A method and system for evaluating the effect of a psychological disease treatment using data analysis

Abstract

The application relates to the field of medical information technology and discloses a psychological disease treatment effect evaluation method and system adopting data analysis. The method comprises the following steps: classifying and standardizing physiological parameters based on a model; obtaining a dimension reduction vector of a sleep quality parameter through kernel principal component analysis, and then obtaining a cross-modal correlation feature through nonlinear mapping of the sleep quality parameter and a voice emotion parameter; calculating a correlation coefficient, adjusting a kernel function parameter in combination with an emotional stress parameter if the correlation coefficient exceeds a threshold value, and then reextracting the feature; obtaining a dynamic weight coefficient through the cross-modal correlation feature and a weight mechanism, and weighting a treatment effect index; and identifying an index abnormal mode through abnormal detection to generate a personalized evaluation report. Through multi-dimensional physiological data analysis and dynamic adjustment, the application solves the problems of lacking objective multi-dimensional continuous monitoring and dynamic personalized analysis in evaluation, and improves the evaluation accuracy and the personalized level.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical information, in particular to a psychological disease treatment effect evaluation method and system using data analysis. BACKGROUND

[0002] Psychological disease treatment effect evaluation is an important research direction in the field of digital medicine. The existing technology mainly uses psychological scale evaluation, single physiological index monitoring and electronic medical record system recording methods to judge the treatment effect. The traditional evaluation method uses subjective scales such as self-report depression scale and anxiety scale, combined with discrete collection of basic physiological parameters such as heart rate and blood pressure, to perform static evaluation at specific time nodes during the treatment process, and the evaluation results are stored in the electronic medical record system for reference by doctors.

[0003] However, the existing technology has significant deficiencies: first, the evaluation is highly subjective, and self-reporting based on scales is easily influenced by patient subjective consciousness and social expectation bias, which cannot objectively reflect the real psychological state changes; second, the monitoring dimension is single, relying only on basic physiological indicators such as heart rate and blood pressure, lacking comprehensive analysis of multi-dimensional physiological behavior data such as sleep quality, emotional stress and voice characteristics; third, the evaluation is not timely, and the existing system can only perform static evaluation at discrete time points, and cannot realize continuous dynamic monitoring and real-time feedback during the treatment process.

[0004] Based on the above deficiencies of the existing technology, it can be found that relying solely on subjective scale evaluation leads to a lack of objective individual baseline reference standard, and thus personalized treatment effect evaluation benchmarks cannot be established; due to the lack of coordinated collection and cross-modal correlation analysis of multi-dimensional physiological data, the existing technology cannot discover the internal relationship and mutual influence mechanism between different physiological systems; and the static discrete evaluation mode makes it impossible for the system to dynamically adjust the weight and optimize the treatment plan according to the real-time data changes during the treatment process, ultimately leading to the technical problems of low treatment effect evaluation accuracy and insufficient personalization. SUMMARY

[0005] The present application provides a psychological disease treatment effect evaluation method and system using data analysis, which solves the technical problem of lack of objective multi-dimensional continuous monitoring and dynamic personalized analysis in psychological disease treatment effect evaluation, and improves the accuracy and personalization level of treatment effect evaluation.

[0006] In a first aspect, the present application provides a psychological disease treatment effect evaluation method using data analysis, which comprises:

[0007] In step S101, multi-dimensional physiological data of a patient is acquired, the multi-dimensional physiological data is standardized to obtain standardized physiological parameters, statistical characteristics and circadian rhythm patterns of each physiological parameter are calculated, and an individualized physiological baseline model is obtained according to individual differences of the patient;

[0008] In step S102, the standardized physiological parameters are divided into sleep quality parameters, emotional stress parameters and voice emotion parameters based on the individualized physiological baseline model; the sleep quality parameters are input into a kernel principal component analysis algorithm to obtain a reduced dimension feature vector, and the reduced dimension feature vector and the voice emotion parameters are input into a nonlinear mapping model to obtain cross-modal correlation features; a correlation coefficient of the cross-modal correlation features is calculated, and if the correlation coefficient exceeds a threshold value, the kernel function parameters of the kernel principal component analysis algorithm are adjusted according to the correlation coefficient and the emotional stress parameters, and the cross-modal correlation features are re-extracted.

[0009] In step S103, a dynamic weight coefficient is obtained according to the cross-modal correlation features and a weight distribution mechanism, a treatment effect index is calculated by weighting the standardized physiological parameters using the dynamic weight coefficient, an abnormal pattern of the treatment effect index is identified by an anomaly detection algorithm, and a personalized treatment effect evaluation report is generated.

[0010] Optionally, the standardized physiological parameters obtained by standardizing the multi-dimensional physiological data include:

[0011] The mean and standard deviation of each physiological parameter during the baseline period are calculated, and the real-time collected physiological data is processed using a standardization formula, wherein the standardized value is equal to the original value minus the mean and then divided by the standard deviation;

[0012] Statistical characteristics of sleep breathing frequency, skin electrical response value and voice fundamental frequency are calculated respectively, and the statistical characteristics include mean, variance, skewness and kurtosis;

[0013] If the absolute value of the skewness of a certain physiological parameter is greater than a preset skewness threshold, power transformation is used for data correction.

[0014] Optionally, the individualized physiological baseline model is obtained according to individual differences of the patient, including:

[0015] The baseline data of the past fourteen days is taken as an individual difference analysis window, circadian rhythm parameters in the analysis window are obtained, and the circadian rhythm parameters include sleep latency, proportion of rapid eye movement sleep, proportion of deep sleep, number of night awakenings, and emotional stress peak occurrence time;

[0016] An individualized physiological baseline model is obtained by fitting each circadian rhythm parameter using a sine function, and the model expression is baseline value equal to amplitude coefficient multiplied by sine function plus direct current component, wherein the period of the sine function is twenty-four hours.

[0017] Optionally, the standardized physiological parameters based on the individualized physiological baseline model are divided into sleep quality parameters, emotional stress parameters, and voice emotion parameters, including:

[0018] Acquire sleep-related data from standardized physiological parameters that deviate from the baseline model by more than the first deviation threshold. If multiple parameters meet the conditions, the parameter with the largest deviation is taken as the main component of the sleep quality parameter.

[0019] Obtain a second deviation threshold. If the current time period is nighttime, increase the second deviation threshold. Among the remaining parameters, those that deviate from the baseline by more than the second deviation threshold and whose change frequency is within a preset frequency range are used as emotional stress parameters and voice emotion parameters.

[0020] Optionally, adjusting the kernel function parameters of the kernel principal component analysis algorithm and re-extracting cross-modal association features based on the correlation coefficient and emotional stress parameters includes:

[0021] The adjustment factor is obtained by weighting the ratio of the correlation coefficient to the baseline correlation coefficient and the ratio of the emotional stress parameter to the baseline stress parameter.

[0022] The kernel function parameter value of the kernel principal component analysis algorithm is obtained by using the formula that the new kernel function parameter is equal to the baseline kernel function parameter multiplied by the adjustment factor plus one. If the new kernel function parameter is greater than the maximum kernel function parameter value, then the maximum kernel function parameter value is taken.

[0023] The kernel principal component analysis algorithm is updated using the adjusted kernel function parameters, and the sleep quality parameters are then re-dimension-reduced using the updated algorithm.

[0024] Optionally, obtaining the dynamic weight coefficients based on the cross-modal association features and weight allocation mechanism includes:

[0025] Calculate the contribution of each cross-modal association feature to the prediction of treatment effect, and use the ratio of the absolute value of the contribution of each feature to the sum of the absolute values ​​of the contribution of all features as the dynamic weight coefficient of that feature.

[0026] Optionally, the step of identifying abnormal patterns in the treatment efficacy index and generating a personalized treatment efficacy evaluation report through anomaly detection algorithms includes:

[0027] The termination threshold for obtaining the treatment effectiveness index;

[0028] Starting from the first monitoring point in the treatment effect index sequence, the time node at which the first treatment effect index value is less than or equal to the termination threshold is obtained;

[0029] The treatment days or treatment cycles experienced from a treatment starting time point to the time node are calculated, and a personalized treatment effect evaluation report containing treatment suggestions is generated in combination with the abnormal pattern analysis result.

[0030] In a second aspect, the present application provides a psychological disease treatment effect evaluation system using data analysis, comprising:

[0031] A baseline model construction module is configured to obtain multi-dimensional physiological data of a patient, perform standardization processing on the multi-dimensional physiological data to obtain standardized physiological parameters, calculate statistical features and circadian rhythm patterns of each standardized physiological parameter, and generate an individualized physiological baseline model in combination with individual differences of the patient.

[0032] A feature processing module is configured to divide the standardized physiological parameters into sleep quality parameters, emotional stress parameters, and voice emotion parameters according to the individualized physiological baseline model, input the sleep quality parameters into a kernel principal component analysis algorithm to obtain a reduced dimension feature vector, input the reduced dimension feature vector and the voice emotion parameters into a nonlinear mapping model to obtain cross-modal correlation features, calculate a correlation coefficient of the features, and if the correlation coefficient exceeds a threshold, adjust kernel function parameters in combination with the correlation coefficient and the emotional stress parameters to re-extract the cross-modal correlation features.

[0033] An evaluation generation module is configured to determine a dynamic weight coefficient according to the cross-modal correlation features and a weight distribution mechanism, perform weighted calculation on the standardized physiological parameters by using the dynamic weight coefficient to obtain a treatment effect index, and generate a personalized treatment effect evaluation report by identifying an abnormal pattern of the treatment effect index through an abnormality detection algorithm.

[0034] In a third aspect, a psychological disease treatment effect evaluation device using data analysis is provided, comprising a memory and at least one processor, wherein the memory stores instructions; and the at least one processor invokes the instructions in the memory to enable the psychological disease treatment effect evaluation device using data analysis to perform the psychological disease treatment effect evaluation method using data analysis described above.

[0035] In a fourth aspect, a computer readable storage medium is provided, wherein the computer readable storage medium stores instructions, and when the instructions are run on a computer, the computer is enabled to perform the psychological disease treatment effect evaluation method using data analysis described above.

[0036] The present application provides a psychological disease treatment effect evaluation method and system using data analysis, which is suitable for the evaluation of psychological disease treatment effect in the field of medical information technology, and can solve the technical problem of lack of objective multi-dimensional continuous monitoring and dynamic personalized analysis in psychological disease treatment effect evaluation. Compared with the prior art, the beneficial effects of the technical solution of the present application are at least as follows:

[0037] First, by acquiring multi-dimensional physiological data of patients and establishing individualized physiological baseline models, an objective evaluation benchmark is provided, reducing dependence on subjective scales and enhancing the authenticity and reliability of evaluation results;

[0038] Second, based on kernel principal component analysis and nonlinear mapping technology, cross-modal correlation features are extracted, realizing the fusion analysis of multi-dimensional physiological data such as sleep quality, emotional stress and speech emotion, and improving the richness and relevance of feature expression;

[0039] Third, by dynamically adjusting algorithm parameters and weight distribution mechanism, real-time response and adaptive optimization of data changes during treatment are realized, enhancing the dynamic monitoring capability and individualized evaluation level of the system;

[0040] Fourth, abnormal patterns of treatment effect index are identified using anomaly detection algorithms, and individualized evaluation reports containing treatment recommendations are generated, improving the timeliness of treatment effect evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can be obtained based on these drawings without creative labor.

[0042] Figure 1 It is a flowchart of a psychological disease treatment effect evaluation method using data analysis in the present application;

[0043] Figure 2 It is a schematic diagram of the processing process of a psychological disease treatment effect evaluation method using data analysis in the present application;

[0044] Figure 3 It is a structural schematic diagram of a psychological disease treatment effect evaluation system using data analysis in the present application;

[0045] Figure 4 It is a structural schematic diagram of a psychological disease treatment effect evaluation device using data analysis in the present application. DETAILED DESCRIPTION

[0046] The embodiment of the present application provides a psychological disease treatment effect evaluation method and system using data analysis. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0047] For ease of understanding, the specific process of the embodiment of the present application is described below. Please refer to Figure 1 One embodiment of the psychological disease treatment effect evaluation method using data analysis in the embodiment of the present application comprises the following steps.

[0048] Step S101, acquiring multi-dimensional physiological data of a patient, performing standardization processing on the multi-dimensional physiological data to obtain standardized physiological parameters, calculating statistical characteristics and circadian rhythm patterns of each physiological parameter, and obtaining an individualized physiological baseline model according to individual differences of the patient;

[0049] Step S102, dividing the standardized physiological parameters into sleep quality parameters, emotional stress parameters and voice emotion parameters based on the individualized physiological baseline model; inputting the sleep quality parameters into a kernel principal component analysis algorithm to obtain a dimension reduction feature vector, and inputting the dimension reduction feature vector and the voice emotion parameters into a nonlinear mapping model to obtain cross-modal correlation features; calculating a correlation coefficient of the cross-modal correlation features, if the correlation coefficient exceeds a threshold value, adjusting kernel function parameters of the kernel principal component analysis algorithm according to the correlation coefficient and the emotional stress parameters and re-extracting cross-modal correlation features;

[0050] Step S103, obtaining a dynamic weight coefficient according to the cross-modal correlation features and a weight distribution mechanism, performing weighted calculation on the standardized physiological parameters by using the dynamic weight coefficient to obtain a treatment effect index, identifying an abnormal mode of the treatment effect index by using an anomaly detection algorithm and generating a personalized treatment effect evaluation report.

[0051] Specifically, multi-dimensional physiological data of the patient is collected, covering sleep breathing frequency, skin conductance value, voice fundamental frequency and other indicators related to psychological state, and is recorded continuously for a certain period to form baseline data. The data is standardized, and the mean and standard deviation of each type of physiological parameter during the baseline period are calculated. The standardized value of a certain real-time physiological data is equal to the original value minus the baseline mean of the corresponding parameter and then divided by the baseline standard deviation. The statistical characteristics of each parameter, including mean, variance, skewness and kurtosis, are also calculated. If the absolute value of skewness exceeds a preset threshold, power transformation is used to correct the data distribution. Day-night rhythm parameters such as sleep latency and rapid eye movement sleep proportion are extracted, and a sinusoidal function with a period of one day is used to fit these parameters to obtain an individualized physiological baseline model, providing an individual reference for subsequent evaluation and solving the problem of lack of objective individual reference in the prior art.

[0052] Based on the individualized physiological baseline model, the standardized physiological parameters are classified, a first deviation threshold is set, and sleep-related data that deviates from the baseline by more than the threshold is selected. If multiple data meet the criteria, the one with the largest deviation is selected as the main component of the sleep quality parameter. A second deviation threshold is set, and if it is a night period (according to the day-night rhythm in the baseline model), the second deviation threshold is increased. From the remaining parameters, select the parameter that deviates from the adjusted threshold and has a change frequency within a preset range - skin conductance value is classified as emotional stress parameter, voice fundamental frequency is classified as voice emotion parameter, and multi-dimensional classification solves the problem of single monitoring dimension.

[0053] The sleep quality parameter is input into the kernel principal component analysis algorithm, which is mapped to a low-dimensional space through a kernel function, and a reduced feature vector is obtained after removing redundant information. The vector and the voice emotion parameter are input into a nonlinear mapping model, and the nonlinear correlation between the two types of data is captured through an activation function to generate cross-modal correlation features, realize multi-modal data collaborative analysis, and solve the problem of inability to discover the internal relationship of the physiological system.

[0054] The correlation coefficient of the cross-modal correlation feature is calculated, and if it exceeds a preset threshold, the ratio of the coefficient to the baseline correlation coefficient and the ratio of the emotional stress parameter to the baseline stress parameter are calculated, and the adjustment factor is obtained by weighted summation according to the weight. According to the formula "new kernel function parameter = baseline kernel function parameter x (adjustment factor + 1)", the parameter value is calculated, and if it does not exceed the maximum value, it is directly used. The adjusted parameter is used to update the kernel principal component analysis algorithm, and the sleep quality parameter is reduced again to generate new cross-modal correlation features, dynamically adjusting to solve the problem of static evaluation not adapting to real-time data changes.

[0055] The contribution value of each cross-modal correlation feature to the treatment effect prediction is calculated, and the feature is taken as the independent variable and the treatment effect quantitative index is taken as the dependent variable through linear regression. The regression coefficient is the contribution value. The dynamic weight coefficient is obtained by taking the absolute value of the contribution value and calculating the proportion. The dynamic weight is used to weight the corresponding standardized physiological parameters, and the weighted values of sleep quality, emotional stress, and voice emotion parameters are summed to obtain the treatment effect index. The dynamic weight solves the problem of lack of individualization of fixed weight, and the quantitative index improves objectivity.

[0056] The termination threshold of the treatment effect index is obtained, which is determined in combination with the clinical standard and the baseline model fluctuation range. From the day of treatment, the daily index is compared with the threshold in sequence, and the time node of the first compliance is recorded. The time interval from the start of treatment to the node is calculated, and the index sequence is analyzed by an anomaly detection algorithm. If the index fluctuation in a certain period exceeds the normal range, it is determined as an abnormal mode. In combination with the above information, a report is generated to determine the treatment compliance time, abnormal mode and corresponding suggestions, solving the problem of poor timeliness of evaluation and providing precise guidance for clinical treatment.

[0057] It can be understood that the execution subject of the present application can be a psychological disease treatment effect evaluation system using data analysis, and can also be a terminal or a server, which is not limited here. The server is taken as an example for illustration in the embodiments of the present application.

[0058] In a specific embodiment, the process of standardizing the multi-dimensional physiological data to obtain standardized physiological parameters can specifically include the following steps:

[0059] The mean and standard deviation of each physiological parameter during the baseline period are calculated, and the real-time collected physiological data is processed using a standardization formula, wherein the standardized value is equal to the original value minus the mean and then divided by the standard deviation;

[0060] The statistical features of sleep breathing frequency, skin electrical response value and voice fundamental frequency are calculated respectively, and the statistical features include mean, variance, skewness and kurtosis;

[0061] If the absolute value of the skewness of a certain physiological parameter is greater than a preset skewness threshold, power transformation is used for data correction.

[0062] Specifically, in the process of standardizing the multi-dimensional physiological data to obtain standardized physiological parameters, the baseline period data of each physiological parameter needs to be determined first. In the context of psychological disease treatment effect evaluation, the physiological data of the past fourteen consecutive days is usually used as the baseline period. For each type of physiological parameter (such as sleep breathing frequency, skin electrical response value, and voice fundamental frequency), the mean and standard deviation of each parameter in the baseline period are calculated. This step establishes an individual reference benchmark for each parameter, avoiding the problem of ignoring individual differences by using a uniform standard in the prior art. The lack of objective individual baseline is one of the core pain points of the prior art. Subsequently, each real-time physiological data is standardized by using the corresponding baseline mean and standard deviation. That is, the standardized value of a real-time physiological data is equal to the original value of the data minus the baseline mean of the parameter to which the data belongs, and then divided by the baseline standard deviation of the parameter. For example, the original value of the real-time collected sleep breathing frequency needs to be subtracted by the baseline mean of the sleep breathing frequency, and then divided by the baseline standard deviation of the sleep breathing frequency to obtain its standardized value. Similarly, the real-time collected skin electrical response value and voice fundamental frequency are processed. This processing can eliminate the influence of different physiological parameters due to the difference in dimensions (such as sleep breathing frequency unit: times / minute, skin electrical response value unit: microsiemens), so that different dimensional parameters are in the same numerical order of magnitude, laying a foundation for subsequent multi-dimensional data comprehensive analysis, and solving the problem of single monitoring dimension and inability to realize multi-physiological system data collaborative analysis in the prior art.

[0063] After completing the standardization process, the statistical characteristics of sleep breathing frequency, skin electrical response value, and voice fundamental frequency, which are closely related to psychological state, are calculated. These statistical characteristics include mean, variance, skewness, and kurtosis. The statistical characteristic mean of a parameter is the average level of all standardized data of the parameter, reflecting the overall trend of the parameter in the monitoring period. The statistical characteristic variance is the average of the square of the deviation of the standardized data of the parameter from the statistical characteristic mean, reflecting the stability of the parameter fluctuations. For example, the variance of sleep breathing frequency can reflect the stability of the patient's breathing during sleep, directly related to sleep quality evaluation. The statistical characteristic skewness describes the degree of asymmetry of the distribution of the standardized data, and the statistical characteristic kurtosis reflects the steepness of the data distribution. Through the extraction of these four types of statistical characteristics, the physiological parameter characteristics can be described from multiple angles such as overall trend, fluctuation, and distribution form, breaking through the limitations of relying on single physiological indicator value evaluation in the prior art, enriching the representation dimension of physiological data, and providing more comprehensive information for subsequent capture of the correlation between physiological data and psychological state.

[0064] After the skewness of each parameter is calculated, the absolute value thereof is compared with a preset skewness threshold. If the absolute value of the skewness of a physiological parameter is greater than the preset skewness threshold, it indicates that the standardized data distribution of the parameter has obvious asymmetry. Such non-normal distribution data will affect the processing effect of the subsequent kernel principal component analysis algorithm (used for sleep quality parameter dimension reduction). Since the kernel principal component analysis can more accurately extract key features when the data is approximately normally distributed, power transformation is needed to correct the standardized data of the parameter at this time. The data distribution form is adjusted through power transformation, so that the corrected data is closer to the normal distribution, ensuring the accuracy and reliability of the subsequent algorithm processing, avoiding feature extraction deviation caused by abnormal data distribution, solving the problem of low evaluation accuracy caused by non-standard data processing in the prior art, and providing data quality guarantee for the effective extraction of subsequent cross-modal correlation features.

[0065] In a specific embodiment, the process of obtaining an individualized physiological baseline model according to individual differences of a patient can specifically include the following steps:

[0066] The baseline data of the next fourteen days are taken as an individual difference analysis window, and circadian rhythm parameters in the analysis window are obtained. The circadian rhythm parameters include sleep latency, rapid eye movement sleep proportion, deep sleep proportion, number of night awakenings, and emotional stress peak occurrence time.

[0067] An individualized physiological baseline model is obtained by fitting each circadian rhythm parameter with a sine function. The model expression is that the baseline value is equal to the amplitude coefficient multiplied by the sine function plus the direct current component, where the period of the sine function is twenty-four hours.

[0068] Specifically, the baseline data of the next fourteen days are taken as an individual difference analysis window. In this window, the circadian rhythm parameters are obtained for the core indicators related to physiological rhythm in the evaluation of the treatment effect of psychological diseases. These parameters specifically include sleep latency, rapid eye movement sleep proportion, deep sleep proportion, number of night awakenings, and emotional stress peak occurrence time. Each parameter is extracted from fourteen days of continuous monitoring data. For example, sleep latency data needs to record the time from preparation for sleep to sleep state of the patient in fourteen days, rapid eye movement sleep proportion needs to count the proportion of rapid eye movement sleep stage time in total sleep time in each day, and emotional stress peak occurrence time needs to capture the specific time of the peak of emotional stress response of the patient in fourteen days. The collection of all parameters is in the form of continuous time series, rather than discrete time point data. This processing method solves the problems of poor timeliness and inability to realize continuous dynamic monitoring of the treatment process in the prior art. Through fourteen days of continuous data collection, a data source covering a complete physiological cycle without data breakpoints is provided for the subsequent establishment of individualized benchmarks, avoiding evaluation deviation caused by discrete data.

[0069] After obtaining fourteen days of continuous data of each circadian rhythm parameter, for each parameter, a sinusoidal function is used for fitting to construct an individualized physiological baseline model. The period of the sinusoidal function is set to twenty-four hours, fully matching the circadian characteristics of human physiological activities (such as sleep-wake cycle, emotional fluctuation rule). The model expression is specifically the baseline value equal to the amplitude coefficient multiplied by the sinusoidal function plus the direct current component. The amplitude coefficient is used to quantify the fluctuation amplitude of the parameter within the circadian cycle. For example, the amplitude coefficient of deep sleep proportion can reflect the difference between the individual's night deep sleep proportion and the daytime baseline level. The direct current component represents the basic stable level of the parameter within the circadian cycle. For example, the direct current component related to the voice emotion parameter can reflect the voice fundamental frequency baseline value of the individual in the daily state. In the specific fitting process, for the sleep latency parameter, the sleep latency data recorded at different time points in the fourteen days is substituted into the sinusoidal function one by one. The amplitude coefficient and the direct current component corresponding to the parameter are determined by calculation, and then the sleep latency individualized baseline model applicable only to the patient is formed. Similarly, the rapid eye movement sleep proportion, deep sleep proportion, number of night awakenings, and emotional stress peak occurrence time each need to use fourteen days of continuous data to perform sinusoidal function fitting independently. Each circadian rhythm parameter and the baseline model obtained by fitting form a one-to-one correspondence, ensuring that each model can accurately reflect the unique rhythm characteristics of the individual in a specific physiological indicator.

[0070] This processing logic based on fourteen days of continuous data and sinusoidal function fitting directly addresses the problem that existing technologies lack objective individual baseline reference standards and cannot establish personalized treatment effect evaluation benchmarks. Existing technologies rely on unified physiological indicator standards or discrete subjective reports, ignoring individual differences in physiological rhythms. In this scheme, fourteen days of continuous data collection ensures complete coverage of individual physiological rhythms, avoiding the problem that short-term data cannot reflect the true rhythm. The continuous data is converted into a quantifiable individualized baseline model through sinusoidal function fitting, so that each physiological parameter has a reference benchmark based on the individual itself, rather than a unified external standard. When classifying standardized physiological parameters based on the model, parameters that deviate from the individual's own baseline can be accurately identified. For example, when judging sleep quality parameters, the patient's exclusive sleep latency baseline model can be used instead of a universal threshold, effectively reducing misjudgments caused by individual differences. At the same time, the selected five circadian rhythm parameters cover sleep quality (sleep latency, rapid eye movement sleep proportion, deep sleep proportion, number of night awakenings) and emotional state (emotional stress peak occurrence time), two dimensions closely related to the treatment effect of psychological diseases, solving the problem of single monitoring dimension in existing technologies, which only relies on basic indicators such as heart rate and blood pressure. This provides multi-dimensional and individualized basic data support for subsequent cross-modal correlation feature extraction.

[0071] In a specific embodiment, the process of classifying the standardized physiological parameters into sleep quality parameters, emotional stress parameters and voice emotion parameters based on the individualized physiological baseline model can specifically include the following steps:

[0072] Obtaining sleep-related data in the standardized physiological parameters that deviates from the baseline model by more than a first deviation threshold, and if there are multiple parameters that meet the condition, the parameter with the largest deviation is selected as the main component of the sleep quality parameter.

[0073] Obtaining a second deviation threshold, if the current period is a night period, the second deviation threshold is increased, and the parameter in the remaining parameters that deviates from the baseline by more than the second deviation threshold and has a change frequency within a preset frequency range is selected as the emotional stress parameter and the voice emotion parameter.

[0074] Specifically, when classifying the standardized physiological parameters based on the individualized physiological baseline model, sleep-related data needs to be selected from all standardized physiological parameters. These sleep-related data specifically include standardized sleep respiratory frequency, sleep latency, rapid eye movement sleep proportion, deep sleep proportion, number of night awakenings, etc., all of which are directly related to sleep quality monitoring in psychological disease treatment effect evaluation. By calculating the deviation of each sleep-related data from the corresponding parameter in the individualized physiological baseline model, it is determined whether the deviation exceeds a preset first deviation threshold. For example, if the difference between the standardized sleep latency of a patient and the sleep latency in the individualized physiological baseline model exceeds the first deviation threshold, the sleep latency data is included in the selected range. If the deviation of multiple sleep-related data exceeds the first deviation threshold, the deviation values of each data need to be further calculated, and the data with the largest deviation is selected as the main component of the sleep quality parameter. For example, when the standardized deep sleep proportion deviates from the baseline by 40% and the number of night awakenings deviates from the baseline by 25%, if both exceed the first threshold, the deep sleep proportion is selected as the main component of the sleep quality parameter. This processing process addresses the problem of single monitoring dimension in the prior art, which cannot focus on core physiological indicators. By directionally selecting sleep-related data and locking the core parameter with the largest deviation, sleep quality evaluation can accurately anchor the sleep indicators that have the most significant impact on psychological state, avoiding evaluation bias caused by parameter clutter, and providing clear core data support for subsequent sleep quality parameter dimension reduction.

[0075] After the sleep quality parameter screening is completed, the remaining standardized physiological parameters (including standardized electrodermal response values, voice fundamental frequency, and emotion stress peak time related data) need to be classified. First, a preset second deviation threshold is obtained, and then it is determined whether the current period is a night period (usually combined with the 24-hour period of the circadian rhythm parameter in the individualized physiological baseline model, such as 22:00 to 6:00 the next day as the night period). If it is a night period, the second deviation threshold is increased, for example, the daytime second deviation threshold of 15% is adjusted to 25% at night. This is because the human body is relatively stable in physiological activity at night, and slight fluctuations are mostly normal rhythm performance. Increasing the threshold value can avoid misjudging normal fluctuations as abnormal data, which meets the needs of adjusting the judgment standard in combination with the circadian rhythm in psychological disease assessment, and solves the problem that the prior art lacks dynamic threshold adjustment and cannot adapt to individual physiological rhythms. Then, the deviation of the remaining standardized physiological parameters from the corresponding parameters in the individualized physiological baseline model is calculated, and parameters with a deviation exceeding the adjusted second deviation threshold are screened out. At the same time, it is necessary to verify whether the frequency of change of these parameters is within the preset frequency range. The preset frequency range is set according to the common rules of emotion and voice response in psychological disease assessment, for example, the change frequency of emotion stress parameters (such as electrodermal response values) is usually 1-5 minutes / time, and the change frequency of voice emotion parameters (such as voice fundamental frequency) is usually 10 seconds-2 minutes / time. If a parameter meets both the deviation threshold and the corresponding frequency range, it is classified into the corresponding category. Among them, electrodermal response values, emotion stress peak time related data are usually classified as emotion stress parameters, and voice fundamental frequency data are classified as voice emotion parameters. This classification logic combines period adjustment threshold and frequency screening to ensure that the screening of emotion stress parameters and voice emotion parameters meets the characteristics of individual physiological rhythms and accurately matches the dynamic change characteristics of the two types of parameters, solving the problem that the prior art cannot distinguish the dynamic characteristics of different physiological parameters, leading to classification confusion. At the same time, a parameter system of three dimensions of sleep quality, emotion stress, and voice emotion is formed, providing multi-dimensional and highly correlated basic data for subsequent cross-modal correlation feature extraction, breaking through the limitations of relying on a single physiological indicator in the prior art.

[0076] In a specific embodiment, the process of adjusting the kernel function parameter of the kernel principal component analysis algorithm according to the correlation coefficient and the emotion stress parameter and re-extracting the cross-modal correlation feature can specifically include the following steps:

[0077] Calculate the weighted sum of the ratio of the correlation coefficient to the reference correlation coefficient and the ratio of the emotion stress parameter to the reference stress parameter to obtain an adjustment factor;

[0078] According to the formula that the new kernel function parameter is equal to the reference kernel function parameter multiplied by the adjustment factor plus one, the kernel function parameter value of the kernel principal component analysis algorithm is obtained. If the new kernel function parameter is greater than the maximum kernel function parameter value, the maximum kernel function parameter value is taken.

[0079] The kernel principal component analysis algorithm is updated using the adjusted kernel function parameters, and the updated algorithm is used to re-process the sleep quality parameters for dimension reduction.

[0080] Specifically, when calculating the adjustment factor, the specific values of the correlation coefficient and the reference correlation coefficient need to be determined first. The correlation coefficient is derived from the previous calculation results of the cross-modal correlation features, which are generated by nonlinear mapping of the dimension-reduced feature vectors of the sleep quality parameters and the speech emotion parameters. The reference correlation coefficient is a reference value preset for the psychological disease treatment effect evaluation scene, which is usually statistically obtained based on a large number of stable state patient cross-modal correlation features, for example, when the patient's mood does not fluctuate significantly and the sleep state is stable, the correlation coefficient of the cross-modal correlation features is often set to 0.6 as the reference. The emotional stress parameter is a parameter classified based on the individualized physiological baseline model, such as the standardized skin conductance response value or the emotional stress peak value occurrence time deviation value. The reference stress parameter corresponds to the reference level of the emotional stress parameter in the individualized physiological baseline model, for example, the reference value of the skin conductance response value in the baseline model of a certain patient is 5 microsiemens, which is generated by fitting the baseline data for fourteen consecutive days. Then, two ratios are calculated, i.e. the ratio of the correlation coefficient to the reference correlation coefficient and the ratio of the emotional stress parameter to the reference stress parameter. In the psychological disease evaluation scene, the emotional stress state has a more significant impact on the physiological data correlation pattern, so the weight of the emotional stress parameter correlation ratio is usually set to 0.6, and the weight of the correlation coefficient correlation ratio is set to 0.4. The weighted sum of the two gives the adjustment factor, for example, when the correlation coefficient is 0.8 and the reference correlation coefficient is 0.6, the ratio is 0.8 / 0.6≈1.33; when the emotional stress parameter is 8 microsiemens and the reference stress parameter is 5 microsiemens, the ratio is 8 / 5=1.6, and the adjustment factor is 1.33×0.4+1.6×0.6≈0.532+0.96=1.492.

[0081] The benchmark kernel function parameter of the kernel principal component analysis algorithm needs to be preset according to the type of the sleep quality parameter. For example, for continuous parameters such as sleep breathing frequency and deep sleep proportion, the RBF kernel function is often used, and the benchmark gamma value is set to 0.3. According to the formula, the new kernel function parameter is equal to the benchmark kernel function parameter multiplied by (adjustment factor plus one), that is, 0.3 x (1.492 + 1) = 0.3 x 2.492 ≈ 0.7476. At the same time, the maximum value of the kernel function parameter needs to be preset to avoid overfitting of the algorithm caused by too large parameter, and in the processing of physiological data of mental illness, the maximum value of the gamma of the RBF kernel is usually set to 0.8. If the new kernel function parameter does not exceed the maximum value, the calculation result is directly used; if the adjustment factor is larger, for example, the adjustment factor is 2.5, the new kernel function parameter is 0.3 x (2.5 + 1) = 1.05, which exceeds the maximum value 0.8, and 0.8 is taken as the final kernel function parameter.

[0082] The kernel principal component analysis algorithm is updated using the adjusted kernel function parameter, the sleep quality parameters obtained by the previous classification (such as standardized sleep latency, rapid eye movement sleep proportion, etc.) are re-input into the updated algorithm for dimensionality reduction processing, and a new dimensionality reduction feature vector is generated. Compared with before adjustment, the vector can more accurately reflect the key information of the sleep quality parameter under the current emotional stress state, and the new dimensionality reduction feature vector and the speech emotion parameter are re-input into the nonlinear mapping model to obtain the adjusted cross-modal correlation feature. The existing technology has the problem of static discrete evaluation, and cannot optimize the analysis model according to the real-time data changes in the treatment process. By dynamically adjusting the kernel function parameter according to the correlation coefficient and the emotional stress parameter, the kernel principal component analysis algorithm can adapt to the changes of the physiological state of the patient. For example, when the emotional stress of the patient increases, the adjusted kernel function parameter can enhance the ability to capture abnormal fluctuation characteristics in the sleep quality parameter, avoid feature extraction deviation caused by fixed model parameters, and solve the problem of low evaluation accuracy in the prior art. At the same time, the adjustment process is based on individualized data of the patient rather than general parameters, further improving the individualization level of evaluation, and meeting the needs of individual difference adaptation in the evaluation of the treatment effect of mental illness.

[0083] In a specific embodiment, the process of obtaining a dynamic weight coefficient according to the cross-modal correlation feature and the weight distribution mechanism can specifically include the following steps:

[0084] The contribution value of each cross-modal correlation feature to the prediction of the treatment effect is calculated, and the ratio of the absolute value of the contribution value of each feature to the sum of the absolute values of the contribution values of all features is taken as the dynamic weight coefficient of the feature.

[0085] Specifically, when obtaining the dynamic weight coefficient according to the cross-modal correlation feature and the weight distribution mechanism, firstly, the composition of the cross-modal correlation feature needs to be determined. The cross-modal correlation feature is a dimension reduction feature vector obtained by kernel principal component analysis on the sleep quality parameter, and a fusion feature generated by nonlinear mapping of the speech emotion parameter. In the psychological disease treatment effect evaluation scene, the cross-modal correlation feature can specifically include sleep rhythm and speech fundamental frequency correlation feature, sleep depth and speech emotion intensity correlation feature, and the like. Each feature corresponds to the internal correlation of the sleep and speech two-dimensional physiological data, and is a key intermediate indicator reflecting the treatment effect.

[0086] When calculating the contribution value of each cross-modal correlation feature to the treatment effect prediction, the objective reference index of the psychological disease treatment effect needs to be combined. Generally, the cross-modal correlation feature is taken as the independent variable, and the quantitative index of the treatment effect, such as the improvement rate of the patient's depressive symptoms in the treatment period, the reduction amplitude of the anxiety emotion attack frequency, the sleep disorder relief degree, and the like, is taken as the dependent variable. The feature importance evaluation method is used for calculation. Taking a linear regression model as an example, all cross-modal correlation features and the quantitative index of the treatment effect are substituted into the model for fitting. The regression coefficient corresponding to each independent variable (i.e., each cross-modal correlation feature) output by the model is the contribution value of the feature to the treatment effect prediction. The numerical value of the regression coefficient directly reflects the influence degree of the feature on the treatment effect prediction. A positive coefficient indicates that the feature positively promotes the accuracy of the treatment effect evaluation, and a negative coefficient indicates that the feature has a reverse influence on the treatment effect prediction. This calculation method can accurately capture the correlation strength between different cross-modal correlation features and the treatment effect, and avoid the deviation caused by subjective setting of feature importance.

[0087] After obtaining the contribution value of each cross-modal correlation feature, the absolute value of each contribution value needs to be taken, because the weight distribution needs to focus on the influence degree of the feature on the treatment effect prediction, rather than the influence direction, to avoid the mutual offset of positive and negative contribution values leading to distortion of the weight calculation. Subsequently, the sum of the absolute values of the contribution values of all cross-modal correlation features is calculated. The ratio operation between the absolute value of the contribution value of a single cross-modal correlation feature and the sum is performed, and the result obtained is the dynamic weight coefficient corresponding to the cross-modal correlation feature. For example, the cross-modal correlation features of a patient include feature A, feature B, and feature C. After fitting, the contribution value of feature A is 1.5, the contribution value of feature B is -0.9, and the contribution value of feature C is 0.6. The absolute values of the contribution values of the three features are 1.5, 0.9, and 0.6, respectively, and the sum is 3.0. Therefore, the dynamic weight coefficient of feature A is 1.5 / 3.0 = 0.5, the dynamic weight coefficient of feature B is 0.9 / 3.0 = 0.3, and the dynamic weight coefficient of feature C is 0.6 / 3.0 = 0.2. The weight coefficient of each feature directly corresponds to the contribution degree of the feature to the treatment effect prediction. The greater the contribution, the higher the weight.

[0088] The prior art has the problem of lack of dynamic individualization analysis. The fixed weight is used to process multi-dimensional data, which cannot adapt to the correlation differences between different patient physiological characteristics and treatment effects. The dynamic weight coefficient is calculated based on the actual contribution value of the patient's own cross-modal correlation characteristics. The weight distribution of each patient matches the influence degree of individual characteristics on the treatment effect, solving the technical problem of insufficient individualization. At the same time, the existing technology has insufficient feature representativeness due to single monitoring dimension. The cross-modal correlation characteristics cover sleep and speech dimensions. The dynamic weight can highlight the characteristics that have more significant impact on the treatment effect, avoid evaluation bias caused by unbalanced single-dimensional feature weight, further improve the accuracy of treatment effect evaluation, and meet the needs of individual difference adaptation and effective use of multi-dimensional data in psychological disease treatment effect evaluation.

[0089] In a specific embodiment, the process of identifying the abnormal pattern of the treatment effect index by the anomaly detection algorithm and generating the individualized treatment effect evaluation report can specifically include the following steps:

[0090] Obtaining a termination threshold of the treatment effect index;

[0091] Starting from the first monitoring point of the treatment effect index sequence, obtaining a time node at which the first treatment effect index value is less than or equal to the termination threshold;

[0092] Calculating the number of treatment days or treatment cycles experienced from the treatment start time point to the time node, and generating an individualized treatment effect evaluation report containing treatment recommendations in combination with the abnormal pattern analysis result.

[0093] Specifically, when obtaining the termination threshold of the treatment effect index, the clinical standard of psychological disease treatment effect evaluation and the individualized physiological baseline model of the patient need to be determined. The reference range of the treatment effect index when the treatment of the same psychological disease (such as depression and anxiety) is effective in the clinic is referred to. At the same time, the fluctuation interval of the treatment effect index in the patient's own individualized physiological baseline model is superimposed. The lower limit value of the treatment effect index when the treatment of the same disease and similar physiological characteristics of the patient reaches clinical remission is statistically determined. In combination with the influence amplitude of the index by the normal fluctuation of the patient's circadian rhythm parameters (such as sleep latency and emotional stress peak time), the termination threshold is finally determined. For example, for a patient with depression, the treatment effect index corresponding to the clinical remission of depression is usually converted into an index threshold of 28 based on the Hamilton Depression Scale score. The fluctuation of the emotional stress parameter in the patient's individualized physiological baseline model will cause a deviation of ±4 in the treatment effect index. Therefore, the termination threshold is comprehensively determined as 32.

[0094] The treatment effect index sequence is generated by daily calculation of the standardized physiological parameters (sleep quality parameters, emotional stress parameters, voice emotion parameters) with dynamic weight coefficients. The index value of each monitoring point corresponds to the current day. Starting from the first monitoring point (i.e., the day the treatment starts), the treatment effect index value of each monitoring point is compared with the termination threshold. When the treatment effect index value of a monitoring point is less than or equal to the termination threshold for the first time, the specific time corresponding to the monitoring point is recorded as the target time node. For example, the treatment effect index of the patient on the first day of treatment is 52 (corresponding to the standardized value of sleep breathing frequency 1.2, the standardized value of skin electrical response 1.5, and the standardized value of voice fundamental frequency 1.3), 47 on the second day, 41 on the third day, 36 on the fourth day, and 31 on the fifth day. The termination threshold is 32. The first time node that meets the condition is the fifth day of treatment, and the index value of this node is 31, which corresponds to the physiological state improvement of the standardized value of sleep depth ratio 0.8 and the standardized value of emotional stress parameter 0.6 on the current day.

[0095] The time interval from the treatment start time point to the target time node is calculated, and the treatment days are counted by natural days or converted into treatment cycle numbers by a preset treatment cycle (such as 7 days in clinical practice). The index change pattern is analyzed by an abnormality detection algorithm (such as a sliding window abnormality detection based on the treatment effect index sequence). If the single-day change amplitude of the treatment effect index in a certain period of time exceeds the preset fluctuation threshold (in psychological disease treatment, the normal fluctuation range of the treatment effect index is usually not more than 10, and if it exceeds, it indicates that there may be physiological data abnormalities or abnormal treatment intervention responses), it is determined as an abnormal pattern. For example, the patient's index on the second day is 47, on the third day is 41 (decreased by 6, within the normal range), on the fourth day is 36 (decreased by 5, normal), and on the fifth day is 31 (decreased by 5, normal), without abnormal pattern. If a patient's index on the third day is 40 and on the fourth day is 22 (decreased by 18, exceeding the fluctuation threshold), it is determined as an abnormal pattern, and the corresponding physiological parameter changes (such as the sudden decrease of the voice emotion parameter standardized value by 2.0 on the current day, which may be related to the patient's sudden emotional stress event on the current day) are recorded simultaneously.

[0096] Integrate the number of treatment days (or cycles), abnormal pattern analysis results (including abnormal occurrence time, corresponding physiological parameter abnormal items and fluctuation amplitude), and combine the clinical treatment guidelines for psychological diseases to generate a personalized treatment effect evaluation report. The report needs to clearly mark the time when the treatment reaches the termination threshold and the key changes in physiological parameters during the period. If there is an abnormal pattern, specific recommendations should be made (such as if the abnormal pattern is caused by sudden changes in emotional stress parameters, it is recommended to increase the monitoring frequency of emotional stress parameters and adjust the emotional relief program in psychological intervention). If there is no abnormal pattern, it is recommended to maintain the current treatment plan and continue to monitor the changes in deep sleep proportion in sleep quality parameters to consolidate the effect. The entire report content directly relates to the individual physiological data changes and treatment effect of the patient, avoiding the evaluation bias caused by relying on subjective scales in the prior art, solving the problems of strong subjectivity and poor timeliness of evaluation, and simultaneously achieving dynamic monitoring of the treatment process through abnormal pattern recognition, improving the individualization level of evaluation.

[0097] The psychological disease treatment effect evaluation method using data analysis in the embodiments of the present application is described above, please refer to Figure 2 The psychological disease treatment effect evaluation processing process using data analysis in the embodiments of the present application is described below:

[0098] First, the multi-dimensional physiological data of the patient is obtained, which covers multiple aspects and provides a basis for subsequent analysis. Then, the obtained data is standardized to make data from different sources and different scales comparable. Then, statistical characteristics and circadian rhythms are calculated to extract representative statistical characteristics and information reflecting the circadian variation of the patient's physiological state from the data. Based on this information, an individualized physiological baseline model is established, which reflects the patient's physiological characteristics in a normal state. Next, the parameter classification link is entered, and the data is divided into three categories: speech emotion parameters, sleep quality parameters, and emotional stress parameters for processing. For speech emotion parameters, cross-modal feature extraction is performed first to extract multiple features related to emotions from speech data; then the correlation coefficient is calculated to determine the correlation of these features with the treatment effect. If the correlation coefficient exceeds the threshold, the kernel function parameter is adjusted, otherwise kernel principal component analysis is performed for dimension reduction. Sleep quality parameters are directly processed by kernel principal component analysis for dimension reduction. Emotional stress parameters are first calculated with dynamic weights according to their importance, and then the treatment effect index is calculated. After the above processing, abnormal pattern recognition is performed to find possible abnormalities in the processed data. Finally, an evaluation report is generated to summarize the treatment effect of psychological diseases based on all analysis results.

[0099] The psychological disease treatment effect evaluation method using data analysis in the embodiments of the present application is described above, and the psychological disease treatment effect evaluation system 300 using data analysis in the embodiments of the present application is described below, please refer toFigure 3 The psychological disease treatment effect evaluation system using data analysis in the embodiments of the present application comprises one embodiment:

[0100] The baseline model construction module 301 is configured to obtain multi-dimensional physiological data of a patient, perform standardization processing on the multi-dimensional physiological data to obtain standardized physiological parameters, calculate statistical features and circadian rhythm patterns of each standardized physiological parameter, and generate an individualized physiological baseline model in combination with individual differences of the patient.

[0101] The feature processing module 302 is configured to divide the standardized physiological parameters into sleep quality parameters, emotional stress parameters and voice emotion parameters according to the individualized physiological baseline model, input the sleep quality parameters into a kernel principal component analysis algorithm to obtain a reduced dimension feature vector, input the reduced dimension feature vector and the voice emotion parameters into a nonlinear mapping model to obtain cross-modal correlation features, calculate a correlation coefficient of the features, and if the correlation coefficient exceeds a threshold, adjust kernel function parameters in combination with the correlation coefficient and the emotional stress parameters to re-extract the cross-modal correlation features.

[0102] The evaluation generation module 303 is configured to determine a dynamic weight coefficient according to the cross-modal correlation features and a weight distribution mechanism, perform weighted calculation on the standardized physiological parameters by using the dynamic weight coefficient to obtain a treatment effect index, identify an abnormal pattern of the treatment effect index by using an anomaly detection algorithm, and generate a personalized treatment effect evaluation report.

[0103] Through the synergistic cooperation of each component part, the baseline model construction module 301 first acquires the multi-dimensional physiological data of the patient and completes the standardization processing, calculates the statistical characteristics and the circadian rhythm mode to generate an individualized physiological baseline model, which provides the core classification basis for the feature processing module. The feature processing module divides the standardized physiological parameters into sleep quality, emotional stress and voice emotion parameters according to the deviation threshold set by the baseline model, and the circadian rhythm cycle data in the baseline model also provides a reference for the feature processing module to adjust the second deviation threshold in the night period, avoiding misjudgment of the physiological parameter fluctuation in the night. The feature processing module 302 obtains the dimension reduction feature vector by kernel principal component analysis of the sleep quality parameter, and generates the cross-modal correlation feature by nonlinear mapping of the voice emotion parameter. If the correlation coefficient of the feature is above the threshold, the emotional stress parameter and the reference stress parameter in the baseline model are called to calculate the adjustment factor, and the feature is extracted again after optimizing the kernel function parameter, and the updated cross-modal correlation feature is directly delivered to the evaluation generation module to provide multi-modal data support for dynamic weight calculation. The evaluation generation module 303 determines the dynamic weight coefficient by combining the cross-modal correlation feature and the weight distribution mechanism, and calculates the treatment effect index by weighting the standardized physiological parameters. In the process, the individual reference fluctuation range of the physiological parameters in the baseline model needs to be referred to, to ensure that the index fits the patient differences; then the abnormal detection algorithm is used to identify the abnormal mode of the index, and the parameter change frequency data of the feature processing module and the circadian rhythm rule of the baseline model are referred to during the identification, and finally the core information of each module is integrated to generate a personalized evaluation report containing treatment suggestions, forming a complete synergistic link from data acquisition to evaluation output, effectively solving the problem of lack of objective and dynamic analysis in traditional evaluation.

[0104] The above Figure 3 From the perspective of modular functional entities, the psychological disease treatment effect evaluation system using data analysis in the embodiment of the application is described in detail below. From the perspective of hardware processing, the psychological disease treatment effect evaluation device 400 using data analysis in the embodiment of the application is described in detail.

[0105] Referring to Figure 4 In the embodiment of the application, a psychological disease treatment effect evaluation device 400 using data analysis is also provided. The psychological disease treatment effect evaluation device using data analysis can be a server, and its internal structure can be as follows Figure 4The psychological disease treatment effect evaluation device using data analysis shown in the figure includes a processor 402, a memory 403, a display screen 404, an input device 405, a network interface 406 and a database 407 connected through a system bus 401. Among them, the computer designed processor 402 is used to provide computing and control ability. The memory 403 of the psychological disease treatment effect evaluation device using data analysis includes a non-volatile storage medium 4031 and an internal memory 4032. The non-volatile storage medium 4031 stores an operating system and a computer program. The internal memory 4032 provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database 407 of the psychological disease treatment effect evaluation device using data analysis is used to store the corresponding data in this embodiment. The network interface 406 of the psychological disease treatment effect evaluation device using data analysis is used to communicate with the external terminal through the network connection. The computer program executed by the processor can realize the above-mentioned method.

[0106] Those skilled in the art can understand that, Figure 4 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the psychological disease treatment effect evaluation device using data analysis to which the scheme of the present application is applied.

[0107] The present application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium, and can also be a volatile computer readable storage medium, and the computer readable storage medium stores instructions, when the instructions run on the computer, make the computer execute the steps of the psychological disease treatment effect evaluation method using data analysis.

[0108] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-mentioned system, system and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

[0109] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a data analysis-based psychological disease treatment effect evaluation device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0110] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for evaluating the effect of a mental disease treatment using data analysis, characterized by, The method comprises: Step S101, acquiring multi-dimensional physiological data of a patient, performing standardization processing on the multi-dimensional physiological data to obtain standardized physiological parameters, calculating statistical features and circadian rhythm patterns of each physiological parameter, and obtaining an individualized physiological baseline model according to individual differences of the patient, including: calculating the mean and standard deviation of each physiological parameter during the baseline period, processing the real-time collected physiological data by using a standardization formula, wherein the standardized value is equal to the original value minus the mean and then divided by the standard deviation; calculating the statistical features of sleep breathing frequency, skin electrical response value and voice fundamental frequency respectively, the statistical features including mean, variance, skewness and kurtosis; if the absolute value of the skewness of a certain physiological parameter is greater than a preset skewness threshold, then power transformation is used for data correction; taking the baseline data of the continuous fourteen days as an individual difference analysis window, acquiring the circadian rhythm parameters in the analysis window, the circadian rhythm parameters including sleep latency, rapid eye movement sleep proportion, deep sleep proportion, night awakening times, and emotional stress peak occurrence time; fitting each circadian rhythm parameter by using a sine function to obtain the individualized physiological baseline model, the model expression being that the baseline value is equal to the amplitude coefficient multiplied by the sine function plus the direct current component, wherein the period of the sine function is twenty-four hours; Step S102, dividing the standardized physiological parameters into sleep quality parameters, emotional stress parameters and voice emotion parameters based on the individualized physiological baseline model; inputting the sleep quality parameters into a kernel principal component analysis algorithm to obtain a reduced dimension feature vector, and inputting the reduced dimension feature vector and the voice emotion parameters into a nonlinear mapping model to obtain cross-modal correlation features; calculating the correlation coefficient of the cross-modal correlation features, if the correlation coefficient exceeds a threshold, adjusting the kernel function parameters of the kernel principal component analysis algorithm according to the correlation coefficient and the emotional stress parameters and reextracting the cross-modal correlation features; Step S103, obtaining a dynamic weight coefficient according to the cross-modal correlation features and a weight distribution mechanism, performing weighted calculation on the standardized physiological parameters by using the dynamic weight coefficient to obtain a treatment effect index, identifying the abnormal mode of the treatment effect index by an anomaly detection algorithm and generating a personalized treatment effect evaluation report.

2. The method of claim 1, wherein, The individualized physiological baseline model is used to divide the standardized physiological parameters into sleep quality parameters, emotional stress parameters and voice emotion parameters, including: Obtaining sleep-related data in the standardized physiological parameters that deviates from the baseline model by more than a first deviation threshold, if there are multiple parameters that meet the condition, then the parameter with the largest deviation degree is taken as the main component of the sleep quality parameter; Obtaining a second deviation threshold, if the current period is a night period, then the second deviation threshold is increased, and the parameters in the remaining parameters that deviate from the baseline by more than the second deviation threshold and have a change frequency within a preset frequency range are taken as the emotional stress parameters and the voice emotion parameters.

3. The method of claim 1, wherein, The correlation coefficient and the baseline correlation coefficient are calculated, and the weighted sum of the ratio of the correlation coefficient to the baseline correlation coefficient and the ratio of the emotional stress parameter to the baseline stress parameter is used to obtain an adjustment factor. ​ The kernel function parameter value of the kernel principal component analysis algorithm is obtained according to a formula that a new kernel function parameter is equal to a reference kernel function parameter multiplied by an adjustment factor plus one, and if the new kernel function parameter is greater than a maximum kernel function parameter value, the maximum kernel function parameter value is taken; The kernel principal component analysis algorithm is updated using the adjusted kernel function parameter, and the sleep quality parameters are reprocessed for dimension reduction using the updated algorithm.

4. The method of claim 1, wherein, The dynamic weight coefficient is obtained according to the cross-modal correlation feature and the weight distribution mechanism, including: The contribution value of each cross-modal correlation feature to the treatment effect prediction is calculated, and the ratio of the absolute value of the contribution value of each feature to the sum of the absolute values of the contribution values of all features is taken as the dynamic weight coefficient of the feature.

5. The method of claim 1, wherein, The abnormal pattern of the treatment effect index is identified through the anomaly detection algorithm, and a personalized treatment effect evaluation report is generated, including: An end threshold of the treatment effect index is obtained; A time node is obtained at which a first treatment effect index value is less than or equal to the end threshold, starting from a first monitoring point of the treatment effect index sequence; The number of treatment days or treatment cycles experienced from the treatment start time point to the time node is calculated, and a personalized treatment effect evaluation report containing treatment suggestions is generated based on the abnormal pattern analysis result. 6.A system for evaluating treatment effect of a psychological disease using data analysis, characterized by, The psychological disease treatment effect evaluation system for implementing the psychological disease treatment effect evaluation method using data analysis according to any one of claims 1 to 5, including: A baseline model construction module is configured to obtain multi-dimensional physiological data of a patient, perform standardization processing on the multi-dimensional physiological data to obtain standardized physiological parameters, calculate statistical characteristics and circadian rhythm patterns of each standardized physiological parameter, and generate an individualized physiological baseline model in combination with individual differences of the patient; A feature processing module is configured to divide the standardized physiological parameters into sleep quality parameters, emotional stress parameters, and voice emotion parameters according to the individualized physiological baseline model, input the sleep quality parameters into a kernel principal component analysis algorithm to obtain a dimension-reduced feature vector, input the dimension-reduced feature vector and the voice emotion parameters into a nonlinear mapping model to obtain cross-modal correlation features, calculate a correlation coefficient of the features, adjust a kernel function parameter if the correlation coefficient exceeds a threshold, and re-extract the cross-modal correlation features in combination with the emotional stress parameters; An evaluation generation module is configured to determine a dynamic weight coefficient according to the cross-modal correlation features and a weight distribution mechanism, perform weighted calculation on the standardized physiological parameters using the dynamic weight coefficient to obtain a treatment effect index, identify an abnormal pattern of the treatment effect index through an anomaly detection algorithm, and generate a personalized treatment effect evaluation report.

7. A mental disease treatment effect evaluation device using data analysis, characterized by, The computer program causes the processor to execute the psychological disease treatment effect evaluation method using data analysis according to any one of claims 1 to 5 when the computer program is run on the processor.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program causes the processor to execute the psychological disease treatment effect evaluation method using data analysis according to any one of claims 1 to 5 when the computer program is run on the processor.

Images