Abnormality monitoring and early warning system for psychiatric patient

Abstract

The invention discloses a psychiatric patient abnormity monitoring and early warning system, which relates to the technical field of medical health information, and comprises the steps of collecting patient information in real time, preprocessing the collected data, extracting features, fusing to form a multi-modal feature set, carrying out preliminary emotional state evaluation on the multi-modal features, and carrying out early warning. Setting an initial early warning threshold according to the basic information of the patient and the preliminary evaluation result, dynamically adjusting the early warning threshold according to the historical data and the emotion evaluation result, when the emotional state of the patient exceeds the dynamically adjusted early warning threshold, analyzing the behavior pattern and the physiological signal of the patient, extracting text data from the social media interaction record, and sending the text data to the user. And identifying the emotion triggering content by using a natural language processing technology. And specific events causing emotion changes are identified in combination with the information, and personalized intervention suggestions are generated. According to the method and the system, medical staff can respond to emergency situations quickly, customized treatment schemes can be provided for patients, and finally the purpose of improving mental health conditions is achieved.

Description

Technical Field

[0001] The present invention relates to the field of medical and health information technology, and particularly to an abnormal monitoring and early warning system for psychiatric patients. Background Art

[0002] With the increase in social pressure and the growing emphasis on mental health, the early identification and monitoring of psychiatric patients have become an important public health issue. Traditional psychiatric diagnosis and treatment mainly rely on doctors' clinical experience and face-to-face communication, which is not only time-consuming and laborious but also difficult to achieve continuous monitoring. Therefore, existing psychiatric patient monitoring systems still have some limitations, especially in personalized early warning and real-time intervention. When most psychiatric patient monitoring systems evaluate the emotional state, they often focus on a single data source, such as only paying attention to physiological signals or behavioral characteristics, while ignoring the complementarity and synergy between different data sources, which limits the accuracy and sensitivity of the monitoring system. At the same time, due to the significant individual differences in mental illnesses, fixed thresholds cannot meet the personalized needs of all patients, resulting in warning signals that may be too frequent or not timely enough. Summary of the Invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the present invention provides an abnormal monitoring and early warning system for psychiatric patients to solve the problems of inaccurate emotional state evaluation, non-personalized warning thresholds, and lack of effective recognition and intervention of emotional trigger events in existing psychiatric patient monitoring systems.

[0005] To solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides an abnormal monitoring and early warning system for psychiatric patients, which includes a data collection and preprocessing module that collects patients' information in real time and performs preprocessing;

[0007] A feature extraction and fusion module that extracts features from the preprocessed patients' information and then performs fusion to form multi-modal features;

[0008] An emotional state evaluation and threshold initialization module that obtains a preliminary emotional evaluation result based on the multi-modal features and an emotion recognition model, and sets an initial warning threshold according to the patient's personal basic information and the preliminary emotional evaluation result;

[0009] A personalized warning threshold dynamic adjustment module that uses machine learning algorithms to dynamically adjust the warning threshold according to the patient's historical data and the preliminary emotional evaluation result;

[0010] An emotion trigger event recognition module, when the patient's emotional state exceeds the dynamically adjusted warning threshold, analyzes the patient's behavior patterns and physiological signals before and after the emotional change through a long short-term memory network, extracts text data from the patient's social media interaction records, uses natural language processing technology to identify emotion-triggering content, and combines the results of the behavior patterns, physiological signals, and text data to identify the events that cause the patient's emotional change;

[0011] A personalized intervention recommendation generation module generates intervention recommendations based on the identified emotion trigger events and the patient's current emotional state.

[0012] As a preferred solution of the abnormal monitoring and warning system for psychiatric patients of the present invention, wherein: the information of the patient is collected by a high-definition camera and a microphone to capture the patient's facial expressions and voices in real time, and the physiological signals of the patient's heart rate and skin conductance are collected by wearable devices, and the collected video and voice data are subjected to noise reduction, filtering, and smoothing processing.

[0013] As a preferred solution of the abnormal monitoring and warning system for psychiatric patients of the present invention, wherein: feature extraction is performed based on the preprocessed patient information, emotion features are extracted from the facial expression information, emotional features are extracted from the voice data, and features of heart rate variability and skin conductance change trends are extracted from the physiological signals;

[0014] The features are fused to form multi-modal features.

[0015] As a preferred solution of the abnormal monitoring and warning system for psychiatric patients of the present invention, wherein: based on the multi-modal features and the emotion recognition model, a preliminary emotion assessment result is obtained.

[0016] As a preferred solution of the abnormal monitoring and warning system for psychiatric patients of the present invention, wherein: the patient's personal basic information is age, gender, and diagnosis result, and an initial warning threshold T is set according to the patient's personal basic information and the preliminary emotion assessment result. T is divided into three levels, indicating the fluctuation level of the patient's emotion. The first level is emotional stability, the second level is emotional fluctuation, and the third level is large emotional fluctuation.

[0017] As a preferred solution of the abnormal monitoring and warning system for psychiatric patients of the present invention, wherein: the patient's historical data includes historical emotional state, historical physiological signals, historical behavior patterns, diagnosis results, and treatment responses;

[0018] Using machine learning algorithms, the warning threshold is dynamically adjusted according to the patient's historical data and the preliminary emotion assessment result;

[0019] The dynamically adjusted warning threshold is divided into three levels, the first level is emotional stability, the second level is emotional fluctuation, and the third level is large emotional fluctuation.

[0020] As a preferred embodiment of the abnormal monitoring and early warning system for psychiatric patients of the present invention, when the emotional state of the patient exceeds the dynamically adjusted early warning threshold, the long short-term memory network is used to analyze the behavior patterns and physiological signals before and after the emotional change of the patient, and a comprehensive evaluation of the behavior patterns and physiological signals is obtained.

[0021] As a preferred embodiment of the abnormal monitoring and early warning system for psychiatric patients of the present invention, text data is extracted from the patient's social media interaction records, and natural language processing technology is used to identify emotional trigger content points.

[0022] As a preferred embodiment of the abnormal monitoring and early warning system for psychiatric patients of the present invention, by combining the results of behavior patterns, physiological signals and text data, the events that cause the emotional change of the patient are identified.

[0023] As a preferred embodiment of the abnormal monitoring and early warning system for psychiatric patients of the present invention, according to the events that cause the emotional change of the patient and the patient's current emotion, through natural language generation technology, easy-to-understand guidance suggestions are provided for the patient;

[0024] The generated guidance suggestions will be sent to the patient's guardians, caregivers and doctors at the same time for assisting the next treatment and intervention measures.

[0025] The beneficial effects of the present invention are as follows: The abnormal monitoring and early warning system for psychiatric patients aims to accurately evaluate the emotional state of patients by real-time collecting and preprocessing the facial expressions, voices and physiological signal data of patients. The system adopts multi-modal feature extraction and fusion technology, combined with deep learning algorithms, to ensure comprehensive capture of the emotional changes of patients from different dimensions. In order to further enhance the practical value of the system, the system can also automatically identify the key events that cause emotional fluctuations and generate personalized intervention suggestions based on these events. This not only helps medical staff to quickly respond to emergencies, but also provides customized treatment plans for patients, ultimately achieving the purpose of improving mental health status. The application of this system significantly improves the efficiency and quality of mental health management and provides strong support for clinical practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0027] Figure 1 It is a flowchart of the abnormal monitoring and early warning system for psychiatric patients in Embodiment 1.

[0028] Figure 2 It is the decision diagram of the abnormal monitoring and early warning system for psychiatric patients in Embodiment 1. Specific implementation manners

[0029] To make the above objects, features, and advantages of the present invention more obvious and understandable, the specific implementation manners of the present invention will be described in detail below with reference to the accompanying drawings of the specification.

[0030] Many specific details are set forth in the following description to facilitate a thorough understanding of the present invention, but the present invention may be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0031] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other with other embodiments.

[0032] Embodiment 1, referring to Figure 1 and Figure 2 is the first embodiment of the present invention. This embodiment provides an abnormal monitoring and early warning system for psychiatric patients, including the following steps:

[0033] S1. Data acquisition and preprocessing module, which collects the information of patients in real time and performs preprocessing.

[0034] The collection of the information of patients is to capture the facial expressions and voices of patients in real time through a high-definition camera and a microphone. The high-definition camera can capture subtle changes in expressions, which is very important for accurately identifying the emotional state of patients. Voice recordings can be used to analyze voice features such as intonation and rhythm to evaluate the emotional state of patients; physiological signals such as heart rate and skin conductance of patients are collected through wearable devices. These physiological signals reflect the immediate reactions of the patients' bodies and are crucial for monitoring emotional changes. Noise reduction processing is performed on the collected video and voice data to ensure data quality; filtering and smoothing processing are performed on the physiological signals to remove noise interference.

[0035] S2. Feature extraction and fusion module, which extracts features from the preprocessed patient information and then performs fusion to form multi-modal features.

[0036] Feature extraction is performed based on the preprocessed patient information. Emotional features are extracted from the facial expression information, including but not limited to the movement patterns of facial muscles (such as smiling, frowning, etc.), as well as the change trends and intensities of these movement patterns.

[0037] Extract emotional features from speech data, and extract emotional information from speech through speech emotion recognition technology. This information may include changes in intonation, volume, speech rate, etc., and the association between these changes and specific emotional states.

[0038] Extract features of heart rate variability and skin conductance change trends from physiological signals. Heart rate variability reflects the degree of change in cardiac rhythm and is usually used to evaluate the balance state of the autonomic nervous system (especially the sympathetic and parasympathetic nerves); the change trend of skin conductance (also known as skin resistance or galvanic skin response) reflects the change in sweat gland activity, which is related to the degree of emotional arousal.

[0039] Fuse the features to form multimodal features, and its expression is:

[0040]

[0041] Among them, M represents the fused multimodal features, F(X) represents the feature fusion function, X represents the feature matrix of emotional features, affective features, and physiological signal features, n represents the number of feature vectors, and x i represents the i-th feature vector, and α i represents the attention weight.

[0042] S2.1. α i is the attention weight, which is calculated by the attention mechanism and represents the importance of each feature. The attention weight α i has the following expression:

[0043]

[0044] Among them, α i represents the attention weight, C a represents the weight matrix of the attention mechanism, U a represents the bias vector of the attention mechanism, and x Y represents an index variable. Here, Y is used to traverse all elements in the feature vector.

[0045] The fused multimodal dataset contains all relevant information extracted from facial expressions, speech, and physiological signals, and is used for subsequent emotional state evaluation and warning threshold adjustment.

[0046] S3. Emotional state evaluation and threshold initialization module. Based on multimodal features and an emotion recognition model, obtain preliminary emotion evaluation results, and set an initial warning threshold according to the patient's personal basic information and preliminary emotion evaluation results.

[0047] Based on multimodal features and an emotion recognition model, obtain preliminary emotion evaluation results, and its expression is:

[0048]

[0049] Among them, E(M) represents the preliminary emotion assessment result, μ represents the average emotion intensity, σ represents the degree of fluctuation of the emotion intensity, z represents a continuous variable used for integral calculation, which represents a potential emotion intensity value. The integral is a weighted sum of all possible emotion intensity values, dz represents the integral of the variable z, and it is a differential symbol.

[0050] S3.1. The personal basic information of the patient is age, gender, and diagnosis result. Among them, the diagnosis result can be the professional judgment of a doctor, a previous diagnosis report, or the result obtained through other medical means.

[0051] Set an initial warning threshold T according to the patient's personal basic information and the preliminary emotion assessment result, which represents the fluctuation level of the patient's emotion.

[0052] The value range of T is [0, 1]. T is divided into two levels. 0.1 to 0.5 is the first level, indicating that the patient's emotional state is relatively stable, and the warning system will reduce the alarm frequency and reduce unnecessary interventions; 0.5 to 1.0 is the second level, indicating that the patient's emotional state may have large fluctuations, and the warning system will increase the alarm threshold to take intervention measures in a timely manner.

[0053] S4. A personalized warning threshold dynamic adjustment module uses machine learning algorithms to dynamically adjust the warning threshold according to the patient's historical data and the preliminary emotion assessment result.

[0054] The patient's historical data includes historical emotional state, historical physiological signals, historical behavior patterns, diagnosis results, and treatment responses;

[0055] Use machine learning algorithms to dynamically adjust the warning threshold according to the patient's historical data and the preliminary emotion assessment result. Its expression is:

[0056]

[0057] Among them, T' represents the dynamically adjusted warning threshold, H represents the historical emotional state, P represents the historical physiological signals, B represents the historical behavior patterns, D represents the diagnosis result, R represents the treatment response, m represents the number of patients, p represents the number of diagnosis results, q represents the number of gender codes, r represents the number of ages, t represents the number of diagnosis results, vr represents the number of gender codes, A b represents the age of the b-th patient and is used in the numerator to calculate the contribution of each patient's age to the warning threshold. A l represents the age of the l-th patient and is used in the denominator to calculate the square of the sum of the ages of all patients. G kDenote the gender code of the k-th patient, N denote the normalization function, and RL denote the reinforcement learning model;

[0058] The dynamically adjusted warning threshold is in the range of [0, 1] and is divided into three levels. When it is between 0.1 and 0.5, it is the first level, indicating that the patient's emotional state is relatively stable, and the warning system can reduce the alarm frequency and reduce unnecessary interventions. When it is between 0.5 and 0.8, it is the second level, indicating that the patient's emotional state may start to fluctuate, and the warning system needs to be more vigilant and be prepared to intervene in a timely manner. When the value is greater than or equal to 0.8, it is the third level, indicating that the patient's emotional state has significant fluctuations, and the warning system should immediately notify the patient's guardian and attending physician to ensure that the patient can receive timely attention and support.

[0059] S5. Emotional trigger event recognition module. When the patient's emotional state exceeds the dynamically adjusted warning threshold, analyze the patient's behavioral patterns and physiological signals before and after the emotional change through a long short-term memory network, extract text data from the patient's social media interaction records, use natural language processing technology to identify emotional trigger content, and combine the results of the behavioral patterns, physiological signals, and text data to identify the events that cause the patient's emotional change.

[0060] When the patient's emotional state exceeds the dynamically adjusted warning threshold, analyze the patient's behavioral patterns and physiological signals before and after the emotional change through a long short-term memory network to obtain a comprehensive evaluation of the behavioral patterns and physiological signals. Its expression is:

[0061]

[0062] Among them, L(M) represents the comprehensive evaluation result of the behavioral patterns and physiological signals, W represents the length of the time series, o represents the number of samples of the behavioral patterns and physiological signals observed within this time window, f represents the time point, f y represents the occurrence time of the behavioral patterns and physiological signals, y represents the index of each sample point of the behavioral patterns and physiological signals within the time window, df represents the tiny variable in time, and τ represents the time scale parameter.

[0063] The value range of L(M) is [0, 1]. A value close to 0 indicates that the emotional change is small or there is no significant trigger event; a value close to 1 indicates that the emotional change is large and there may be significant emotional trigger events. The value of L(M) can be used to identify which specific events or behavioral patterns have triggered the patient's emotional change.

[0064] S5.1. Extract text data from the patient's social media interaction records and use natural language processing technology to identify emotional trigger content. Its expression is:

[0065]

[0066] Among them, S(Q) represents the score of the emotion-triggering content, Q represents the social media interaction text data, and Q g represents the emotion intensity of the g-th text segment, BERT represents the evaluation value of the emotion intensity of the text segment, e represents the number of text segments, and d g represents the timestamp of the g-th text segment, and d represents the reference time point.

[0067] The value range of S(Q) is between [0, 1], representing the intensity of the emotion-triggering content. When the value of S(Q) is close to 0, it indicates that the emotion-triggering content is weak or insignificant. When the value of S(Q) is close to 1, it indicates that the emotion-triggering content is strong and may be an important factor causing the patient's emotional changes.

[0068] Furthermore, through the above steps, we use natural language processing technology to extract text data from the patient's social media interaction records and identify emotion-triggering content. This method not only considers the emotion intensity of the text content but also takes into account the time factor to ensure that the identified content is the most relevant and up-to-date.

[0069] Integrate S(Q) into the overall emotion-triggering event recognition process and analyze it together with other information such as behavior patterns and physiological signals to determine the specific events that cause the patient's emotional changes.

[0070] S5.2 Combine the results of behavior patterns, physiological signals, and text data to identify the events that cause the patient's emotional changes. Its expression is:

[0071]

[0072] Among them, E represents the final comprehensive evaluation result, L(M) represents the comprehensive evaluation result of behavior patterns and physiological signals, and S(Q) represents the score of the emotion-triggering content.

[0073] The value range of E is [0, 1]. If E is close to 0, it indicates that the patient's emotional state has changed. This change is caused by a specific emotion-triggering event, or the influence of the emotion-triggering event is relatively large. This situation may be because the patient's emotional state is affected by other unmonitored factors or is a natural fluctuation of the emotional state. If E is close to 1, it means that the change in the patient's emotional state is caused by one or more significant emotion-triggering events. In this case, the change in the patient's emotional state can be clearly linked to the emotion-triggering events.

[0074] To further illustrate, when the emotional state exceeds the warning threshold T' but E is close to 0, in this case, the patient's emotional state has significantly deviated from the normal range, but there is no obvious single emotional trigger event as the direct cause of this emotional change. This situation may be because the patient's emotional state is affected by multiple factors, including but not limited to environmental factors, changes in internal emotional states, long-term accumulated stress, etc., rather than just a single emotional trigger event.

[0075] When the emotional state exceeds the warning threshold T' but E is close to 1, in this case, the patient's emotional state has significantly deviated from the normal range, and there is one or more significant emotional trigger events as the direct cause of this emotional change.

[0076] This situation indicates that the change in the patient's emotional state is mainly caused by specific emotional trigger events, which have a significant impact on the patient's emotional state.

[0077] S6. Personalized intervention suggestion generation module, which generates intervention suggestions based on the identified emotional trigger events and the patient's current emotional state.

[0078] Adjust the intervention suggestions according to the patient's personal preferences, historical records, and other relevant information to make them more personalized; considering factors such as the patient's age, gender, cultural background, etc., generate suggestions that are more in line with the patient's actual needs.

[0079] Based on the events that cause the patient's emotional change and the patient's current emotion, through natural language generation technology, provide the patient with easy-to-understand guiding suggestions. The intervention suggestions need to be clear and easy to understand so that the patient, guardians, and medical teams can understand and take actions.

[0080] The generated guiding suggestions will be sent to the patient's guardians, caregivers, and doctors simultaneously to ensure that all relevant parties can obtain information in a timely manner for assisting in the next treatment and intervention measures.

[0081] Example intervention suggestions

[0082] Suppose a patient's emotional state significantly deviates from the normal range and the trigger event is work stress, then the intervention suggestions may include

[0083] Suggestion to the patient: "We notice that you have been feeling very tense due to work recently. Try to arrange some relaxation time, such as taking a walk or meditating, which is very helpful for reducing stress."

[0084] Suggestion to the guardian: "Please pay attention to observing [patient's name]'s sleep pattern and eating habits. If any abnormalities are found, please contact us in a timely manner."

[0085] Advice to the doctor: "[Patient's name] has recently shown significant emotional changes due to work pressure. Please consider arranging a psychological consultation to assess whether further psychological intervention is needed."

[0086] In summary, the beneficial effects of the present invention are that the abnormal monitoring and early warning system for psychiatric patients aims to accurately evaluate the emotional state of patients by collecting and preprocessing the facial expressions, voices, and physiological signal data of patients in real time. The system uses multi-modal feature extraction and fusion technology, combined with deep learning algorithms, to ensure that the emotional changes of patients are comprehensively captured from different dimensions. To further enhance the practical value of the system, the system can also automatically identify the key events that cause emotional fluctuations and generate personalized intervention suggestions based on these events. This not only helps medical staff quickly respond to emergencies, but also provides customized treatment plans for patients, ultimately achieving the goal of improving mental health conditions. The application of the system significantly improves the efficiency and quality of mental health management and provides strong support for clinical practice.

[0087] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of the abnormal monitoring and early warning system for psychiatric patients are given.

[0088] In this example, three psychiatric patients were selected as research objects, labeled as Patient A, Patient B, and Patient C respectively. Each patient wore a wearable device for monitoring heart rate and skin conductance and was continuously monitored by a high-definition camera and microphone during a specific period. In addition, the social media interaction records of each patient were collected. During the experiment, the collected data was denoised, filtered, and smoothed.

[0089] Data collection and preprocessing:

[0090] Collect data at a fixed time every day to ensure data consistency.

[0091] Use a high-definition camera and microphone to capture the facial expressions and voice data of patients.

[0092] Collect the heart rate and skin conductance data of patients through wearable devices.

[0093] Feature extraction and fusion:

[0094] Extract emotional features from facial expression data, extract emotional features from voice data, and extract physiological signal features from heart rate and skin conductance data.

[0095] Use the attention mechanism for feature fusion to form a multi-modal feature vector.

[0096] Emotional state assessment and threshold initialization:

[0097] The initial threshold T is set according to the patient's personal basic information (age, gender, diagnosis result) and the preliminary emotion assessment result.

[0098] The emotion assessment result is obtained through an emotion recognition model.

[0099] Personalized warning threshold dynamic adjustment:

[0100] Use machine learning algorithms to dynamically adjust the warning threshold according to the patient's historical data (historical emotion state, historical physiological signals, historical behavior patterns, diagnosis results, treatment response).

[0101] The dynamically adjusted warning threshold is divided into three levels: level one (emotionally stable), level two (emotionally fluctuating), and level three (emotionally greatly fluctuating).

[0102] Emotion trigger event recognition:

[0103] When the patient's emotion state exceeds the warning threshold, use the long short-term memory network (LSTM) to analyze the behavior patterns and physiological signals before and after the patient's emotion change.

[0104] Extract text data from social media interaction records and use natural language processing technology to identify emotion trigger content.

[0105] Combine the results of behavior patterns, physiological signals, and text data to identify the events that cause the patient's emotion change.

[0106] Personalized intervention suggestion generation:

[0107] Generate easy-to-understand guiding suggestions according to the identified emotion trigger events and the patient's current emotion state.

[0108] Send the guiding suggestions to the patient's guardians, caregivers, and doctors.

[0109] Specifically as shown in Table 1:

[0110] Table 1 Operating result table of the abnormal detection and warning system for psychiatric patients

[0111]

[0112] Data analysis

[0113] Preliminary emotion assessment:

[0114] The preliminary emotion assessment of Patient A is 0.5, at a medium level, indicating that the patient's emotion is relatively stable.

[0115] The preliminary emotion assessment of Patient B is 0.4, slightly lower than the average level, indicating that the patient's emotion is relatively stable.

[0116] The preliminary emotional assessment of Patient C is 0.6, which is higher than the average level, indicating that the patient may have a potential risk of emotional fluctuations.

[0117] Dynamic threshold adjustment:

[0118] The dynamic threshold adjustment of Patient A is 0.6. According to the preliminary emotional assessment, the system sets the warning threshold higher, indicating that the system believes the emotional state of this patient is relatively stable.

[0119] The dynamic threshold adjustment of Patient B is 0.5. The low threshold may indicate that the system believes the emotions of this patient are more likely to fluctuate.

[0120] The dynamic threshold adjustment of Patient C is 0.7. The high threshold indicates that the system believes the emotional state of this patient is unstable and requires more attention.

[0121] Behavior pattern assessment:

[0122] The behavior pattern assessment of Patient A is 0.7, indicating that the change in the patient's behavior pattern before and after emotional changes is not significant.

[0123] The behavior pattern assessment of Patient B is 0.6, suggesting that the patient's behavior pattern has changed.

[0124] The behavior pattern assessment of Patient C is 0.8, indicating a significant change in the patient's behavior pattern.

[0125] Social media score:

[0126] The social media score of Patient A is 0.8, indicating that there is a relatively large amount of emotion-triggering content in social media interactions.

[0127] The social media score of Patient B is 0.7, and there is a certain amount of emotion-triggering content in social media interactions.

[0128] The social media score of Patient C is 0.9, indicating that there is a large amount of emotion-triggering content in social media interactions.

[0129] Comprehensive assessment result:

[0130] The comprehensive assessment result of Patient A is 0.75, which is close to the behavior pattern assessment value, indicating that social media interactions have a greater impact on the patient's emotional changes.

[0131] The comprehensive assessment result of Patient B is 0.65, which is between the preliminary emotional assessment and the behavior pattern assessment, indicating that both social media interactions and the behavior pattern have a certain impact on the patient's emotional changes.

[0132] The comprehensive evaluation result of Patient C is 0.85, which is close to the social media score, indicating that social media interaction has a very significant impact on their mood changes.

[0133] Guardian feedback:

[0134] The guardians of all three patients gave positive feedback (+), indicating that the guardians believe the intervention suggestions provided by the system are helpful and can effectively prevent the deterioration of the mood.

[0135] Conclusion

[0136] From the above experimental data, it can be seen that the abnormal monitoring and early warning system for psychiatric patients of the present invention can effectively identify the key factors leading to the mood changes of patients and give targeted intervention suggestions. Compared with traditional monitoring means, this system has the following innovations and advantages:

[0137] Personalized early warning threshold:

[0138] The system can dynamically adjust the early warning threshold according to the patient's personal basic information and historical data, improving the accuracy and personalization level of the early warning.

[0139] Comprehensive evaluation result:

[0140] By integrating behavior patterns, physiological signals and social media data, the system can more comprehensively evaluate the patient's mood state, thus more accurately identifying mood-triggering events.

[0141] Timely intervention suggestions:

[0142] The system can generate intervention suggestions in a timely manner according to the identified mood-triggering events and the patient's current mood state, which helps to prevent the deterioration of the mood at an early stage.

[0143] In summary, compared with the prior art, the abnormal monitoring and early warning system for psychiatric patients of the present invention has obvious innovations and advantages in terms of personalization, accuracy, timeliness and effectiveness, and can significantly improve the treatment effect and quality of life of psychiatric patients.

[0144] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A psychiatric patient abnormal monitoring and early warning system, characterized in that: include, Data collection and preprocessing module, collects patient information in real time and performs preprocessing; The feature extraction and fusion module extracts features from the preprocessed patient information and then fuses them to form multimodal features; The emotional state assessment and threshold initialization module obtains preliminary emotional assessment results based on multimodal features and emotion recognition models, and sets the initial warning threshold according to the patient's basic personal information and preliminary emotional assessment results; A personalized warning threshold dynamic adjustment module uses a machine learning algorithm to dynamically adjust the warning threshold based on the patient's historical data and preliminary emotional assessment results; The emotion trigger event recognition module analyzes the patient's behavior patterns and physiological signals before and after the emotional change through the long short-term memory network when the patient's emotional state exceeds the dynamically adjusted warning threshold, extracts text data from the patient's social media interaction records, uses natural language processing technology to identify emotional trigger content, and combines the results of the behavior patterns, physiological signals and text data to identify the events that caused the patient's emotional change; The personalized intervention suggestion generation module generates intervention suggestions based on the identified emotion triggering events and the patient's current emotional state.

2. The abnormal monitoring and early warning system for psychiatric patients as claimed in claim 1, characterized in that: The patient information is collected by capturing the patient's facial expressions and voice in real time through a high-definition camera and microphone, collecting the patient's heart rate and skin conductance physiological signals through wearable devices, and performing noise reduction, filtering and smoothing on the collected video and voice data.

3. The abnormal monitoring and early warning system for psychiatric patients as claimed in claim 2, characterized in that: Feature extraction is performed based on the preprocessed patient information. Emotional features are extracted from facial expression information, emotional features are extracted from voice data, and features of heart rate variability and skin conductance change trends are extracted from physiological signals. The features are fused to form multimodal features.

4. The abnormal monitoring and early warning system for psychiatric patients as claimed in claim 3, characterized in that: The patient's basic personal information includes age, gender, and diagnosis results. The initial warning threshold T is set based on the patient's basic personal information and preliminary emotion assessment results.

5. The abnormal monitoring and early warning system for psychiatric patients as claimed in claim 4, characterized in that: The patient's historical data includes historical emotional states, historical physiological signals, historical behavioral patterns, diagnosis results, and treatment responses; Use machine learning algorithms to dynamically adjust warning thresholds based on patients’ historical data and preliminary emotional assessment results; The dynamically adjusted warning threshold is divided into three levels: level one for stable emotions, level two for fluctuating emotions, and level three for large fluctuations in emotions.

6. The abnormal monitoring and early warning system for psychiatric patients as claimed in claim 5, characterized in that: When the patient's emotional state exceeds the dynamically adjusted warning threshold, the patient's behavioral patterns and physiological signals before and after the emotional changes are analyzed through the long and short-term memory network to obtain a comprehensive evaluation of the behavioral patterns and physiological signals.

7. The abnormal monitoring and early warning system for psychiatric patients as claimed in claim 6, characterized in that: Extract text data from patients’ social media interaction records and use natural language processing technology to identify emotional triggers.

8. The abnormal monitoring and early warning system for psychiatric patients as claimed in claim 7, characterized in that: Combine results from behavioral patterns, physiological signals, and text data to identify events that lead to changes in patients’ emotions.

9. The abnormal monitoring and early warning system for psychiatric patients as claimed in claim 8, characterized in that: Provide patients with easy-to-understand guidance suggestions based on the events that caused the patient's emotional changes and the patient's current emotions through natural language generation technology; The generated guidance recommendations will be sent simultaneously to the patient's guardian, caregiver and doctor to assist in the next step of treatment and intervention.

Images