristics includes: When the existence of the analyte fluctuation event is determined, an early warning event corresponding to the analyte fluctuation event is triggered after a preset delay. Within the preset time period, the target physical motion characteristics and target physiological signal characteristics of the target object are acquired, wherein the target physical motion characteristics include features used to characterize the instantaneous intensity of the physical motion behavior of the target object, and the target physiological signal characteristics include features used to characterize the abnormality of the fluctuation of the analyte. When the target physical motion characteristics and the target physiological signal characteristics meet preset conditions, a pre-set first artifact confidence level is used as the artifact confidence level, wherein the first artifact confidence level is used to indicate the suppression of the warning event.

6. The method according to claim 1, characterized in that, Determining whether the fluctuation event of the analyte is an artifact based on the artifact confidence level includes: The artifact confidence level is compared with a first threshold and a second threshold respectively to obtain a comparison result, wherein the first threshold is less than the second threshold; If the comparison result indicates that the confidence level of the artifact is less than the first threshold, the analyte fluctuation event is determined to be a real event; If the comparison result indicates that the confidence level of the artifact is greater than or equal to the second threshold, the analyte fluctuation event is determined to be an artifact event; If the comparison result indicates that the confidence level of the artifact is not less than the first threshold and less than the second threshold, the analyte fluctuation event is determined to be an artifact based on the value range of the artifact confidence level.

7. The method according to claim 1, characterized in that, The method further includes: In the event of an abnormality in the physiological signal characteristics, a fourth physical motion characteristic is determined within a first abnormal time window, wherein the first abnormal time window is located before the occurrence of the abnormality, and the fourth physical motion characteristic is used to characterize the physical motion behavior that causes the analyte sensor to detach from the body surface. A third physiological signal feature is determined within a second abnormal time window, wherein the second abnormal time window is located after the occurrence of the abnormality, and the third physiological signal feature is used to characterize the recovery state of the target object's physiological signals after the occurrence of the abnormality. The presence or absence of the analyte sensor is determined based on the fourth physical motion characteristic and the third physiological signal characteristic.

8. The method according to claim 1, characterized in that, The method further includes: When the physiological signal characteristics indicate that the rate of decrease in the analyte value of the target object is greater than a preset rate and the acceleration signal in the physical motion characteristics exhibits an instantaneous spike, the second artifact confidence level is used as the artifact confidence level, wherein the second artifact confidence level is used to indicate that the analyte fluctuation event is an artifact; or... If the artifact confidence level is within the target value range, the observation window used to determine the physical motion characteristics and the physiological signal characteristics is extended, and the third artifact confidence level is re-determined based on the data within the observation window.

9. The method according to claim 1, characterized in that, The method further includes: Obtain the historical artifact events corresponding to the target object, and determine the artifact feature vector of the historical artifact events; The compression pattern of the target object is determined based on the artifact feature vector, wherein the compression pattern is used to reflect the compression habit of the target object on the analyte sensor in a non-moving state; During the compression period indicated by the compression mode, the detection trigger threshold corresponding to the physical motion feature is reduced, wherein the detection trigger threshold is used to reflect the sensitivity to compression detection.

10. An electronic device, characterized in that, include: A memory and a processor, the memory being used to store program instructions; the processor being connected to the memory and used to execute the method for artifact detection according to any one of claims 1 to 9.

Images