A depression risk assessment system and method based on a multi-sign bracelet

Through the multi-signal bracelet system, physiological and non-physiological data are collected and processed, and deep learning models are constructed and personalized updates are used to solve the problem of low accuracy and reliability of depression risk assessment in the existing technology, and the long-term risk of depression is achieved Evaluate the effectiveness of timely medical advice.

CN115568852BActive Publication Date: 2024-11-29BEIJING ANXINXI TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202211130079.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-09-16
Publication Date
2024-11-29
Estimated Expiration
2042-09-16

AI Technical Summary

Technical Problem

The prior art is difficult to collect multiple sign information of patients with depression for a long time through wearable devices, resulting in low accuracy and reliability of the risk assessment of depression, and prone to misjudgment and misdiagnosis.

Method used

Adopting a system based on multi-signal bracelet, including data acquisition module, data cleaning module, feature extraction module, deep learning module and risk assessment module, collect and process physiological and non-physiological data through multiple sensors, and use deep learning models to build and personalize Updated to achieve long-term depression risk assessment and timely medical advice.

Benefits of technology

A complete system process from extracting data from multi-signal bracelets to real-time assessment of depression risk. It does not require complex large-scale monitoring equipment, and can monitor and provide accurate depression risk index for a long time, remind users to seek medical treatment in a timely manner, improving the accuracy of assessment and reliability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115568852B_ABST
    Figure CN115568852B_ABST
Patent Text Reader

Abstract

The present invention discloses a depression risk assessment system and method based on a multi-sign bracelet, wherein the system includes a data acquisition module: used for continuously collecting a variety of multi-sign information photoelectric signals that affect depression; a data cleaning module: performing data cleaning and data processing on the collected multi-sign information photoelectric signals to obtain multi-sign information that affects depression; a feature extraction module: performing feature extraction on the multi-sign information to obtain high-dimensional statistical features of periodic changes; a deep learning module: using a deep learning model to classify the extracted high-dimensional statistical features to obtain depression classification probability information; an online reasoning module: performing online reasoning on the obtained depression classification probability information to obtain a depression risk index; a risk assessment module: performing risk assessment according to the depression risk index and giving corresponding medical advice. The present invention can continuously detect human depression through a multi-sign bracelet and achieve accurate assessment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of depression risk assessment, and more particularly to a depression risk assessment system and method based on a multi-sign bracelet. Background Art

[0002] Depression is a common and serious mental illness that places a heavy burden on affected individuals and the public health system. In the past 30 years, the incidence of depression in my country has increased 10 times, but the medical prevention and treatment of depression in China is still at a relatively early stage. The recognition rate of depression in hospitals is less than 20%, and less than 10% of patients receive relevant drug treatment. However, the diagnosis of depression is usually delayed, and patients often choose to seek medical treatment only when the condition has become serious, which leads to a long treatment cycle for depression and easy recurrence. Medical research has found that the heart rate, exercise status and other physiological signs of patients with depression are different from those of normal people.

[0003] However, current technical solutions mostly use sensors such as electroencephalograms and electrocardiograms that are difficult to integrate into wearable devices to monitor and collect user's physical signs and activities, making it difficult to collect user's physical sign information for a long time. Therefore, most existing technical solutions judge physical signs in a short period of time, while the physiological signs of different users vary significantly in different time periods, which leads to misjudgment and misdiagnosis in the risk assessment of depression, resulting in limited assessment results and low reliability.

[0004] Therefore, how to improve the accuracy of depression assessment and provide a depression risk assessment system and method based on a multi-sign bracelet is an urgent problem that technicians in this field need to solve. Summary of the invention

[0005] In view of this, the present invention provides a depression risk assessment system and method based on a multi-sign bracelet, which is used to monitor the user's depression risk in the long term and provide the user's long-term depression risk index.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] On the one hand, the present invention discloses a depression risk assessment system based on a multi-sign bracelet, comprising:

[0008] Data acquisition module: including a multi-sign bracelet, which is used to continuously collect a variety of multi-sign information photoelectric signals that affect depression;

[0009] Data cleaning module: used to clean and process the continuously collected multi-sign information photoelectric signals to obtain multi-sign information that affects depression;

[0010] Feature extraction module: used to extract features from multiple vital signs information and obtain high-dimensional statistical features of periodic changes;

[0011] Deep learning module: used to classify the extracted high-dimensional statistical features using a deep learning model to obtain depression classification probability information;

[0012] Online reasoning module: used to perform online reasoning on the acquired depression classification probability information to obtain the depression risk index;

[0013] Risk assessment module: used to conduct risk assessment based on the depression risk index and give corresponding medical advice based on the risk assessment results.

[0014] Preferably, the multi-sign bracelet includes a photoplethysmography sensor, an electrocardiogram sensor, an IMU inertial sensor, a skin conductivity sensor, an electromyography sensor and a temperature sensor.

[0015] Preferably, the multi-vitamin information includes real-time heart rate, resting heart rate, heart rate variability, respiratory rate, pulse rate, blood oxygen saturation, activity range, number of steps, sleep status and wrist movement frequency.

[0016] Preferably, the data cleaning module performs data cleaning and data processing on the collected multi-sign information photoelectric signals, specifically including: data screening on the collected continuous multi-sign information photoelectric signals, data processing on the screened multi-sign information photoelectric signals, and obtaining multi-sign information affecting depression.

[0017] Preferably, the deep learning model includes a recurrent neural network or a long short-term memory network.

[0018] Another aspect of the present invention discloses a depression risk assessment system based on a multi-sign bracelet, comprising the following steps:

[0019] Data collection: Use a multi-sign bracelet to continuously collect a variety of multi-sign information photoelectric signals that affect depression;

[0020] Data cleaning: Data cleaning and data processing are performed on the continuously collected multi-sign information photoelectric signals to obtain multi-sign information that affects depression;

[0021] Feature extraction: Extract features from multiple signs to obtain high-dimensional statistical features of periodic changes;

[0022] Deep learning model construction: Build a deep learning model and use it to classify the extracted high-dimensional statistical features to obtain depression classification probability information;

[0023] Online reasoning: Perform online reasoning on the obtained depression classification probability information to obtain the depression risk index;

[0024] Risk assessment: Risk assessment is conducted based on the depression risk index, and corresponding medical advice is given based on the risk assessment results.

[0025] Preferably, data cleaning is performed on the continuously collected multi-sign information, specifically including:

[0026] Performing data screening on the collected multi-vital sign information, and screening out the multi-vital sign information collected when the user wears the multi-vital sign bracelet;

[0027] The screened multi-sign information is subjected to data thinning processing.

[0028] Preferably, feature extraction is performed on the continuous multi-sign information after data cleaning to obtain high-dimensional statistical features of periodic changes, specifically including:

[0029] The continuous multi-sign information obtained after data processing is used as the original signal, and the trend signal is extracted from the original signal using the seasonal_decompose method;

[0030] Use the centralized moving average method to decompose trend signals;

[0031] Subtract the value of the decomposed trend signal from the value of the original signal at the same time point to get the value of the periodic signal at that time point, and connect the values ​​of all periodic signals within a certain period of time to get the periodic signal with periodic changes;

[0032] The obtained periodic signal is converted into high-dimensional statistical features of periodic changes for deep learning model input.

[0033] Preferably, a deep learning model is constructed, and the extracted high-dimensional statistical features are classified using the deep learning model to obtain depression classification probability information, specifically including:

[0034] Constructing a deep learning model, wherein the deep learning model includes a recurrent neural network or a long short-term memory network;

[0035] Marking the existing multi-sign information of depression patients as positive data, marking the existing multi-sign information of healthy people as negative data, and using the negative data and positive data to train a deep learning model;

[0036] The extracted high-dimensional statistical features of periodic changes are input into the trained deep learning model to obtain the probability information of depression classification.

[0037] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a depression risk assessment system based on a multi-sign bracelet, which has the following beneficial effects:

[0038] 1. The present invention completes a complete system process of extracting data from a multi-sign bracelet - server-side modeling - real-time assessment of the user's depression risk. The multi-sign bracelet can collect long-term continuous multi-sign signals (multi-sign signals include physiological data such as heart rate and respiratory rate that are directly affected by depression, and non-physiological data such as the number of steps, exercise speed and exercise acceleration that are indirectly affected by depression) to complete the creation of a depression risk assessment model without the need for other complex large-scale monitoring equipment (such as audio and video signals-cameras, EEG signals-EEG devices).

[0039] 2. The present invention can complete personalized model updates for different users based on the overall model, making it closer to the designated user.

[0040] 3. The system can monitor the user's depression risk over a long period of time, provide the user with a long-term and short-term depression risk index, and remind the user to seek medical treatment in a timely manner on edge smart devices (such as mobile phones) when the user shows continuous high risk. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0042] Figure 1 A schematic diagram of a method flow of a depression risk assessment system based on a multi-sign bracelet provided in an embodiment of the present invention;

[0043] Figure 2 Schematic diagram of original signal, trend signal and periodic signal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] On the one hand, an embodiment of the present invention discloses a depression risk assessment system based on a multi-sign bracelet, comprising:

[0046] Data acquisition module: including a multi-sign bracelet, which is used to continuously collect a variety of multi-sign information photoelectric signals that affect depression;

[0047] Data cleaning module: used to clean and process the continuously collected multi-sign information photoelectric signals to obtain multi-sign information that affects depression;

[0048] Feature extraction module: used to extract features from multiple vital signs information and obtain high-dimensional statistical features of periodic changes;

[0049] Deep learning module: used to classify the extracted high-dimensional statistical features using a deep learning model to obtain depression classification probability information;

[0050] Online reasoning module: used to perform online reasoning on the acquired depression classification probability information to obtain the depression risk index;

[0051] Risk assessment module: used to conduct risk assessment based on the depression risk index and give corresponding medical advice based on the risk assessment results.

[0052] In a specific embodiment, the multi-sign wristband includes a photoplethysmography sensor, an electrocardiogram sensor, a motion sensor, a six-axis acceleration sensor, a skin electrical sensor, an electromyography sensor, and a temperature sensor. Correspondingly, the multi-sign information includes real-time heart rate, resting heart rate, heart rate variability, respiratory rate, pulse rate, blood oxygen saturation, sleep status, exercise steps, activity range, and wrist movement frequency.

[0053] Another aspect of the present invention discloses a depression risk assessment method based on a multi-sign bracelet, and the method is implemented based on a depression risk assessment system of the multi-sign bracelet.

[0054] The flowchart of the method of depression risk assessment system based on multi-sign bracelet is as follows Figure 1 As shown. The edge refers to the multi-sign bracelet and its connected smart devices (such as mobile phones, tablets, etc.), which are used for data collection, user interaction, risk display and other functions; the server refers to the server in the cloud, which is used for data processing, model training and other functions. Specifically, the edge includes a data acquisition module for collecting multi-sign information (mainly composed of multi-sign bracelets), an online reasoning module and a risk assessment module for user interaction and risk display; the server includes a data cleaning module, a feature extraction module and a deep learning module. The steps of the depression risk assessment method based on the multi-sign bracelet are as follows:

[0055] 1) Data collection

[0056] The data is collected through a multi-sign bracelet. The bracelet integrates sensors including but not limited to photoplethysmography sensors (PPG sensors), electrocardiogram sensors (ECG sensors), IMU inertial sensors, temperature sensors, etc., which can collect photoelectric signals of physiological data including but not limited to real-time heart rate, resting heart rate, heart rate variability, respiratory rate, sleep conditions, wrist movement frequency, and optical signals of non-physiological data such as exercise steps, exercise speed, and exercise acceleration. In the present invention, physiological data is used as multi-sign information directly affected by depression, and non-physiological data is used as multi-sign information indirectly affected by depression. Physiological data and non-physiological data together constitute multi-sign information that can reflect the user's depression. This information can be used for subsequent model construction and risk reasoning. The multi-sign bracelet can be used for half a month or more on a single charge, so continuous and long-term vital sign signals of the user can be obtained.

[0057] 2) Data cleaning

[0058] After the photoelectric signal data is collected by the multi-sign bracelet, it will be sent to the server at a regular interval, and the server will perform subsequent data cleaning and data processing.

[0059] Whenever the wearer exercises, the relative position between the wearer's wrist and the bracelet will change significantly, which will cause errors in the data measured by the sensor; at the same time, the wearer may take off the bracelet due to reasons such as charging the bracelet, and the photoelectric signal data collected at this time is meaningless. Therefore, the first step in data cleaning is to eliminate the photoelectric signal data collected when not worn and the unstable photoelectric signal data with large errors. By judging whether the user is wearing the device and calculating the data fluctuation of the vital sign signal, the photoelectric signal data of not wearing the bracelet and the photoelectric signal data with large fluctuations are eliminated.

[0060] The second step is to calculate the vital sign data. After eliminating the error photoelectric signal data, the system will calculate the user's heart rate, pulse rate, wrist movement frequency and other multi-vital sign information data based on the collected original sensor photoelectric signal data.

[0061] 3) Feature extraction

[0062] The multi-sign data inferred by the server contains noise. If the original data is used directly, it will lead to large model deviation. Therefore, it is necessary to extract high-dimensional statistical features from the original data within a certain time window. Specifically, assuming the time window is 5 seconds, the statistical values ​​of this time period will be calculated based on the vital sign data such as heart rate, pulse rate, wrist movement frequency, etc. collected within 5 seconds, including but not limited to the mean, standard deviation, kurtosis, skewness, etc. of the signal.

[0063] To obtain multi-sign information that affects depression, the seasonal_decompose method is used to extract trend signals from multi-sign information recorded over a long period of time, and the centralized moving average method is used to decompose the trend signal; the value of the periodic signal at that time point is obtained by subtracting the value of the decomposed trend signal from the value of the original signal at the same time point, and the values ​​of all periodic signals within a certain period of time are connected to obtain a periodic signal of periodic changes. The periodic signal can be used to represent its long-term changes in signs, and it is usually stable within a certain range. Figure 2 As shown, the original signal is the upper figure, the trend signal is the middle figure, and the periodic signal is the lower figure. In the embodiment of the present invention, the original signal refers to the multi-sign information after data cleaning. The obtained periodic signal is further converted into a high-dimensional statistical feature of periodic changes for deep learning model input.

[0064] 4) Deep learning model construction

[0065] Constructing a deep learning model, wherein the deep learning model includes RNN (Recurrent Neural Network) or LSTM (Long Short-Term Memory);

[0066] Existing multi-sign information data are collected from patients with depression and healthy people, divided into positive and negative data and used as training sets. Deep learning models such as RNN (Recurrent Neural Network) or LSTM (Long Short-Term Memory) are used for training to complete the coarse-grained classification of depression based on big data.

[0067] For different users, the extracted periodic feature signals are input into the model to further fine-tune the model and perform personalized adaptation on the model, so that the model can be closer to the user's actual situation.

[0068] Since both RNN and LSTM are time series models, that is, the input signal is time series data, the present invention uses RNN and LSTM deep learning models to classify the collected continuous multi-vital sign information data.

[0069] 5) Online Reasoning

[0070] After the server-side model is built, it will be sent to the user's edge smart device, such as a smartphone. During actual use, whenever the user needs to assess the current risk of depression, recent data can be automatically input into the model to complete online reasoning and obtain the risk index of depression.

[0071] The original output of the deep learning model is the probability of someone suffering from depression. The classification threshold is usually 0.5, which means that if the probability is greater than 0.5, the person suffers from depression, and if it is less than 0.5, the person does not suffer from depression. The risk index is the probability of someone suffering from depression, which is the original output of the model.

[0072] 6) Risk Assessment

[0073] The system will automatically complete the depression risk assessment every day and automatically generate the user's long-term and short-term depression risk index. The user can make timely adjustments and choose to seek medical treatment based on the assessment index. After the user uses the assessment function to obtain the depression risk index, the system will push corresponding treatment suggestions to the user. For users with a lower risk index, they will be advised to exercise actively and arrange a regular and stable schedule; for users with a higher risk, they are advised to seek medical attention in time. At the same time, when the user's depression risk index is greater than 0.7 for many consecutive days, the system will promptly send a feedback warning to the user to remind the patient to seek medical treatment in time.

[0074] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0075] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A depression risk assessment system based on a multi-sign bracelet, characterized in that: include: Data acquisition module: including a multi-sign bracelet, which is used to continuously collect a variety of multi-sign information photoelectric signals that affect depression; Data cleaning module: used to clean and process the continuously collected multi-sign information photoelectric signals to obtain multi-sign information that affects depression; The multi-sign information includes real-time heart rate, resting heart rate, heart rate variability, respiratory rate, pulse rate, blood oxygen saturation, number of steps, wrist movement frequency, activity range and sleep status; Feature extraction module: used to extract features from multiple vital signs information and obtain high-dimensional statistical features of periodic changes; specifically includes: The continuous multi-sign information obtained after data processing is used as the original signal, and the trend signal is extracted from the original signal using the seasonal_decompose method; Use the centralized moving average method to decompose trend signals; Subtract the value of the decomposed trend signal from the value of the original signal at the same time point to get the value of the periodic signal at that time point, and connect the values ​​of all periodic signals within a certain period of time to get the periodic signal with periodic changes; Convert the obtained periodic signal into high-dimensional statistical features of periodic changes for deep learning model input; Deep learning module: used to classify the extracted high-dimensional statistical features using a deep learning model to obtain depression classification probability information; Online reasoning module: used to perform online reasoning on the acquired depression classification probability information to obtain the depression risk index; Risk assessment module: used to conduct risk assessment based on the depression risk index and give corresponding medical advice based on the risk assessment results.

2. The depression risk assessment system based on a multi-sign bracelet according to claim 1 is characterized in that: The multi-sign bracelet includes a photoplethysmography sensor, an electrocardiogram sensor, an IMU inertial sensor, a skin conductivity sensor, an electromyography sensor and a temperature sensor.

3. The depression risk assessment system based on a multi-sign bracelet according to claim 1 is characterized in that: The data cleaning module performs data cleaning and data processing on the collected multi-sign information photoelectric signals, specifically including: data screening on the collected continuous multi-sign information photoelectric signals, data processing on the screened multi-sign information photoelectric signals, and obtaining multi-sign information affecting depression.

4. The depression risk assessment system based on a multi-sign bracelet according to claim 1, characterized in that: The deep learning model includes a recurrent neural network or a long short-term memory network.

5. A method for assessing depression risk based on a multi-sign bracelet, characterized in that: The following steps are involved: Data collection: Use a multi-sign bracelet to continuously collect a variety of multi-sign information photoelectric signals that affect depression; Data cleaning: Data cleaning and data processing are performed on the continuously collected multi-sign information photoelectric signals to obtain multi-sign information that affects depression; The multi-sign information includes real-time heart rate, resting heart rate, heart rate variability, respiratory rate, pulse rate, blood oxygen saturation, number of steps, wrist movement frequency, activity range and sleep status; Feature extraction: Extract features from multiple signs to obtain high-dimensional statistical features of periodic changes; specifically, including: The continuous multi-sign information obtained after data processing is used as the original signal, and the trend signal is extracted from the original signal using the seasonal_decompose method; Use the centralized moving average method to decompose trend signals; Subtract the value of the decomposed trend signal from the value of the original signal at the same time point to get the value of the periodic signal at that time point, and connect the values ​​of all periodic signals within a certain period of time to get the periodic signal with periodic changes; Convert the obtained periodic signal into high-dimensional statistical features of periodic changes for deep learning model input; Deep learning model construction: Build a deep learning model and use it to classify the extracted high-dimensional statistical features to obtain depression classification probability information; Online reasoning: Perform online reasoning on the obtained depression classification probability information to obtain the depression risk index; Risk assessment: Risk assessment is conducted based on the depression risk index, and corresponding medical advice is given based on the risk assessment results.

6. The method for assessing depression risk based on a multi-sign bracelet according to claim 5, characterized in that: Data cleaning and data processing are performed on the continuously collected multi-sign information photoelectric signals, including: The collected multi-sign information photoelectric signals are screened to select the multi-sign information photoelectric signals collected when the user wears the multi-sign wristband; The screened multi-sign information photoelectric signal is subjected to data processing to obtain multi-sign information affecting depression.

7. The method for assessing depression risk based on a multi-sign bracelet according to claim 5, characterized in that: A deep learning model is constructed and used to classify the extracted high-dimensional statistical features to obtain depression classification probability information, including: Constructing a deep learning model, wherein the deep learning model includes a recurrent neural network or a long short-term memory network; Marking the existing multi-sign information of depression patients as positive data, marking the existing multi-sign information of healthy people as negative data, and using the negative data and positive data to train a deep learning model; The extracted high-dimensional statistical features of periodic changes are input into the trained deep learning model to obtain the probability information of depression classification.

Citation Information

Patent Citations

  • Network card traffic model construction method, traffic prediction method, equipment and storage medium

    CN113139643A

  • Blod glucose data set optimization for improved hypoglycemia prediction based on machine learning implementation ingestion

    CN113168917A