Method for monitoring unexpected events and electronic device

Abstract

The application discloses a method for monitoring an unexpected event and an electronic device. The method comprises: collecting multi-dimensional data of a target object, wherein the multi-dimensional data comprises physiological parameters and motion parameters associated with analyte value abnormalities; in the case that the motion parameters indicate that the target object is in a non-motion state, predicting the multi-dimensional data by using a prediction model to obtain a first risk score, wherein the prediction model is trained based on historical unexpected event data, and the first risk score is used to quantitatively represent the probability of the occurrence of a first unexpected event caused by the analyte value abnormality; and determining whether to give a warning for the first unexpected event according to the first risk score. The application solves the technical problem of the lag of the prediction and intervention mechanism for unexpected events in analyte monitoring in the related art.

Description

Technical Field

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

[0002] In continuous analyte monitoring systems, abnormal analyte values ​​are often accompanied by unexpected events such as confusion, limb weakness, sudden falls, respiratory depression, or even sudden cardiac death. However, the active detection of these unexpected events in analyte monitoring relies solely on fixed analyte threshold alarms for post-event alerts, resulting in a high false alarm rate and an inability to intervene in a timely manner when the patient is still capable of self-rescue.

[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 monitoring unexpected events, in order to at least solve the technical problem that the prediction and intervention mechanisms for unexpected events in the monitoring of analytes are lagging behind in related technologies.

[0005] According to one aspect of the embodiments of this application, a method for monitoring unexpected events is provided, comprising: collecting multidimensional data of a target object, wherein the multidimensional data includes physiological parameters and motion parameters associated with abnormal analytical values; when the motion parameters indicate that the target object is in a non-motion state, using a prediction model to predict the multidimensional data to obtain a first risk score, wherein the prediction model is trained based on historical unexpected event data, and the first risk score is used to quantify the probability that the abnormal analytical values ​​lead to the occurrence of a first unexpected event; and determining whether to issue an early warning for the first unexpected event based on the first risk score.

[0006] According to another aspect of the embodiments of this application, a monitoring device for unexpected events is also provided, comprising: a data acquisition module for acquiring multidimensional data of a target object, wherein the multidimensional data includes physiological parameters and motion parameters associated with abnormal analytical values; a prediction module for predicting the multidimensional data using a prediction model to obtain a first risk score when the motion parameters indicate that the target object is in a non-motion state, wherein the prediction model is trained based on historical unexpected event data, and the first risk score is used to quantify the probability that abnormal analytical values ​​lead to the occurrence of a first unexpected event; and an early warning module for determining whether to issue an early warning for the first unexpected event based on the first risk score.

[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 monitoring unexpected events.

[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-mentioned method for monitoring unexpected events 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 aforementioned method for monitoring unexpected events.

[0010] In this embodiment, a method is adopted whereby, when the target object is in a non-moving state, a prediction model trained based on historical accidental event data is used to fuse and analyze multidimensional data including physiological parameters and motion parameters associated with abnormal analyte values. By quantifying the probability of an accidental event caused by abnormal analyte values ​​through a first risk score, and determining whether to issue an early warning for the accidental event based on the first risk score, the aim of identifying accidental risks and triggering intervention in advance while the patient is still conscious is achieved. This realizes the transformation from a single blood glucose threshold alarm to proactive prediction of multiple physical quantity precursor features, significantly reducing the false alarm rate and extending the patient's self-rescue time window. In turn, it solves the technical problem of the lag in the prediction and intervention mechanism of accidental events in analyte monitoring. 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 1 This is a hardware structure block diagram of a computer terminal for a method of monitoring unexpected events according to an embodiment of this application;

[0013] Figure 2 This is a flowchart of a method for monitoring unexpected events according to an embodiment of this application;

[0014] Figure 3 This is an architecture diagram of an accident monitoring system according to an embodiment of this application;

[0015] Figure 4 This is a structural diagram of an analytical monitoring device for a method of monitoring unexpected events according to an embodiment of this application;

[0016] Figure 5 This is a schematic diagram of the overall process of a method for monitoring unexpected events according to an embodiment of this application;

[0017] Figure 6 This is a schematic diagram of the structure of an accident monitoring device according to an embodiment of this application. Detailed Implementation

[0018] 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.

[0019] 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.

[0020] To better understand the embodiments of this application, the technical terms involved in the embodiments of this application are explained below:

[0021] Continuous Analyte Monitoring System: A medical monitoring system that monitors the concentration of analytes in interstitial fluid in real time through an electrochemical sensor implanted under the skin and transmits the data to a receiving terminal to assist users in managing daily physiological parameters. In this embodiment, the system provides basic analyte concentration and trend data to trigger anomaly detection and perform fusion analysis with multi-physical quantity data.

[0022] Skin impedance sensor: An electronic device used to measure the electrical impedance characteristics of skin and subcutaneous tissue at different frequencies. Its measured values ​​reflect ion concentration, degree of sweating and autonomic nervous system function. In the embodiments of this application, the sensor is used to capture changes in sweating and tissue fluid ion concentration caused by sympathetic nerve excitation in the early stage of analyte abnormality, as an important physiological characteristic signal for identifying the prodromal period of coma.

[0023] Synthetic acceleration amplitude: Based on the X, Y, and Z axis acceleration data collected by the triaxial accelerometer, a scalar value is calculated through vector synthesis. It is used to characterize the intensity of the overall motion of the target object. In the embodiments of this application, this parameter is used to distinguish between the resting state and the moving state. The multi-physical quantity prediction model is only enabled when the synthetic acceleration is lower than a certain threshold (such as 0.1g) to eliminate motion interference and ensure the accuracy of the prediction of unexpected risks caused by the abnormality of the analyzed object.

[0024] For ease of explanation, let's take a continuous glucose monitoring (CGM) system as an example. CGM is widely used in the daily blood glucose management of diabetic patients. Patients obtain real-time blood glucose concentration data through CGM, which guides adjustments to diet, exercise, and medication, effectively reducing the risk of hypoglycemia and hyperglycemia. However, the CGM systems used in these technologies still have the following technical shortcomings in practical use:

[0025] (1) Lack of proactive detection of hypoglycemia-related accidents: When diabetic patients experience severe hypoglycemia, they may experience confusion, weakness in the limbs, sudden falls, or even coma. At this time, patients often lose the ability to actively seek help. The relevant CGM system can only trigger an alarm through the blood glucose threshold and cannot detect abnormal body posture (such as sudden fall or prolonged stillness) or behavioral changes in patients.

[0026] (2) The single blood glucose threshold alarm has limitations: hypoglycemic events and accidents such as falls do not necessarily occur simultaneously. Some patients experience a temporary increase in blood glucose after a fall due to stress response, which may cause the alarm threshold to not be triggered, thus missing the opportunity for rescue. In addition, for patients with "hypoglycemia unawareness", the CGM system lacks an independent body status perception channel, resulting in blind spots in safety monitoring.

[0027] (3) The "accelerometer + blood glucose" fusion scheme used in the relevant technologies is still a post-event detection and cannot predict the prodromal period of coma. The relevant technologies attempt to combine accelerometers with CGM for fall detection. The basic logic is: after detecting acceleration characteristics (free fall, impact, stationary), a fall is determined, and then the blood glucose value is combined to confirm whether it is caused by hypoglycemia. However, this type of scheme has two fundamental defects: on the one hand, it cannot predict "imminent hypoglycemic coma", that is, when the patient's blood glucose is extremely low but has not yet fallen, consciousness may have gradually been lost, and the acceleration signal does not have a significant change; it is completely ineffective for "silent hypoglycemia" (that is, the patient falls into a coma directly due to hypoglycemia without drastic changes in posture). In other words, the relevant technologies lack the ability to identify the electrophysiological characteristics of the prodromal period of hypoglycemic coma and cannot provide early warning when the patient is still conscious.

[0028] (4) Failure to utilize other measurable physiological signals during hypoglycemia: Clinical studies have shown that during severe hypoglycemia, patients experience autonomic dysfunction, manifested as decreased skin impedance (cold sweats), peripheral vasoconstriction leading to a decrease in interstitial fluid temperature, etc. These physiological parameter changes often occur before blood glucose levels reach the danger threshold and can serve as important precursors to the worsening of hypoglycemia. However, relevant CGM systems only focus on glucose concentration measurement and do not integrate or utilize these auxiliary physical quantities for trend prediction, resulting in the system's inability to issue dietary intervention recommendations when the patient is still capable of self-rescue (e.g., still conscious and able to swallow).

[0029] (5) Unable to provide proactive warning and self-rescue guidance before hypoglycemic coma: Even if the relevant CGM system detects hypoglycemia, it can only issue a local alarm and cannot automatically distinguish whether the patient has entered a state of confusion or pre-coma. Furthermore, it cannot provide specific intervention suggestions (such as "immediately supplement 15g of carbohydrates") when the patient is still able to rescue themselves. Patients face extremely high safety risks when alone (such as sleeping at night, living alone at home, or exercising outdoors).

[0030] To address the aforementioned technical problems, this application provides corresponding solutions, which are detailed below.

[0031] The accident monitoring method provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal for implementing a method for monitoring unexpected events is shown. Figure 1 As shown, the computer terminal 10 may include a processor 102, a memory 104 for storing data, and a transmission module 106 for communication functions. The processor 102 may include one or more processors. For ease of explanation, Figure 1 The illustration uses multiple processors, including a first processor 102a, a second processor 102b, ..., an nth processor 102n. Processor 102 may include, but is not limited to, processing devices such as microcontroller units (MCUs) or field-programmable gate arrays (FPGAs). In addition, it may 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.

[0032] 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).

[0033] 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 accident monitoring 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 realizing the aforementioned accident monitoring 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.

[0034] 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 data interaction with external networks or devices via wired and / or wireless network connections.

[0035] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows the user to interact with the user interface of the computer terminal 10. Furthermore, the computer terminal 10 can be connected to a cursor control device and a keyboard via an input / output interface. The cursor control device (e.g., a mouse, trackball, or touchpad) is primarily used to transmit directional information and command selections to the terminal and control the movement of the cursor on the display; the keyboard is primarily used to receive letters, numbers, and other control commands input by the user, thereby enabling interaction between the user and the computer terminal 10.

[0036] 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.

[0037] In the above operating environment, this application provides an embodiment of a method for monitoring unexpected events. 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.

[0038] Figure 2 This is a flowchart of a method for monitoring unexpected events according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following steps:

[0039] Step S202: Collect multidimensional data of the target object, including physiological and motion parameters associated with abnormal values ​​of the analyzed object.

[0040] In step S202 above, multidimensional data refers to a data set composed of various physical quantities or physiological signals. For example, it may include a comprehensive information set containing externally measurable signals (physiological parameters) reflecting the internal metabolic state of the target object and motion signals (motion parameters) reflecting the behavioral state of the target object. It should be noted that physiological parameters refer to biophysical quantities reflecting the autonomic nervous system function, microcirculation state, and metabolic characteristics of the target object, including skin impedance, interstitial fluid temperature, heart rate, and heart rate variability. These are used to capture and analyze early autonomic nervous system responses triggered by abnormal values. For example, in the early stages of hypoglycemia, sympathetic nerve excitation leads to sweating, which in turn causes a decrease in skin impedance and a drop in local temperature. Motion parameters refer to physical quantities reflecting the macroscopic limb activity state and posture changes of the target object, including acceleration data collected by accelerometers, such as synthetic acceleration amplitude and posture angles. These are used to distinguish the true activity state of the target object and eliminate the interference of motion artifacts on the interpretation of physiological parameters.

[0041] It should be noted that the analytes include, but are not limited to, at least one of glucose, ketone bodies (trihydroxybutyric acid), lactic acid, and uric acid. For ease of explanation, the embodiments of this application use the glucose concentration in interstitial fluid as an example, but it should be understood that the application scenarios of this application are not limited to this.

[0042] Abnormal analyte values ​​refer to a state in which the concentration level of a specific physiological analyte in the target body deviates from the normal physiological range or fluctuates drastically, leading to physiological dysfunction. This includes, but is not limited to, hypoglycemia (i.e., blood glucose concentration is below the safe threshold or shows a rapid downward trend), lactic acid accumulation, hyperkalemia, or hypokalemia. Different abnormality judgment conditions can be set according to different analytes, and no limitation is made here.

[0043] In some embodiments of this application, the wearable monitoring device worn on the skin surface of the target object incorporates an electrochemical glucose oxidase sensor, a skin impedance sensor, a temperature sensor, and a triaxial accelerometer. A microprocessor sequentially or synchronously reads the raw signals from each sensor via a hardware interface. Specifically, the glucose oxidase sensor measures the glucose concentration in the interstitial fluid of the subcutaneous tissue at a frequency of once per minute; the skin impedance sensor measures the complex impedance of the skin and subcutaneous tissue at a specific frequency (e.g., 1 kHz, 10 kHz) to reflect sweat gland activity; the temperature sensor measures the temperature of the interstitial fluid close to the skin at a frequency of 0.2 Hz; and the accelerometer acquires triaxial acceleration data at a high sampling frequency.

[0044] In order to accurately capture early physiological stress signs caused by analyte abnormalities, in some embodiments of this application, multidimensional data can be determined in the following ways: acquiring analyte data, skin impedance data, temperature data, and acceleration data of the target object from multiple sensors respectively; determining the target analyte value and the rate of change of the analyte value within a preset time period based on the analyte data; determining the low-frequency impedance value and the rate of change of impedance within a preset time period based on the skin impedance data; determining the target temperature value and the rate of change of temperature within a preset time period based on the temperature data; determining the first composite acceleration amplitude based on the acceleration data; and using the target analyte value, the rate of change of the analyte value, the low-frequency impedance value, the rate of change of impedance, the target temperature value, the rate of change of temperature, and the first composite acceleration amplitude as multidimensional data.

[0045] It should be noted that the analyte data refers to the original concentration values ​​of specific biochemical substances in the target object directly measured by implanted or contact sensors; the low-frequency impedance value refers to the complex impedance modulus value of the skin and subcutaneous tissue measured at a specific low-frequency test frequency (such as 1 kHz); the temperature data refers to the original temperature data of interstitial fluid or epidermis continuously collected by a temperature sensor close to the skin surface; and the acceleration data refers to the original linear acceleration signal of the target object in space collected by a triaxial accelerometer.

[0046] By integrating multidimensional data such as acceleration, blood glucose, skin impedance, temperature, and heart rate, early signs of hypoglycemic exacerbation can be identified before the patient falls or loses consciousness. Specifically, taking continuous glucose monitoring as an example, this includes:

[0047] (1) Blood glucose characteristics: Let the original blood glucose sequence be G[t], and the sampling interval be Δt. Taking Δt = 1 minute as an example, calculate:

[0048] Current blood glucose level (i.e., target analyte value): ;

[0049] Blood glucose change rate (i.e., analyte value change rate): (Unit: mmol / L / min).

[0050] It should be noted that the above calculation uses a 3-minute difference to smooth short-term fluctuations as an example.

[0051] (2) Skin impedance characteristics: complex impedance measurement values Where Z(f) represents the impedance magnitude, which represents the total resistance encountered when current passes through tissue, and Φ(f) represents the phase angle of the complex impedance measured at frequency f, defined as follows:

[0052] Low-frequency impedance mode: This reflects the activity of the skin's stratum corneum and sweat glands;

[0053] High-frequency impedance mode: This reflects the ion concentration in subcutaneous tissue;

[0054] Impedance change rate: (Unit: kΩ / min), the detection impedance drops rapidly (autonomic nerve stress).

[0055] (3) Temperature characteristics: interstitial fluid temperature T[t] (unit: °C), where:

[0056] Rate of temperature change: (Unit: °C / min)

[0057] (4) Motion characteristics: Taking the magnitude of the composite acceleration (i.e., the magnitude of the first composite acceleration) as an example: , where A syn This represents the first composite acceleration amplitude, where i represents the coordinate axis index variable of the accelerometer, N represents the total number of coordinate axes of the accelerometer, and a i This represents the acceleration component value on the i-th coordinate axis at time t.

[0058] Step S204: When the motion parameters indicate that the target object is in a non-motion state, a prediction model is used to predict the multidimensional data to obtain a first risk score. The prediction model is trained based on historical accident data, and the first risk score is used to quantify the probability that the abnormal value of the analyzed object will lead to the occurrence of the first accident.

[0059] In step S204 above, the non-motion state refers to the physiological state in which the target object is at rest, asleep, or slightly active (such as sitting or lying down) and has not performed large-scale limb movements. This is used to eliminate the interference of motion artifacts on the interpretation of physiological parameters. For example, sweating during strenuous exercise will lead to a decrease in skin impedance and an increase in body temperature. If the motion state is not distinguished at this time, the system may misjudge it as autonomic nerve stress caused by hypoglycemia, thereby generating a false positive alarm.

[0060] In some embodiments of this application, if the first synthesized acceleration amplitude is less than a preset amplitude, the target object is determined to be in a non-motion state. That is, to eliminate motion interference, for example, only when... Enable the multi-physical quantity prediction model when the content is <0.1g (resting state).

[0061] Predictive models are algorithmic models based on machine learning or statistics, including but not limited to lightweight logistic regression models, decision trees, or neural network models. They are used to map multidimensional input features to risk probability values. They are trained on historical data and learn to analyze the nonlinear relationship between physical values, physiological parameters, and unexpected events, thereby having the ability to extract precursor features of unexpected events such as coma from complex physiological signals.

[0062] The first unexpected event refers to a serious physiological crisis event that is directly caused or significantly aggravated by abnormal analytical values ​​(such as hypoglycemia), including but not limited to hypoglycemic coma, loss of consciousness or resulting falls, respiratory depression, sudden cardiac death, etc.

[0063] In some embodiments of this application, the extracted feature vectors (such as blood glucose, blood glucose change rate, low-frequency impedance, impedance change rate, temperature, temperature change rate, and synthetic acceleration in the above embodiments) can be used as input vectors and substituted into the pre-trained linear equations to obtain the first risk score.

[0064] Specifically, taking continuous blood glucose monitoring as an example, a lightweight logistic regression model is used, with the input feature vector... ,in:

[0065] x1= (Current blood glucose); x2 = G[t] (rate of change in blood glucose); x3 = [t](1kHz impedance); x4= [t](rate of change of impedance); x5=T[t](temperature); x6= [t](rate of temperature change); x7= [t](composite acceleration, close to 0 at rest).

[0066] The model outputs the probability of coma precursor P∈[0,1], which is calculated using the following formula:

[0067]

[0068] Among them, the weight w and the bias b are obtained by training with clinical data (the training dataset contains features of the first preset duration (e.g., 30 minutes) before hypoglycemic coma and normal state samples).

[0069] To achieve adaptive prediction in multi-device collaborative scenarios, prediction can be performed as follows: determine the feature vector corresponding to the multi-dimensional data; determine the prediction model corresponding to the feature vector from multiple candidate prediction models, wherein the input feature vectors of the multiple candidate prediction models have different dimensions; use the prediction model to predict the multi-dimensional data to obtain the first risk score.

[0070] It should be noted that a candidate prediction model refers to a set of pre-trained algorithm models that differ in structure, input dimension, or complexity. For example, it may include a base model and an augmentation model. The base model may only accept input of a first preset dimension (relying solely on the wearable device's own sensors), while the augmentation model may accept input of a second preset dimension (used to fuse data such as heart rate and electrocardiogram from external devices such as smartwatches).

[0071] In some embodiments of this application, the multidimensional data includes first raw data collected by an analyte monitoring device and / or second raw data collected by an external device connected to the analyte monitoring device. Based on this, the feature vector corresponding to the multidimensional data can be determined in the following manner: when the second raw data is absent, a first feature vector with a first preset dimension is determined based on the first raw data; when the second raw data exists, a second feature vector with a second preset dimension is determined based on the first and second raw data, wherein the second preset dimension is greater than the first preset dimension, and when there are missing features in the second preset dimension of the second feature vector, they are filled with preset values.

[0072] It should be noted that the first raw data refers to the raw values ​​of physiological signals directly collected by the built-in sensors of the wearable analyte monitoring device (such as a CGM transmitter), which may include data such as interstitial fluid glucose concentration, skin impedance, skin temperature, and acceleration. The second raw data refers to the raw values ​​of enhanced physiological signals collected by an external wearable device (such as a smartwatch, smart bracelet, or ECG patch) that is wirelessly connected to the analyte monitoring device, which may include data such as heart rate, heart rate variability, ECG QT interval, and skin conductance.

[0073] Specifically, taking continuous blood glucose monitoring as an example, when there is no second original data, the microprocessor reads the glucose concentration, 1kHz skin impedance, interstitial fluid temperature and triaxial acceleration data collected by the CGM device, and after feature calculation (differential calculation of change rate), it sequentially extracts the current blood glucose value, blood glucose change rate, low frequency impedance modulus, impedance change rate, current temperature value, temperature change rate and composite acceleration amplitude, and arranges these seven values ​​in a fixed order to form a 7-dimensional first feature vector.

[0074] When second raw data is available, the system receives data from the smartwatch, including heart rate, heart rate variability, heart rate LF / HF ratio, and skin conductance level. First, it constructs a vector containing seven basic features. Then, it concatenates four (or more) enhanced features from the smartwatch to form an 11-dimensional (or 14-dimensional) second feature vector. If the smartwatch only provides some enhanced features (e.g., only heart rate, not skin conductance), zero values ​​(preset values) are filled into the corresponding enhanced feature positions to maintain a fixed vector dimension.

[0075] In the absence of external devices, the above embodiments utilize basic feature vectors to ensure the independent operation and basic monitoring functions of the system. When external devices are present, the enhanced feature vectors are used to fuse multimodal physiological signals, which can improve the accuracy and sensitivity of identifying the prodromal phase of unexpected events such as hypoglycemic coma.

[0076] To facilitate understanding of the data interaction with external devices and the process of constructing feature vectors, the following explanations will be provided in conjunction with some specific examples.

[0077] The system can connect to the user's existing smartwatch, smart bracelet or other wearable devices via wireless communication (such as BLE) to obtain multi-dimensional physiological data from their built-in sensors. This data can be used as optional enhancement features to input into the prediction model, thereby significantly improving the sensitivity and specificity of identifying the precursor to hypoglycemic coma. It should be noted that these external sensor parameters are not mandatory, but if the user wears the corresponding device, the system can automatically adapt and enable the optimization mode.

[0078] As an example, the available external sensor parameters and their correlation with hypoglycemia are shown in Table 1.

[0079] Table 1: Sensor Parameter Table.

[0080]

[0081] Taking receiving a data packet every 5 seconds via a watch as an example, the following enhanced features can be defined:

[0082] (1) Rate of change of heart rate:

[0083] (bpm / min)

[0084] Where HR[t] represents the heart rate value collected at time t, and Δt hr This indicates the time interval used to calculate the rate of change in heart rate.

[0085] During the hypoglycemic stress period, heart rate (HR) first increases and then decreases; the reversal of the sign of the rate of change can serve as an early warning indicator.

[0086] (2) HRV low-frequency / high-frequency power ratio (requires watch to support RR interval output):

[0087] (m / Hz)

[0088] Where P(f) represents the power spectral density function, the power distribution in the frequency domain is obtained by performing a Fourier transform on the RR interval sequence (the time interval between consecutive heartbeats), LF represents low-frequency power, and HF represents high-frequency power.

[0089] It should be noted that a value >1.5 indicates sympathetic dominance, which is commonly seen in hypoglycemic stress.

[0090] (3) Rate of change in skin conductivity:

[0091]

[0092] Where SCL[t] represents the skin conductance level collected at time t. When SCL > 0.5 μS / s and lasts for more than 30 seconds, it indicates sudden sweating. Δt eda This indicates the time interval used to calculate the rate of change in skin conductivity.

[0093] (4) QT interval variation (ECG required):

[0094]

[0095] Where QT[t] represents the QT interval value collected at time t, QT baseline It represents the baseline value of the QT interval. When ΔQT>30ms and blood glucose<4.0mmol / L, it has extremely high specificity.

[0096] (5) Non-invasive blood glucose trend characteristics (NIR): Extract the absorbance ratio of the second wavelength to the reference wavelength. Its short-term decline slope is positively correlated with the decline in blood glucose. Here, ρ represents the ratio of absorbance at the second wavelength to absorbance at the reference wavelength, Aλ2 represents the light absorption intensity or absorbance measured at the second wavelength λ2, and Aλ1 represents the light absorption intensity or absorbance measured at the first wavelength λ1.

[0097] The basic model's input feature vector x is 7-dimensional (blood glucose, rate of change of blood glucose, 1kHz impedance, rate of change of impedance, temperature, rate of change of temperature, and composite acceleration). When external devices are available, the system dynamically expands to a 14-dimensional feature vector. ,in:

[0098] x8 = HR[t] (heart rate); x9 = (Heart rate variability); x10 = HRV[t] (heart rate variability ratio); x11 = SCL[t] (skin conductance level); x12 = [t](skin conductance change rate); x13=ΔQT[t](QT interval shift, if available); x14=ρ[t](NIR absorbance ratio, if available).

[0099] It should be noted that the augmented model can use the same logistic regression structure, but the weight vector... The bias b needs to be retrained (based on a dataset containing external features). Furthermore, the system automatically identifies the watch model and available sensors via Bluetooth, dynamically adjusts the input dimensions, and fills in missing features with zero values.

[0100] Furthermore, the adaptive switching logic is as follows:

[0101] If the watch is not connected or data is missing, the system will still run the basic model independently;

[0102] If the watch provides certain external features (such as only HR, HRV), the system will enable the lightweight enhancement model;

[0103] If the watch provides all ECG, EDA, NIR, etc., the system will activate the full enhanced model;

[0104] The switching process is transparent to the user and does not increase the power consumption of the core device (external data is sampled by the watch itself and sent via Bluetooth, and the core device only performs fusion calculations).

[0105] For example, when the base model outputs P=0.65 (the medium-risk threshold), but external features show that the HR increases from 85 to 120 bpm, =0.6μS / s, ΔQT=35ms, the enhanced model will recalculate P to 0.85, triggering a high-risk alarm. Conversely, if the basic model gives a false alarm (e.g., due to sweating from environmental heat), but external characteristics show stable HR and no change in SCL, the enhanced model can reduce the risk level and decrease false alarms.

[0106] In some embodiments of this application, when the motion parameters indicate that the target object is in a non-motion state, the following steps may also be performed: when the analyte data in the physiological parameters indicates that the target object has experienced an abnormal analyte value event, the motion intensity of the target object within the monitoring time window is determined based on the motion parameters, and the temperature drop rate of the target object within the monitoring time window is determined based on the temperature data in the physiological parameters, wherein the monitoring time window is a preset time window after the occurrence of the abnormal analyte value event; a second risk score is determined based on the motion intensity and the temperature drop rate; and a warning is issued for the first unexpected event based on the second risk score.

[0107] It should be noted that the analyte data here refers to the analyte concentration data continuously monitored after the target object has experienced an abnormal analyte value event (such as a confirmed diagnosis of hypoglycemia). The monitoring time window refers to a period of time set by the system after the abnormal analyte value event is detected, which is used to retrospectively or prospectively statistically analyze changes in physiological parameters (such as 10 minutes, 20 minutes or 30 minutes after the abnormality occurs).

[0108] Body movement intensity refers to the intensity or activity level of a target's limbs, quantified based on acceleration data within a monitoring time window. This can include, for example, average acceleration amplitude, activity count, or percentage of effective movement time. It reflects the degree of neuromuscular dysfunction caused by hypoglycemia. For instance, in severe hypoglycemia, patients may experience limb weakness, paralysis, or inability to move actively. A significant decrease in body movement intensity is an indicator of whether the patient has lost their ability to save themselves.

[0109] The second risk score refers to a risk quantification index calculated based on two specific dimensions: physical intensity and rate of temperature decrease, after confirming the anomaly of the analyzed material.

[0110] Specifically, regarding body motion intensity, the system monitors whether the synthetic acceleration amplitude of the target object remains below a certain activity threshold (e.g., 0.1g or 0.05g) within a preset time window (e.g., 20 minutes) after the occurrence of an abnormality in the analysis value. If the proportion of time with acceleration below the threshold exceeds a set proportion (e.g., 80%) within this window, or if the continuous static time exceeds a specific duration (e.g., 5 minutes), then the body motion intensity is determined to be "low" or "disappeared".

[0111] Regarding the rate of temperature decrease, the system calculates the rate of temperature change in adjacent time periods at a high frequency (e.g., every 10 seconds) within the monitoring time window, and takes the average of the most recent N rates of change as the current rate of temperature decrease to smooth out noise.

[0112] After determining the exercise intensity and temperature drop rate, a second risk score can be determined as follows: if the exercise intensity is less than the activity threshold and the temperature drop rate is less than a preset rate, the first score value is used as the second risk score; if the exercise intensity drops to zero within the monitoring time window and the temperature drop rate is not less than a preset rate, the second score value is used as the second risk score, wherein the second score value is greater than the first score value; if the exercise intensity is zero, the temperature drop rate is not less than a preset rate, and the duration of the state is greater than a preset duration, the third score value is used as the second risk score, wherein the third score value is greater than the second score value.

[0113] Specifically, (1) the system sets the activity threshold to a synthetic acceleration amplitude of 0.1g and the preset rate to a temperature change rate of -0.1°C / min. If the current body motion intensity (such as the average acceleration over the past minute) is <0.1g and the temperature change rate is >-0.1°C / min (i.e., the decrease is slow or not at all), the risk is determined to be low, and the first score value (such as level label 1 or a specific value such as 0.3) is output.

[0114] (2) If the system detects that within the monitoring time window (e.g., the most recent 5-10 minutes), the synthetic acceleration amplitude is continuously lower than 0.05g or continues to decrease until it reaches 0 (determined as "decreased to zero"), and the temperature change rate is ≤ -0.1°C / min (i.e. the rate of decrease reaches or exceeds the preset rate), then the second score value (e.g., level identifier 2 or specific value such as 0.7) will be output immediately.

[0115] (3) The system starts a timer. When the body movement intensity is detected to be zero and the temperature drop rate is ≥ the preset rate, the timer starts. If the state continues for more than the preset time (e.g., 15 minutes or 30 minutes), the risk is judged to be extremely high and the third score value (e.g., level label 3 or specific value such as 0.9) is output.

[0116] To facilitate understanding of the process of determining the second risk score, the following explanation is provided in conjunction with some specific embodiments.

[0117] Taking continuous glucose monitoring as an example, during the data acquisition phase, the system acquires blood glucose values ​​from the CGM, body motion data from the accelerometer, body temperature data from the temperature sensor, and data from the impedance sensor in real time. During this process, a sensor fault troubleshooting mechanism is executed to ensure that the sensor is only included in subsequent analysis when it is confirmed that the sensor is properly attached and the signal is normal.

[0118] Furthermore, the system makes judgments based on the following criteria: if the blood glucose level is lower than a preset threshold (e.g., 3.5 mmol / L) and shows a continuous downward trend, it is judged as abnormal blood glucose; if, after a hypoglycemic event, the target subject's physical activity intensity remains below the activity threshold for more than 20 consecutive minutes, it is judged as a significant reduction or disappearance of physical activity; if the rate of decrease in core or skin temperature exceeds 0.3℃ / 10 minutes, it is judged as an abnormal temperature drop; at the same time, the system needs to continuously confirm the stability of the impedance signal to eliminate sensor malfunction interference.

[0119] Furthermore, based on the extracted features, the system performs a three-level risk classification: when hypoglycemia is detected accompanied by reduced body movement, a Level 1 warning is triggered, and a gentle reminder is sent via mobile phone; when hypoglycemia is detected accompanied by the disappearance of body movement and a drop in temperature, a Level 2 high-risk alarm is triggered, activating the device's strong vibration and sound alarms, and pushing notifications to preset emergency contacts; if the above high-risk state lasts for more than 15 minutes and the target does not show any body movement response, a Level 3 emergency alarm is triggered, and the system automatically sends a stop infusion command to the insulin pump and calls the emergency center for intervention.

[0120] It should be noted that the triggering times for the first and second risk scores can differ. Specifically, unexpected events can be monitored in the following ways:

[0121] Collect multidimensional data of the target object, including physiological and motor parameters associated with abnormal analytical values; when motor parameters indicate that the target object is in a non-motor state, perform the following operations:

[0122] (1) If the analyte data in the physiological parameters indicate that no abnormal analyte value event has occurred in the target object, then the prediction model is used to predict the multidimensional data to obtain the first risk score, and the first risk score is used to determine whether to issue an early warning for the first unexpected event.

[0123] (2) If the analyte data in the physiological parameters indicates that the target object has an abnormal analyte value event, then the body movement intensity of the target object within the monitoring time window is determined based on the motion parameters, and the temperature drop rate of the target object within the monitoring time window is determined based on the temperature data in the physiological parameters; a second risk score is determined based on the body movement intensity and the temperature drop rate; and a warning is issued for the first unexpected event based on the second risk score.

[0124] Alternatively, if the analyte data in the physiological parameters indicates that the target object has experienced an abnormal analyte value event, a first risk score and a second risk score are determined, and a third risk score is determined based on the two; a warning is then issued for the first unexpected event based on the third risk score.

[0125] It should be noted that the first risk score is mainly used to provide early warning before anomalies occur, the second risk score is used to dynamically assess the deterioration of analytical material anomalies that have already occurred, and the third risk score is used to integrate information from pre-prediction and post-assessment, eliminate the limitations of a single scoring perspective, and ensure the comprehensiveness and accuracy of risk assessment in complex scenarios.

[0126] Step S206: Determine whether to issue an early warning for the first unexpected event based on the first risk score.

[0127] In step S206 above, the first risk score refers to a numerical index calculated by a prediction model that integrates multiple physical quantities, which quantifies the probability that the target object will soon experience hypoglycemic coma or other unexpected events, and the value ranges from 0 to 1.

[0128] In some embodiments of this application, whether to issue a warning for a first unexpected event can be determined in the following ways: if the first risk score is less than a first threshold, no warning event is triggered; if the first risk score is not less than the first threshold and less than a second threshold, and the rate of change of the analytical value is less than a preset rate of change of the analytical value, a first warning event is triggered, wherein the second threshold is greater than the first threshold; if the first risk score is greater than the second threshold, or if the target analytical value is less than a first preset analytical value, and the low-frequency impedance value is less than a preset impedance value, and the temperature change rate is less than a preset temperature change rate, a second warning event is triggered, wherein the priority of the second warning event is higher than the priority of the first warning event.

[0129] Specifically, taking continuous blood glucose monitoring as an example, the system classifies the risk into three levels based on the risk probability P (i.e., the first risk score, ranging from 0 to 1) output by the prediction model, and executes corresponding monitoring or alarm strategies:

[0130] (1) When the risk probability P is less than 0.3, it is judged as low risk, the system maintains normal monitoring status and does not perform additional intervention.

[0131] (2) When the risk probability P is between 0.3 and 0.7, and the rate of change of blood glucose G[t] (unit is mmol / L / min, which represents the change of blood glucose concentration per unit time) is less than -0.5 mmol / L / min, it is judged as medium risk, and the system triggers an early warning (i.e. the first early warning event). Furthermore, the device can push a self-rescue prompt to inform the user that "blood glucose drops rapidly, autonomic nerve abnormality, please replenish carbohydrates immediately".

[0132] (3) When the risk probability P is greater than or equal to 0.7, or when the following composite conditions are met, the system is judged as high risk and triggers an emergency alarm (i.e., the second warning event): the composite conditions require that the current blood glucose value is less than 3.0 mmol / L, the skin impedance change rate Z[t] (in kΩ / min, representing the change in skin impedance per unit time) is less than -0.2 kΩ / min, and the temperature change rate T[t] (in °C / min, representing the change in temperature per unit time) is less than -0.1 °C / min. Furthermore, a local high-decibel buzzer and strong vibration can be activated to remind people around to help; if the patient does not confirm safety by pressing the button within a preset time (e.g., 2 minutes), the system will automatically enter the emergency rescue process; at the same time, a warning message (including real-time location, blood glucose trend, and predicted coma time) will be sent to the preset contact person.

[0133] In some embodiments of this application, the following steps may also be performed: after issuing an early warning using an initial early warning event corresponding to the first risk score, detecting whether the target object returns confirmation information within a preset time window; if no confirmation information is received within the preset time window, issuing an early warning again using a target early warning event, wherein the target early warning event has a higher priority than the initial early warning event.

[0134] Specifically, as an example, when the system determines the risk level to be medium, it controls the wearable device to emit low-frequency vibrations and pushes a notification containing specific guidance text (such as "Please immediately replenish 15g of carbohydrates") via a Bluetooth-connected mobile app (i.e., the initial warning event). After issuing the initial warning, a 2-minute timer is started. The patient can be asked to send a confirmation message by clicking the "I have processed" button on the mobile app or by long-pressing a physical button on the wearable device. If no interaction signal is detected within 2 minutes, it is considered that no confirmation has been received. If no confirmation is received within the time limit, the risk level is automatically upgraded to high risk, and an emergency message is sent to a preset contact (i.e., the target warning event).

[0135] To achieve independent detection of physical accidents such as falls, the following steps can also be performed: when the motion parameters indicate that the target object is in motion, determine the motion characteristics of the target object based on the motion parameters; if the motion characteristics match the preset characteristics, determine that a second accident has occurred.

[0136] It should be noted that the preset features refer to the standard movement pattern templates or feature sets used to characterize specific accidental events (such as falls) through training with a large amount of historical data or clinical statistics. The second accidental event refers to the physical injury event (such as a fall) caused by a sudden change in the posture of the target object. This is different from the loss of consciousness directly caused by hypoglycemia (the first accidental event). The second accidental event focuses on physical falls that occur in hypoglycemia or in a normal state.

[0137] In some embodiments of this application, the motion characteristics of the target object can be determined in the following ways: determining the second composite acceleration amplitude corresponding to the sliding time window based on motion parameters; determining the first motion characteristic and the second motion characteristic of the target object corresponding to the first time period and the second time period respectively based on the second composite acceleration amplitude, wherein the second time period is after the first time period, the first motion characteristic is used to characterize the degree of weightlessness of the target object during the falling process, and the second motion characteristic is used to characterize the impact intensity when the target object collides with the ground; determining the attitude information of the target object based on motion parameters, and determining the third motion characteristic of the target object in the third time period based on the attitude information, wherein the third time period is after the second time period, and the third motion characteristic is used to characterize the horizontality of the target object's attitude after the collision; and using the first motion characteristic, the second motion characteristic, and the third motion characteristic together as the motion characteristics.

[0138] Specifically, taking a fall event as an example, a sliding time window (e.g., window length 2 seconds, step size 0.5 seconds) is used to process the acceleration signal and calculate the composite acceleration amplitude (i.e., the second composite acceleration amplitude): .

[0139] Furthermore, identify the three-stage characteristics of a fall:

[0140] (1) Free fall phase: A drops from about 1g to less than 0.3g in 0.1 seconds.

[0141] (2) Impact phase: A impact peak greater than 2.5g occurs within 0.2 seconds thereafter.

[0142] (3) Attitude level phase: Within 2 seconds after impact, the gravity direction component is extracted by low-pass filtering (the accurate vertical direction can be obtained by fitting multi-axis data), and the pitch angle or horizontal tilt angle is calculated. If the absolute value of the tilt angle is less than 25° and the composite acceleration A < 0.1g (at rest), it is determined to be a valid fall event.

[0143] Furthermore, the system sequentially checks whether the motion characteristics meet three preset conditions: first, whether there are free-fall characteristics (sudden drop in acceleration); second, whether there are impact characteristics (followed by a high-amplitude impact); and third, whether there are stationary horizontal characteristics (the absolute value of the tilt angle after impact is less than 25 degrees and the resultant acceleration is less than 0.1g). Only when all three conditions are met in chronological order is it determined that a second accidental event (fall) has occurred.

[0144] It should be noted that in the fall detection logic, the system strictly verifies the integrity of the three-stage characteristics of "free fall-impact-rest": if a free fall stage signal (sudden drop in acceleration) is detected, but no impact peak matching the characteristics is detected in the subsequent time window (e.g., no significant impact in acceleration), or the target object quickly resumes regular activity within a short period after the impact (e.g., acceleration signal shows gait recovery or limb patting movements), the system determines the event as normal behavior (e.g., sitting down, lying down, or self-adjustment), and then automatically cancels the emergency alarm, recording it only as a normal movement event.

[0145] The above embodiments achieve independent detection of physical accidents such as falls by monitoring motion status and identifying specific motion characteristics. For example, for patients with hypoglycemia, falls are often a direct consequence or complication of hypoglycemic coma. By accurately identifying the second accident, the system can trigger targeted emergency rescue (such as calling for emergency medical treatment for external injuries) and determine whether it is necessary to adjust the hypoglycemia warning strategy (such as falls may interfere with the accuracy of subsequent blood glucose monitoring data), thereby providing comprehensive safety monitoring.

[0146] In some embodiments of this application, a mechanism to prevent accidental triggering can also be set. Specifically, if the user resumes activity within 5 seconds after falling (A>0.2g and gait is regular), the emergency alarm is canceled and only the "get up on your own after falling" event is recorded.

[0147] In some embodiments of this application, the following steps may also be performed: determining the static duration of the target object; triggering a third warning event when the static duration is not less than a first duration and the target analytical value of the target object is less than a second preset analytical value; triggering a fourth warning event when the static duration is not less than the second duration and the target analytical value is less than a third preset analytical value, wherein the second duration is greater than the first duration, the third preset analytical value is less than the second preset analytical value, and the priority of the fourth warning event is higher than the priority of the third warning event.

[0148] It should be noted that coma detection can be performed by combining the duration of stillness, either after a second unexpected event (fall) is detected, or when no second unexpected event occurs (such as when the target is asleep or in a non-moving state).

[0149] Specifically, when the system detects that the user has remained stationary for an extended period (A < 0.05g) for more than a set time threshold (configurable, such as 30 seconds, 1 minute, 5 minutes, etc.), it will make a judgment based on the blood glucose level:

[0150] If the resting time is ≥30 seconds and the blood glucose level is <3.0 mmol / L, then the system will enter the "suspected hypoglycemic coma" pre-alarm state.

[0151] If the resting time is ≥2 minutes and the blood glucose level is <2.8 mmol / L, it is considered a severe hypoglycemic coma event, triggering an emergency response.

[0152] Furthermore, to avoid misjudging normal sleep stillness, the system can identify sleep periods (such as 22:00-07:00) through historical schedules and trigger an alarm only when a longer stillness time threshold is applied during sleep (such as 2 hours of no activity and blood glucose consistently below 3.0 mmol / L).

[0153] It should be noted that event handling priorities can be set based on the severity of the risk, such as following a logical hierarchy of "fall detection first, first risk score (pre-event prediction) and second risk score (post-event assessment) next, blood glucose threshold alarm next, and prolonged inactivity alarm last," to ensure that the highest level of rescue measures can be quickly activated when multiple risks overlap. The specific execution logic is as follows:

[0154] First, when the system simultaneously detects a fall (the second unexpected event) and the blood glucose level is below the danger threshold, or when the system calculates a first or second risk score using a multi-physical quantity fusion model that reaches a high-risk threshold (e.g., P ≥ 0.7, or meets the conditions of hypoglycemia accompanied by loss of body movement and a sudden drop in temperature), it is determined to be a high-risk composite event. At this time, the system immediately skips the routine procedures and directly enters the highest level emergency response mode, simultaneously triggering a strong local alarm, sending a distress message to emergency contacts, and can also link the insulin pump to pause infusion to deal with possible coma or serious trauma.

[0155] Secondly, if only a blood glucose level below the danger threshold is detected but there are no signs of a fall and no high-risk score warning is triggered, the standard hypoglycemia alarm procedure will be followed, and a self-rescue prompt will be sent and a graded warning (such as a medium-risk warning) will be activated to remind the patient to replenish carbohydrates in time.

[0156] Finally, if only fall characteristics are detected but blood glucose levels are within the normal range, and neither the first nor the second risk scores indicate a risk of worsening hypoglycemia, the system will mark it as a "fall event" for local recording and log storage for the patient to review later. This usually does not trigger a medical emergency call unless the user has pre-configured a mandatory alarm option for falls with normal blood glucose levels.

[0157] In some embodiments of this application, the following steps may also be performed: when a warning is issued for a first unexpected event and no target confirmation information is received for the target object, or when the target analysis value of the target object is less than a fourth preset analysis value and the target object meets a preset state, an alarm event is triggered, wherein the preset state includes the target object remaining stationary for a longer than a preset duration or the target object experiencing a second unexpected event.

[0158] Specifically, when the multi-physical quantity fusion model outputs a high risk and the patient does not provide interactive confirmation within a limited time, or when the acceleration module detects a fall / prolonged immobility and blood glucose is below the danger threshold, the system performs the following operations:

[0159] (1) Local alarm: The buzzer or vibration motor is activated to emit a continuous and rapid alarm sound.

[0160] (2) Wireless SOS: Send an SOS signal to a pre-set emergency contact's mobile phone or a third-party emergency rescue platform via BLE / Wi-Fi / cellular network. For example, the SOS message may include, but is not limited to:

[0161] User ID and medical information;

[0162] Current GPS / base station location;

[0163] Blood glucose trend, impedance, temperature, and heart rate data for the last 15 minutes;

[0164] Event type (hypoglycemic coma precursor / fall / prolonged immobility) and predicted time.

[0165] (3) Continuous retry and confirmation: Send once every 10 seconds, for 10 consecutive times or until confirmation is received.

[0166] (4) Remote monitoring terminal response: Emergency contacts can view the patient's status and send instructions via the APP (such as turning off the alarm or delaying the call for help).

[0167] Through steps S202 to S206, a method is adopted where, while the target object is in a non-moving state, a prediction model trained based on historical accidental event data is used to fuse and analyze multidimensional data including physiological and motion parameters associated with abnormal analyte values. By quantifying the probability of an accidental event caused by abnormal analyte values ​​through a first risk score, and determining whether to issue an early warning for the accidental event based on the first risk score, the goal of identifying accidental risks and triggering intervention in advance while the patient is still conscious is achieved. This realizes the transformation from a single blood glucose threshold alarm to proactive prediction of multiple physical quantity precursor features, significantly reducing the false alarm rate and extending the patient's self-rescue time window. In turn, it solves the technical problem of the lag in the prediction and intervention mechanism of accidental events in analyte monitoring.

[0168] Figure 3 This is an architecture diagram of an accident monitoring system according to an embodiment of this application, such as... Figure 3 As shown, the system includes an analyte monitoring device 30 and an external device 32, wherein:

[0169] The analyte monitoring device 30 includes a sensor 302, a microprocessor 304, and a wireless communication module 306. The sensor 302 is used to collect multidimensional physiological and motion data of the target object in real time. The microprocessor 304 is used to preprocess, extract features, and perform fusion analysis on the multidimensional data collected by the sensor 302. The wireless communication module 306 is used to realize data interaction between the analyte monitoring device 30 and external devices 32 and cloud servers.

[0170] External device 32 is connected to analyte monitoring device 30 and is used to provide additional multidimensional physiological parameters to analyte monitoring device 30 via a wireless connection (such as Bluetooth BLE).

[0171] It should be noted that, Figure 3 The monitoring system for unexpected events shown is used to perform... Figure 2 The method for monitoring unexpected events shown, therefore Figure 2 The relevant explanations and instructions in the monitoring methods for unexpected events also apply to... Figure 3 The monitoring system for unexpected events shown will not be described in detail here.

[0172] Figure 4 This is a structural diagram of a wearable monitoring device according to an embodiment of the present application for a method of monitoring unexpected events, as shown in the figure. Figure 4 As shown, the wearable monitoring device (i.e., analyte monitoring equipment) integrates:

[0173] Accelerometer 402: Used to collect the user's acceleration data in real time. The sampling frequency is adjustable. The data includes static gravity components and dynamic motion components.

[0174] Blood glucose sensor 404: An electrochemical glucose oxidase sensor used to measure glucose concentration in subcutaneous interstitial fluid, with a configurable sampling frequency (e.g., once per minute).

[0175] Skin impedance sensor 406: can be used in conjunction with or independently of the CGM electrode to measure the impedance of the skin and subcutaneous interstitial fluid at multiple frequencies (such as 1 kHz, 10 kHz, 100 kHz), reflecting ion concentration, degree of sweating and autonomic nervous system function.

[0176] Temperature sensor 408: Placed close to the CGM probe or skin surface, it is used to continuously monitor the temperature of interstitial fluid, reflecting the peripheral circulation status, with an accuracy of ±0.1°C and a sampling rate of 0.2Hz.

[0177] Microprocessor 410: Electrically connected to the aforementioned sensors, used to perform motion feature extraction, multi-physical quantity fusion analysis, hypoglycemic coma precursor feature recognition, graded alarm and emergency response decision-making.

[0178] Wireless communication module 412: used to send warning information, blood glucose data and location information to user terminal or cloud server, and can receive heart rate data from external devices (such as smartwatches) to achieve multi-device collaboration.

[0179] Figure 5 This is a schematic diagram of the overall process of a method for monitoring unexpected events according to an embodiment of this application, as shown below. Figure 5 As shown, it includes:

[0180] The S502 collects multi-dimensional data. Specifically, it collects physiological and motion data such as acceleration, blood glucose, skin impedance, and temperature in real time through sensors, or further receives enhanced features such as heart rate and electrocardiogram transmitted from external devices to complete the synchronous acquisition of multimodal data.

[0181] S504, determine whether the state is in motion. Specifically, calculate the magnitude of the composite acceleration. If the magnitude exceeds the preset motion threshold, the state is determined to be in motion; otherwise, it is determined to be in a non-motion state (resting state), in order to distinguish between daily activities and potential risk states.

[0182] S506, Non-motor state, determine risk score. Specifically, in the resting state, features such as blood glucose change rate, impedance change rate, and temperature change rate are extracted and input into the fusion prediction model to calculate the first risk score, or when hypoglycemia has occurred, a second risk score is calculated based on body movement and temperature data.

[0183] S508, determine whether the preset threshold has been reached. Specifically, compare the calculated risk score with the preset low-risk, medium-risk, and high-risk thresholds. If the score exceeds the medium-risk or high-risk threshold, a warning condition is triggered.

[0184] S510, the preset threshold is reached, and the first unexpected event occurs. Specifically, when the risk score indicates a risk of hypoglycemic coma or a state of severe hypoglycemia, it is determined as the first unexpected event, and the process jumps to the early warning procedure to execute a graded response.

[0185] S512, Motion State, Determine Motion Characteristics. Specifically, under the motion state, the acceleration signal is processed using a sliding time window to extract temporal features such as free fall, impact peak, and attitude level, and to construct structured data reflecting the motion pattern.

[0186] S514, determine whether it matches the preset features. Specifically, compare the extracted motion features with the preset fall feature template to verify whether it fully includes the three stages of "free fall-impact-stationary" to confirm whether it is a real fall.

[0187] S516, Matching, a second unexpected event occurs. Specifically, if the motion feature perfectly matches the preset fall feature and false positive interference is excluded, then a second unexpected event (fall) is determined to have occurred, and the event is recorded for subsequent analysis.

[0188] S518, Warning. Specifically, based on the type of event and risk level, tiered response measures are implemented, including local audible and visual alarms, push notifications for self-rescue guidance, sending distress messages to emergency contacts, and suspending insulin pump infusions.

[0189] It should be noted that the specific implementation of each of the above steps can be found in [reference needed]. Figure 2 The various embodiments of the method for monitoring unexpected events shown are not described in detail here.

[0190] The beneficial effects of this application are as follows:

[0191] (1) Related technologies only detect when a patient falls or loses consciousness, which may cause them to miss the golden window for rescue. The embodiments of this application monitor the early signs of autonomic dysfunction such as decreased skin impedance, decreased temperature and abnormal heart rate, and can issue an early warning 10 to 15 minutes in advance when the patient is still conscious and capable of self-rescue, and provide specific dietary intervention suggestions, thereby significantly reducing the incidence of severe hypoglycemic coma.

[0192] (2) For hypoglycemic coma without falls, but only manifested as gradual loss of consciousness, accelerometers cannot provide effective signals. The embodiments of this application utilize the fusion monitoring of multiple physical quantities such as skin impedance and temperature, which can identify metabolic crisis trends even when the patient is in a static state (such as sleeping at night), effectively avoiding nocturnal hypoglycemia-related deaths.

[0193] (3) The embodiments of this application only run the prediction model in the resting state (synthetic acceleration less than 0.1g), which effectively prevents misjudgment caused by sweating or changes in body temperature due to exercise. In addition, a 2-minute patient confirmation window mechanism is set to further filter false warnings and ensure the accuracy of alarms.

[0194] (4) By accessing the heart rate data of the smartwatch via wireless communication, there is no need to integrate an additional heart rate sensor into the continuous glucose monitoring device, thereby reducing the complexity of system integration and power consumption. Users can freely choose whether to wear the watch according to their needs, which enhances the flexibility and compatibility of the application.

[0195] (5) In the medium-risk warning stage, this application embodiment not only issues an alarm, but also actively pushes specific action suggestions such as "replenish carbohydrates" and "sit down and rest" to help patients complete self-rescue before losing consciousness, reduce dependence on the emergency system, and improve the success rate of self-rescue.

[0196] (6) The embodiments of this application have the ability to automatically identify available sensors and adaptively adjust model weights. Whether the connected watch is a low-end watch that only provides heart rate data or a high-end watch that provides electrocardiogram, electrodermal activity and near-infrared spectral data, it can achieve better performance than the basic model, demonstrating good backward compatibility and upward scalability.

[0197] Figure 6 This is a structural diagram of an accident monitoring device according to an embodiment of this application, such as... Figure 6 As shown, the device includes:

[0198] The acquisition module 602 is used to acquire multidimensional data of the target object, including physiological parameters and motion parameters associated with abnormal values ​​of the analyzed object.

[0199] The prediction module 604 is used to predict multidimensional data using a prediction model when the motion parameters indicate that the target object is in a non-motion state, and obtain a first risk score. The prediction model is trained based on historical accident data, and the first risk score is used to quantify the probability that the abnormal value of the analyzed object will lead to the occurrence of the first accident.

[0200] The early warning module 606 is used to determine whether to issue an early warning for the first unexpected event based on the first risk score.

[0201] It should be noted that, Figure 6 The monitoring device for unexpected events shown is used to perform... Figure 2 The method for monitoring unexpected events shown, therefore Figure 2 The relevant explanations and instructions in the monitoring methods for unexpected events also apply to... Figure 6 The monitoring device for unexpected events shown will not be described in detail here.

[0202] 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 steps of the method for monitoring unexpected events in various embodiments of this application.

[0203] 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 accident monitoring method in various embodiments of this application by running the computer program.

[0204] This application also provides a computer program product, including computer instructions that, when executed by a processor, implement the steps of the unexpected event monitoring method in various embodiments of this application.

[0205] This application also provides a computer program that, when executed by a processor, implements the steps of the unexpected event monitoring method in various embodiments of this application.

[0206] 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.

[0207] 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.

[0208] 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.

[0209] 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.

[0210] 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.

[0211] 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.

[0212] 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 monitoring unexpected events, characterized in that, include: Collect multidimensional data of the target object, wherein the multidimensional data includes physiological parameters and motion parameters associated with abnormal values ​​of the analyzed object; When the motion parameters indicate that the target object is in a non-motion state, a prediction model is used to predict the multidimensional data to obtain a first risk score. The prediction model is trained based on historical accidental event data, and the first risk score is used to quantify the probability that the abnormal value of the analyzed object will cause the first accidental event to occur. Based on the first risk score, determine whether to issue an early warning for the first unexpected event.

2. The method according to claim 1, characterized in that, The method further includes: When the motion parameters indicate that the target object is in motion, the motion characteristics of the target object are determined based on the motion parameters; If the motion characteristics match the preset characteristics, a second unexpected event is determined to have occurred.

3. The method according to claim 1, characterized in that, A prediction model is used to predict the multidimensional data to obtain a first risk score, including: Determine the feature vector corresponding to the multidimensional data; A prediction model corresponding to the feature vector is determined from multiple candidate prediction models, wherein the input feature vectors of the multiple candidate prediction models have different dimensions; The prediction model is used to predict the multidimensional data to obtain the first risk score.

4. The method according to claim 1, characterized in that, Collect multidimensional data of the target object, including: Data on the target object, including analyte data, skin impedance data, temperature data, and acceleration data, are acquired from multiple sensors. The target analyte value and the rate of change of the analyte value within a preset time period are determined based on the analyte data. The low-frequency impedance value and the rate of impedance change within the preset time period are determined based on the skin impedance data. The target temperature value and the rate of temperature change within the preset time period are determined based on the temperature data. The magnitude of the first composite acceleration is determined based on the acceleration data; The target analytical value, the rate of change of the analytical value, the low-frequency impedance value, the rate of change of impedance, the target temperature value, the rate of change of temperature, and the first synthetic acceleration amplitude are used as the multidimensional data.

5. The method according to claim 4, characterized in that, Determining whether to issue an early warning for the first unexpected event based on the first risk score includes: If the first risk score is less than the first threshold, no warning event will be triggered; If the first risk score is not less than the first threshold and less than the second threshold, and the rate of change of the analytical value is less than the preset rate of change of the analytical value, a first warning event is triggered, wherein the second threshold is greater than the first threshold; If the first risk score is greater than the second threshold, or if the target analyte value is less than the first preset analyte value, the low-frequency impedance value is less than the preset impedance value, and the temperature change rate is less than the preset temperature change rate, a second warning event is triggered, wherein the priority of the second warning event is higher than the priority of the first warning event.

6. The method according to claim 2, characterized in that, Determining the motion characteristics of the target object based on the motion parameters includes: The second composite acceleration amplitude within the sliding time window is determined based on the motion parameters. Based on the second composite acceleration amplitude, the first motion characteristic and the second motion characteristic of the target object are determined in the first time period and the second time period, respectively. The second time period is located after the first time period. The first motion characteristic is used to characterize the degree of weightlessness of the target object during the falling process, and the second motion characteristic is used to characterize the intensity of the impact when the target object collides with the ground. The pose information of the target object is determined based on the motion parameters, and the third motion feature of the target object in a third time period is determined based on the pose information, wherein the third time period is after the second time period, and the third motion feature is used to characterize the horizontal degree of the target object's pose after the collision. The first motion feature, the second motion feature, and the third motion feature are collectively used as the motion feature.

7. The method according to claim 1, characterized in that, The method further includes: Determine the duration of stillness of the target object; If the static duration is not less than the first duration and the target analysis value of the target object is less than the second preset analysis value, a third early warning event is triggered. If the static duration is not less than the second duration and the target analyte value is less than the third preset analyte value, a fourth warning event is triggered, wherein the second duration is greater than the first duration, the third preset analyte value is less than the second preset analyte value, and the priority of the fourth warning event is higher than the priority of the third warning event.

8. The method according to claim 1, characterized in that, When the motion parameters indicate that the target object is in a non-motion state, the method further includes: When the analyte data in the physiological parameters indicates that the target object has experienced an abnormal analyte value event, the body movement intensity of the target object within the monitoring time window is determined based on the movement parameters, and the temperature drop rate of the target object within the monitoring time window is determined based on the temperature data in the physiological parameters, wherein the monitoring time window is a preset time window after the occurrence of the abnormal analyte value event; A second risk score is determined based on the intensity of bodily movement and the rate of temperature decrease; Based on the second risk score, determine whether to issue an early warning for the first unexpected event.

9. The method according to claim 8, characterized in that, A second risk score is determined based on the said physical intensity and the said rate of temperature decrease, including: If the physical intensity is less than the activity threshold and the temperature drop rate is less than the preset rate, the first score will be used as the second risk score. If the body dynamic intensity drops to zero within the monitoring time window and the temperature drop rate is not less than the preset rate, the second score value is used as the second risk score, wherein the second score value is greater than the first score value. When the body dynamic intensity is zero, the temperature drop rate is not less than the preset rate, and the duration of the state is greater than the preset duration, the third score is used as the second risk score, wherein the third score is greater than the second score.

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 monitoring unexpected events as described in any one of claims 1 to 9.

Images