Depression severity index prediction method, electronic device, and storage medium
Patent Information
- Application Number
- CN202610896096.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-22
- Publication Date
- 2026-09-15
Smart Images
Figure CN122762233A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to a method for predicting the severity index of depression, an electronic device, and a storage medium. Background Technology
[0002] With increasing attention being paid to mental health issues, the early identification and severity assessment of depression have become important research directions.
[0003] In related technologies, many methods employ single-dimensional PHQ-9 self-assessment scales or facial expression behavior analysis, using logistic regression models to output the probability of depression risk. This approach relies excessively on the authenticity of the subjects' subjective statements, and single-modal data struggles to capture subconscious pathological characteristics, resulting in a persistently high rate of missed diagnoses of moderate and masked depression. Summary of the Invention
[0004] This invention provides a method for predicting the severity index of depression, an electronic device, and a storage medium to address the problem of high missed detection and misjudgment rates in depression assessment caused by reliance on single-dimensional data in related technologies.
[0005] According to one aspect of the present invention, a method for predicting the severity index of depression is provided, comprising: Obtain multidimensional test data of the target object, wherein the test data includes at least subjective layer data, behavioral layer data, and physiological layer data; The multidimensional test data is input into a pre-trained index prediction model to obtain a prediction index of the severity of depression in the target subject.
[0006] According to another aspect of the present invention, a depression severity index prediction device is provided, comprising: The acquisition module is used to acquire multidimensional test data of the target object, wherein the test data includes at least subjective layer data, behavioral layer data and physiological layer data; The prediction module is used to input the multidimensional test data into a pre-trained index prediction model to obtain a prediction index of the severity of depression of the target object.
[0007] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the depression severity index prediction method according to any embodiment of the present invention.
[0008] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the depression severity index prediction method according to any embodiment of the present invention.
[0009] According to another aspect of the present invention, embodiments of this disclosure also provide a computer program product, including a computer program that, when executed by a processor, implements the depression severity index prediction method as described in any of the embodiments of this disclosure.
[0010] The technical solution of this invention acquires multidimensional test data of the target object. Since the test data includes at least subjective, behavioral, and physiological data, it can integrate subjective and objective multi-source data to provide a data foundation for subsequent prediction. The multidimensional test data is input into a pre-trained index prediction model to obtain a prediction index of the severity of depression of the target object. The pre-trained model can intelligently analyze the data and accurately output a quantitative index of depression severity, solving the problem of high missed detection and misjudgment rates in depression assessment caused by reliance on single-dimensional data in related technologies. Multidimensional data fusion modeling significantly improves the objectivity and quantitative accuracy of depression assessment.
[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description
[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0013] Figure 1 This is a flowchart of a method for predicting the severity index of depression according to Embodiment 1 of the present invention; Figure 2 This is a flowchart of a method for predicting the severity index of depression according to Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the structure of a depression severity index prediction device provided in Embodiment 3 of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device that implements the depression severity index prediction method according to embodiments of the present invention. Detailed Implementation
[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0016] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".
[0017] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.
[0018] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.
[0019] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.
[0020] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.
[0021] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.
[0022] It is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.
[0023] Example 1 Figure 1 The flowchart of a method for predicting the severity index of depression is provided in Embodiment 1 of the present invention. This embodiment is applicable to psychological screening and dynamic assessment of home-based depression. The method can be executed by a device for predicting the severity index of depression, which can be implemented in hardware and / or software. Optionally, it can be implemented through an electronic device, such as a mobile terminal, a PC, or a server.
[0024] like Figure 1 As shown, the method may specifically include: S110. Obtain multidimensional test data of the target object, wherein the test data includes at least subjective layer data, behavioral layer data and physiological layer data.
[0025] The target group can be understood as the individual being assessed and tested, such as a patient, subject, or screened individual. They are the primary recipients of the multidimensional test data and the output of the depression severity prediction index. The multidimensional test data can be understood as a dataset reflecting psychological and neurophysiological states, obtained from multiple different dimensions (subjective, behavioral, and physiological). By integrating subjective reports, behavioral tasks, and physiological signals, it overcomes the limitations of a single questionnaire or physiological indicator, improving the objectivity and accuracy of depression assessment.
[0026] One alternative implementation involves constructing a standardized empathy-inducing task. Participants are presented with materials depicting others' pain, negative emotional faces, and social interaction scenarios to induce empathic processing at different levels, including emotional resonance, cognitive understanding, self-involvement, and inference of others' states. The duration of stimulus presentation, stimulus type, trial order, response window, and scoring window are uniformly controlled, and the participants' subjective experiences, behavioral responses, and physiological responses are recorded simultaneously for each trial.
[0027] Furthermore, the first The multi-level original feature set obtained from each trial is represented as follows: in, Indicates the first Multidimensional test data from each trial; Represents subjective layer data; Represents behavioral layer data; This represents physiological layer data. Through the above method, subjective, behavioral, and physiological information in the empathy processing process is organized into a unified hierarchical data structure, providing a data foundation for subsequent cross-level representation learning and mismatch feature modeling.
[0028] Specifically, subjective data includes empathy scores, pain perception scores, emotional involvement scores, and self-other distinction scores; behavioral data includes reaction time, recognition accuracy, judgment and selection, attention allocation characteristics, and task performance indicators; and physiological data includes electroencephalogram (EEG) signals, skin conductance, pupillary response, heart rate variability, eye movement tracking, facial muscle activity, and speech prosody characteristics.
[0029] Based on the above scheme, optionally, the subjective layer data includes at least one of the target object's subjective feelings about the second reference object's emotions and the degree of self-reference; the behavioral layer data includes at least one of the target object's recognition efficiency, judgment accuracy, and reaction deviation in the emotion judgment process; and the physiological layer data includes at least one of the neural oscillation characteristics, nervous system activation degree, and peripheral somatic physiological representation feedback in the emotion processing stage.
[0030] The subjective layer data can be understood as feedback data generated based on the target object's own subjective perception and inner experience, reflecting the object's internal emotions and psychological cognitive state, and is the core subjective basis for judging psychological abnormalities. The second reference object can be understood as an external reference person / simulated object used as an emotional reference, judged by the target object's emotional perception. Setting up an emotional reference subject is used to test the target object's ability to perceive external emotions. The subjective feeling data can be understood as the target object's personal subjective feelings and judgment records regarding the emotions displayed by the second reference object, reflecting the individual's perception and empathy for external emotions, and is an important basis for psychological state assessment. The degree of self-reference can be understood as the degree of association between the target object and external emotions and events and their own experiences and psychological state, i.e., the degree of self-involvement (or identification), reflecting individual cognitive biases. Depressed individuals often exhibit excessive self-association characteristics, which assists in the diagnosis of symptoms.
[0031] The behavioral layer data can be understood as the data related to the target object's external behavior and operational performance in a specified emotion judgment task. It captures the differences in external behavior during the emotion cognition process, compensating for the subjective bias of subjective data. The emotion judgment process can be understood as the complete process by which the target object identifies, distinguishes, and judges the type and intensity of external emotions. It unifies the test scenario and task standards, ensuring consistency in the conditions for collecting all behavioral data. The recognition efficiency can be understood as quantitative indicators such as the time spent and reaction speed by the target object in completing the emotion recognition task, reflecting the speed of emotion cognitive processing. Depression often leads to a decrease in recognition efficiency. The judgment accuracy can be understood as the degree to which the target object's judgments of various emotions match the standard answers, measuring the correctness of emotion cognitive judgments. Abnormal psychological states easily lead to increased judgment errors. The reaction bias can be understood as the habitual bias, judgment deviation, or tendency to misjudge that the target object exhibits in emotion judgment. Capturing abnormal cognitive biases is a key behavioral characteristic distinguishing normal people from depressed individuals.
[0032] The physiological data can be understood as the objective physiological signals and indicators generated by the body and nervous system during the emotional processing stage of the target subject. This data is objective and unmasked, truly reflecting the body's internal physiological stress state under emotional stimulation. The emotional processing stage can be understood as the entire process of the target subject receiving emotional stimuli and the brain processing emotional information. Defining the time period for collecting physiological data ensures that the data corresponds to the true response of emotional processing. The neural oscillation characteristics can be understood as the characteristic signals such as the brain electrical rhythm and fluctuation patterns formed by the synchronous discharge of brain neuron groups, reflecting the activity patterns of the brain's emotional processing areas. Depression alters the neural oscillation rhythm. The nervous system activation level can be understood as the excitation and activity levels of relevant brain regions and the central nervous system when facing emotional stimuli, quantifying the response intensity of the central nervous system. Abnormal activation levels are a typical physiological characteristic of mental illness. The peripheral somatic physiological representation feedback can be understood as the various physiological indicators and changes exhibited by the body's peripheral organs, excluding the central nervous system. Collecting peripheral signs such as heart rate, electromyography, and skin conductance provides physical evidence of abnormal emotional and psychological states.
[0033] This technical solution effectively overcomes the limitations of single scales being susceptible to faking and social expectation bias by integrating complementary data from multiple dimensions such as subjective feelings, behavioral responses, and neurophysiology. It significantly improves the objectivity, accuracy, and anti-interference ability of predicting the severity of depression and reveals the underlying pathological mechanisms.
[0034] S120. Input the multidimensional test data into the pre-trained index prediction model to obtain the severity prediction index of depression of the target object.
[0035] The index prediction model can be understood as a machine learning / deep learning regression or classification model trained with a large amount of "multidimensional test data + known depression severity labels". It can directly receive data from the subjective, behavioral, and physiological levels and output a depression severity prediction index. The severity prediction index can be understood as a numerical indicator calculated by the model to quantify the severity of depression in the target subject, transforming the abstract severity of depression into an intuitive numerical value, and achieving standardized and quantitative assessment.
[0036] Based on the above scheme, optionally, the training process of the index prediction model includes: acquiring multidimensional sample data from multiple tests of the sample object, and the severity index label corresponding to the sample object; for each test's multidimensional sample data, inputting the multidimensional sample data into an initial encoder to obtain a multidimensional initial sample compression space vector, determining the encoding loss based on the multidimensional sample compression space vector, adjusting the model parameters of the initial encoder based on the encoding loss to obtain a pre-trained encoder; inputting the multidimensional sample data into the pre-trained encoder to obtain a multidimensional target sample compression space vector; and inputting the multidimensional target sample compression space vector into the pre-trained encoder to obtain a multidimensional target sample compression space vector. The sample high-order feature vector is obtained by inputting the sample high-order feature vector and the severity index label into the initial prediction module to obtain the sample severity prediction index of the sample object's depression; the average value of the sample severity prediction index from multiple tests is calculated, and the prediction loss is determined based on the average value and the severity index label; the model parameters of the initial feature modeling module and the initial prediction module are adjusted based on the prediction loss to obtain the pre-trained feature modeling module and the initial prediction module; the index prediction model is constructed based on the pre-trained encoder, the feature modeling module, and the initial prediction module.
[0037] The sample objects can be understood as subjects used for model training, collecting test data, and labeling. They provide the data and ground truth labels needed for training and are the main data source for the model's learning patterns. The multidimensional sample data can be understood as a collection of sample object test data collected from multiple dimensions such as subjective, behavioral, and physiological aspects. It carries multidimensional feature information and provides complete input material for model training. The severity index label can be understood as a standardized quantitative value corresponding to the severity of depression of the sample object, obtained through manual / professional assessment. It serves as the training ground truth, used to judge the correctness of the model's prediction results and guide parameter optimization. Multiple tests can be understood as the operation of conducting multiple rounds of repeated tests and collecting data on the same sample object, reducing the randomness of single data and improving the stability and reliability of training data. The initial encoder can be understood as an encoding network module in its original state without parameter optimization. It is responsible for feature compression and dimensional transformation of the original multidimensional data and is the front-end feature processing unit of the model. The initial sample compressed space vector can be understood as a low-dimensional space feature vector obtained after the original multidimensional sample data is transformed by the initial encoder. It simplifies redundant information and transforms the original data into a standardized vector form that is easy for the model to process. The encoding loss can be understood as the output value of a loss function that measures the quality of the encoder output. It is used for backpropagation to update the initial encoder parameters, making the compressed vector more representative and discriminative. The pre-trained encoder can be understood as an encoder with fixed parameters after optimization by the encoding loss. It stabilizes the feature extraction process in subsequent training and prevents training instability caused by drastic changes in the feature space. The target sample compressed space vector can be understood as a stable low-dimensional feature vector output by the pre-trained encoder from the multidimensional sample data. It is used for downstream modeling and serves as the input to the feature modeling module, carrying denoised and structured multidimensional information. The initial feature modeling module can be understood as an untrained feature modeling network used for cross-modal and cross-temporal high-order feature interaction and abstract representation learning in the compressed space. The sample high-order feature vector can be understood as an abstract feature vector output by the feature modeling module. It integrates high-level semantics of multidimensional information and is closer to the potential determinants of depression severity than the compressed vector. It is the direct input to the prediction module. The initial prediction module can be understood as an untrained regression or classification prediction head used to map high-order features to specific depression severity prediction indices. The sample severity prediction index can be understood as the depression severity estimate output by the prediction module for a single test sample, which is compared with the true label to calculate the loss and drive joint optimization between the prediction module and the feature modeling module. The sample severity prediction index from multiple tests can be understood as a set of multiple prediction indices obtained from multiple tests on the same object, used to calculate the average, reduce fluctuations in single predictions, and improve training stability and generalization ability.The prediction loss can be understood as the loss that measures the difference between the predicted average value and the true severity index label. It is used for backpropagation to update the parameters of the feature modeling module and the prediction module, thereby achieving end-to-end optimization.
[0038] This technical solution, which combines a phased training strategy with an averaging mechanism from multiple tests, effectively reduces random noise and individual state fluctuations in a single measurement, significantly improving the model's stability, robustness, and generalization ability in predicting the severity of depression. At the same time, the pre-trained encoder accurately extracts deep cross-modal features, ensuring the reliability and clinical applicability of the assessment results.
[0039] Based on the above scheme, optionally, after inputting the multidimensional test data into the pre-trained index prediction model to obtain the severity prediction index of the target object's depression, the method further includes: standardizing the severity prediction index, and determining the severity level based on the standardized result and multiple preset classification thresholds.
[0040] The standardization can be understood as the process of converting the severity prediction index into a uniform-scale value according to certain rules (such as Z-score, Min-Max normalization, scale benchmarking, etc.), eliminating numerical differences caused by different models and different training distributions, and making it comparable to a preset threshold on the same scale. The classification threshold can be understood as a predefined set of critical values (such as the mild / moderate / severe dividing point) used to classify the standardized index into different severity levels, providing a basis for grading and transforming continuous predicted values into interpretable categorical results. The severity level can be understood as a category label (such as normal, mild, moderate, severe) that discretizes the severity of depression, facilitating intuitive understanding of the prediction results and guiding subsequent interventions or risk assessments.
[0041] By standardizing the predictive index and comparing it with the classification threshold, the continuous model output is transformed into an intuitive clinical grade, thus achieving standardization and interpretability of the assessment results, which facilitates doctors to quickly classify and develop intervention plans.
[0042] Based on the above scheme, optionally, after inputting the multidimensional test data into the pre-trained index prediction model to obtain the severity prediction index of the target object's depression, the method further includes: generating a prediction report based on multiple mismatch distances, the severity prediction index, and the severity level of the target object.
[0043] The mismatch distance can be understood as a set of multiple quantified deviation values calculated from different dimensions, feature levels, or reference sources, supporting multi-angle consistency checks (such as whether subjective-behavioral-physiological consistency exists), thus improving the interpretability and credibility of the report. The mismatch distance can be calculated by compressing vector pairs composed of any two dimensions in the spatial vector (such as subjective-behavioral, behavioral-physiological, physiological-subjective, etc.). The prediction report can be understood as a structured output document or interface display that integrates the severity prediction index, severity level, and multiple mismatch distances, used to summarize all evaluation information and form a complete, viewable, and applicable final assessment conclusion.
[0044] One alternative implementation converts the individual-level severity prediction index into a standardized depression severity index ranging from 0 to 100: ; in, Indicates the first Standardized depression severity index of 100 subjects; Indicates the first Predicted severity index of 100 subjects; and These represent the lowest and highest severity index values in the reference scale or training samples, respectively.
[0045] A grading system is generated based on a standardized severity index and multiple classification thresholds, such as no risk, mild, moderate, and severe risk levels. Simultaneously, an interpretable report is generated by combining multiple mismatch distances, severity prediction indices, and the severity level of the target subject. This report indicates that the primary source of abnormality in the subject is subjective-behavioral mismatch, subjective-physiological mismatch, behavioral-physiological mismatch, or self-other processing bias.
[0046] By adopting this technical solution, which integrates mismatch distance, prediction index, and grade to generate reports, the interpretability and credibility of the results can be enhanced, revealing inherent contradictions and deviations in the data, and providing a more comprehensive and transparent risk assessment basis for clinical decision-making.
[0047] The technical solution of this invention acquires multidimensional test data of the target object. Since the test data includes at least subjective, behavioral, and physiological data, it can integrate subjective and objective multi-source data to provide a data foundation for subsequent prediction. The multidimensional test data is input into a pre-trained index prediction model to obtain a prediction index of the severity of depression of the target object. The pre-trained model can intelligently analyze the data and accurately output a quantitative index of depression severity, solving the problem of high missed detection and misjudgment rates in depression assessment caused by reliance on single-dimensional data in related technologies. Multidimensional data fusion modeling significantly improves the objectivity and quantitative accuracy of depression assessment.
[0048] Example 2 Figure 2 This is a flowchart of a method for predicting the severity index of depression provided in Embodiment 2 of the present invention. This embodiment is based on the above embodiments and further refines the input of multidimensional test data into a pre-trained index prediction model to obtain a predicted index of the severity of depression of the target object. Optionally, the index prediction model includes an encoder, a feature modeling module, and a prediction module; the step of inputting multidimensional test data into the pre-trained index prediction model to obtain a predicted index of the severity of depression of the target object includes: inputting the multidimensional test data into the encoder to obtain a multidimensional compressed space vector, wherein the dimensions of the compressed space vector correspond to the subjective layer data, behavioral layer data, and physiological layer data; inputting the multidimensional compressed space vector into the feature modeling module to obtain a high-order feature vector; and inputting the high-order feature vector into the prediction module to obtain a predicted index of the severity of depression of the target object. Specific implementation details can be found in the description of this embodiment. Technical features that are the same as or similar to those in the foregoing embodiments will not be repeated here.
[0049] like Figure 2 As shown, the method may specifically include: S210. Obtain multidimensional test data of the target object, wherein the test data includes at least subjective layer data, behavioral layer data and physiological layer data.
[0050] S220. Input the multidimensional test data into the encoder to obtain a multidimensional compressed space vector, wherein the dimensions of the compressed space vector correspond to the subjective layer data, behavioral layer data and physiological layer data.
[0051] The compressed space vector can be understood as a set of low-dimensional feature vectors generated after multi-dimensional test data is converted by an encoder. It corresponds to various types of original data in different dimensions, which can unify the data format, reduce the data dimension, retain independent features according to data type, and connect to downstream modules.
[0052] In one optional implementation, since the subjective, behavioral, and physiological data differ in dimensions, dimensions, sampling frequency, and statistical distribution, the present invention constructs corresponding feature mapping functions to map the data of different levels to a unified latent representation space, so as to ensure comparability between different levels in the future.
[0053] Specifically, for the first The subjective, behavioral, and physiological data from each trial were encoded separately to obtain: ; ; ; in, , , Each represents a feature mapping function for the subjective layer, behavioral layer, and physiological layer; , , These represent the compressed spatial vectors of the subjective layer, behavioral layer, and physiological layer after mapping, respectively. The feature mapping function can be implemented by linear projection, kernel mapping, neural network encoder, or other feature transformation models.
[0054] An optional implementation, to ensure that the distance calculations for different levels of representation after encoder output can be performed in a unified space, makes the compressed space vectors of the three types of samples have the same dimension: ; Here, d represents the dimension of the unified latent representation space. During the training phase, cross-level representation consistency constraints are constructed by computation, i.e., encoding loss is constructed, to ensure that the representations of the subjective layer, behavioral layer, and physiological layer maintain structural correlation within the same trial. The parameters here represent the meaning of the training phase, and the parameter expressions used in the S220 application phase are the same as those described above. The encoding loss can be expressed as: ; in, N represents the encoding loss; N represents the number of trials involved in training or modeling. This represents the Euclidean norm. At this point... , , It is a fixed coefficient.
[0055] Another alternative implementation method is through weighted... , , The encoding loss can be expressed as: ; in, , , These are the weighting coefficients.
[0056] Another alternative implementation can achieve feature space unification by aligning the feature vectors of the subjective and physiological layers to the feature vectors of the behavioral layer, with an encoding loss of: .
[0057] By using the above methods, we can obtain an empathic processing representation with unified dimensions and structural comparability, which provides a foundation for subsequent calculations of cross-level mismatch relationships.
[0058] S230. Input the multidimensional compressed space vector into the feature modeling module to obtain a high-order feature vector.
[0059] The higher-order feature vector can be understood as a fused feature vector obtained by deep processing of the compressed space vector through the feature modeling module, which integrates multi-dimensional feature relationships to form a more recognizable deep feature.
[0060] Optionally, the test data further includes state characterization data of the target object, and the higher-order feature vector includes mismatch intensity and offset index. The step of inputting the multidimensional compressed space vector into the feature modeling module to obtain the higher-order feature vector includes: inputting the multidimensional compressed space vector into the feature modeling module so that the feature modeling module determines multiple mismatch distances based on each two-dimensional compressed space vector, and determines the mismatch intensity based on the multiple mismatch distances; obtaining reference state data of the first reference object, and determining the offset index based on the state characterization data and the reference state data.
[0061] The state representation data can be understood as quantifiable data reflecting the current psychological or physiological state of the target object, or potential representations of self-emotional involvement or self-related processing, used to compare with a reference state to calculate an "offset index," reflecting the degree of deviation of the current state from a healthy / baseline state. The mismatch strength can be understood as a comprehensive index obtained by aggregating multiple mismatch distances, representing the overall degree of inconsistency, used to measure the consistency and stability within multidimensional test data, and can serve as a basis for predictive confidence or anomaly alerts. The first reference object can be understood as an individual or group serving as a comparison benchmark (such as healthy individuals, age-group norms, historical baselines, etc.), providing a "normal" or "ideal" state reference for calculating the offset index. The reference state data can be understood as data of the first reference object in the same state representation dimension, serving as a benchmark value for offset calculation, used to measure the degree of deviation of the target object. The offset index can be understood as a quantitative value representing the degree of difference between the target object's current state representation data and the reference state data, reflecting the magnitude of deviation of the individual's current psychological state from a healthy or ideal state, assisting in judging the risk level.
[0062] By introducing mismatch intensity and offset index, this technical solution quantifies the inherent contradictions of multimodal data and the deviation of individuals from the healthy baseline while making predictions. This significantly enhances the interpretability of the results and the ability to diagnose abnormalities, effectively distinguishes between unreliable data and true pathological conditions, and reduces the risk of misjudgment.
[0063] Optionally, determining the mismatch intensity based on the plurality of mismatch distances includes at least one of the following: determining the mismatch intensity based on the weighted sum of the plurality of mismatch distances; constructing a mismatch matrix based on the plurality of mismatch distances; obtaining a target norm based on the constructed mismatch matrix; and determining the target norm as the mismatch intensity.
[0064] The mismatch matrix can be understood as a square matrix formed by arranging the pairwise mismatch distances using the dimensions of the compressed space vector as row and column indices. This organizes the dispersed mismatch distances into a structured form, facilitating overall analysis using linear algebra tools. The target norm can be understood as a matrix norm (such as the Frobenius norm, spectral norm, nuclear norm, etc.) applied to the mismatch matrix, compressing the information of the entire matrix into a scalar to measure the overall mismatch scale or energy intensity.
[0065] An optional implementation defines the first... Mismatch distances between the subjective layer and the behavioral layer, the subjective layer and the physiological layer, and the behavioral layer and the physiological layer in each trial: ; ; ; in, Indicates the first The mismatch distance between the compressed space vector corresponding to the subjective layer data and the compressed space vector corresponding to the behavioral layer data in each trial; This represents the mismatch distance between the compressed space vector corresponding to the subjective layer data and the compressed space vector corresponding to the physiological layer data. This represents the mismatch distance between the compressed space vector corresponding to the behavioral layer data and the compressed space vector corresponding to the physiological layer data.
[0066] Furthermore, the pairwise mismatch distances mentioned above are organized into a mismatch matrix based on empathy processing: ; in, Indicates the first The empathic processing mismatch matrix for each trial; the rows and columns of the matrix correspond to the subjective layer, behavioral layer, and physiological layer, respectively; the diagonal elements are 0, indicating that no mismatch is calculated between the same level and itself; the off-diagonal elements represent the mismatch intensity between different levels.
[0067] Mismatch strength can be defined as follows: ; in: ; in, Indicates the first The mismatch intensity in each trial; , , These represent the weights of the subjective-behavioral, subjective-physiological, and behavioral-physiological mismatch distances, respectively. These weights can be set based on preset rules, training data, or task importance. This index is used to quantify the overall inconsistency between multi-level responses during empathic processing.
[0068] In an alternative implementation, the Frobenius norm of the mismatch matrix can also be used to represent the overall mismatch intensity: .
[0069] Furthermore, to characterize the shift between an individual's self-involvement and their understanding of others' states during the empathy process, a shift index is defined: ; in, Indicates the first The offset index of each trial; State representation data indicating self-emotional involvement or self-related processing; This refers to reference state data used for recognizing, understanding, or inferring the states of other objects. This indicator reflects the degree of deviation between self-related processing and other-related processing.
[0070] Finally, the mismatch strength and the offset exponent are combined to form a higher-order feature vector. : .
[0071] This technical solution determines the mismatch intensity through a dual-path approach of weighted summation and matrix norm, balancing computational flexibility with mathematical rigor. It can adapt to different clinical weight preferences and globally quantify multidimensional data conflicts, significantly improving the robustness and credibility of the assessment.
[0072] S240. Input the higher-order feature vector into the prediction module to obtain the severity prediction index of the target object's depression.
[0073] In obtaining high-order feature vectors Then, a mapping relationship between it and the depression severity index was established. The prediction model uses... As input, the output is the secondary severity prediction index: ; in, Indicates the first The predictive index of depression severity corresponding to each trial; θ represents the severity prediction model; θ represents the model parameters.
[0074] Since the same subject typically comprises multiple valid trials, the trial-level prediction results are further aggregated into individual-level severity prediction values: ; in, Indicates the first Predictive values for the severity of depression in 100 subjects; Indicates belonging to the first The effective set of trials for 10 subjects; Indicates the first The number of valid trials for each subject.
[0075] Severity prediction error is defined as: ; in, Indicates the total number of subjects; Indicates the first The severity index labels for each participant can be derived from depression scale scores; This represents the severity index of depression predicted by the model.
[0076] This technical solution uses an encoder to compress multi-dimensional test data in different dimensions, reducing data volume and computational complexity while preserving the original characteristics of various data types. Then, the feature modeling module deeply mines the intrinsic relationships between data to generate high-order feature vectors rich in deep information. Finally, the prediction module completes the exponential output. By processing the data in layers and refining effective features step by step, the integrity of the original information is ensured, and the feature expression ability is improved, making the prediction results of the severity of depression more accurate and stable.
[0077] Example 3 Figure 3 This is a schematic diagram of a depression severity index prediction device provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes: an acquisition module 310 and a prediction module 320. Wherein, The acquisition module 310 is used to acquire multidimensional test data of the target object, wherein the test data includes at least subjective layer data, behavioral layer data and physiological layer data; the prediction module 320 is used to input the multidimensional test data into a pre-trained index prediction model to obtain the severity prediction index of the target object's depression.
[0078] The technical solution of this invention acquires multidimensional test data of the target object through an acquisition module. Since the test data includes at least subjective, behavioral, and physiological data, it can integrate subjective and objective multi-source data to provide a data foundation for subsequent prediction. The prediction module inputs the multidimensional test data into a pre-trained index prediction model to obtain a prediction index of the severity of depression of the target object. The pre-trained model can intelligently analyze the data and accurately output a quantitative index of depression severity, solving the problem of high missed detection and misjudgment rates in depression assessment caused by reliance on single-dimensional data in related technologies. Multi-dimensional data fusion modeling significantly improves the objectivity and quantitative accuracy of depression assessment.
[0079] Optionally, the prediction module includes a compression submodule, a modeling submodule, and a prediction submodule. The compression submodule is used to input the multidimensional test data into the encoder to obtain a multidimensional compressed space vector, wherein the dimensions of the compressed space vector correspond to the subjective layer data, behavioral layer data, and physiological layer data. The modeling submodule is used to input the multidimensional compressed space vector into the feature modeling module to obtain a high-order feature vector. The prediction submodule is used to input the high-order feature vector into the prediction module to obtain a prediction index of the severity of depression in the target subject.
[0080] Optionally, the test data further includes state characterization data of the target object, and the high-order feature vector includes mismatch strength and offset index; the modeling submodule includes a mismatch strength determination unit and an offset index determination unit. The mismatch strength determination unit is used to input the multidimensional compressed space vector into the feature modeling module, so that the feature modeling module determines multiple mismatch distances based on every two dimensions of the compressed space vector, and determines the mismatch strength based on the multiple mismatch distances; the offset index determination unit is used to obtain reference state data of the first reference object, and determine the offset index based on the state characterization data and the reference state data.
[0081] Optionally, the mismatch intensity determination unit is used to: determine the mismatch intensity based on the weighted sum of multiple mismatch distances; construct a mismatch matrix based on the multiple mismatch distances; obtain a target norm based on the constructed mismatch matrix; and determine the target norm as the mismatch intensity.
[0082] Optionally, the depression severity index prediction device includes a training module. The training module is used to train the index prediction model. The training process of the index prediction model includes: acquiring multidimensional sample data from multiple tests of a sample object, and the severity index label corresponding to the sample object; for each test's multidimensional sample data, inputting the multidimensional sample data into an initial encoder to obtain a multidimensional initial sample compression space vector, determining the encoding loss based on the multidimensional sample compression space vector, adjusting the model parameters of the initial encoder based on the encoding loss, and obtaining a pre-trained encoder; inputting the multidimensional sample data into the pre-trained encoder to obtain a multidimensional target sample compression space vector; inputting the multidimensional target sample compression space vector into an initial feature modeling module to obtain a sample high-order feature vector; inputting the sample high-order feature vector and the severity index label into an initial prediction module to obtain a sample severity prediction index for the sample object's depression; calculating the average value of the sample severity prediction index from multiple tests, determining the prediction loss based on the average value and the severity index label, adjusting the model parameters of the initial feature modeling module and the initial prediction module based on the prediction loss, and obtaining a pre-trained feature modeling module and an initial prediction module; and constructing the index prediction model based on the pre-trained encoder, the feature modeling module, and the initial prediction module.
[0083] Optionally, the depression severity index prediction device includes a severity level determination module. The severity level determination module is used to, after inputting the multidimensional test data into a pre-trained index prediction model to obtain a severity prediction index for the target object's depression, standardize the severity prediction index and determine the severity level based on the standardized result and multiple preset classification thresholds.
[0084] Optionally, the depression severity index prediction device includes a prediction report generation module. The prediction report generation module is used to generate a prediction report based on multiple mismatch distances, the severity prediction index, and the severity level of the target object after obtaining the predicted severity index of depression for the target object by inputting the multidimensional test data into a pre-trained index prediction model.
[0085] Optionally, the subjective layer data includes at least one of the target object's subjective feelings about the second reference object's emotions and the degree of self-reference; the behavioral layer data includes at least one of the target object's recognition efficiency, judgment accuracy, and reaction bias in the emotion judgment process; and the physiological layer data includes at least one of the neural oscillation characteristics, nervous system activation level, and peripheral somatic physiological representation feedback during the emotion processing stage.
[0086] The depression severity index prediction device provided in the embodiments of the present invention can execute the depression severity index prediction method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method execution.
[0087] Example 4 Figure 4 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.
[0088] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.
[0089] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0090] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as a depression severity index prediction method.
[0091] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication unit 19, or installed from storage unit 18, or installed from ROM 12. When the computer program is executed by processor 11, it performs the functions defined in the methods of the embodiments of the present invention.
[0092] In some embodiments, a method for predicting the severity index of depression may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or mounted on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method for predicting the severity index of depression described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform a method for predicting the severity index of depression by any other suitable means (e.g., by means of firmware).
[0093] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.
[0094] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0095] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.
[0096] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).
[0097] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.
[0098] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.
[0099] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.
[0100] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.
Claims
1. A method of predicting the severity index of depression, characterized by, include: Obtain multidimensional test data of the target object, wherein the test data includes at least subjective layer data, behavioral layer data, and physiological layer data; The multidimensional test data is input into a pre-trained index prediction model to obtain a prediction index of the severity of depression in the target subject.
2. The method of claim 1, wherein, The index prediction model includes an encoder, a feature modeling module, and a prediction module; the step of inputting the multidimensional test data into the pre-trained index prediction model to obtain the severity prediction index of the target object's depression includes: The multidimensional test data is input into the encoder to obtain a multidimensional compressed space vector, wherein the dimensions of the compressed space vector correspond to the subjective layer data, behavioral layer data and physiological layer data; The compressed space vector of the multidimensional modeling module is input into the feature modeling module to obtain the high-order feature vector; The higher-order feature vector is input into the prediction module to obtain the severity prediction index of the target object's depression.
3. The method of claim 2, wherein, The test data also includes state representation data of the target object. The higher-order feature vector includes mismatch strength and offset exponent. The process of inputting the multidimensional compressed space vector into the feature modeling module to obtain the higher-order feature vector includes: The multidimensional compressed space vector is input into the feature modeling module so that the feature modeling module determines multiple mismatch distances based on each two-dimensional compressed space vector, and determines the mismatch intensity based on the multiple mismatch distances; Obtain reference state data of the first reference object, and determine the offset index based on the state characterization data and the reference state data.
4. The method of claim 3, wherein, The determination of mismatch intensity based on a plurality of mismatch distances includes at least one of the following: The mismatch intensity is determined based on the weighted summation of multiple mismatch distances. A mismatch matrix is constructed based on multiple mismatch distances, and a target norm is obtained based on the constructed mismatch matrix. The target norm is then determined as the mismatch intensity.
5. The method of claim 1, wherein, The training process of the index prediction model includes: Obtain multidimensional sample data from multiple tests of the sample object, as well as the severity index label corresponding to the sample object; For each test, the multidimensional sample data is input into the initial encoder to obtain a multidimensional initial sample compression space vector. The encoding loss is determined based on the multidimensional sample compression space vector. The model parameters of the initial encoder are adjusted based on the encoding loss to obtain a pre-trained encoder. The multidimensional sample data is input into a pre-trained encoder to obtain a multidimensional target sample compressed space vector; The multidimensional target sample compressed space vector is input into the initial feature modeling module to obtain the sample high-order feature vector; Input the high-order feature vector of the sample and the severity index label into the initial prediction module to obtain the sample severity prediction index of the depression of the sample object; The average value of the severity prediction index of samples from multiple tests is calculated. The prediction loss is determined based on the average value and the severity index label. The model parameters of the initial feature modeling module and the initial prediction module are adjusted based on the prediction loss to obtain the pre-trained feature modeling module and the initial prediction module. The exponential prediction model is constructed based on the pre-trained encoder, the feature modeling module, and the initial prediction module.
6. The method of claim 1, wherein, After obtaining the severity prediction index of the target subject's depression from the multidimensional test data input into the pre-trained index prediction model, the method further includes: The severity prediction index is standardized, and the severity level is determined based on the standardized result and multiple preset classification thresholds.
7. The method of claim 3, wherein, After obtaining the severity prediction index of the target subject's depression from the multidimensional test data input into the pre-trained index prediction model, the method further includes: A prediction report is generated based on multiple mismatch distances, severity prediction indices, and the severity level of the target object.
8. The method according to claim 1, characterized in that, The subjective layer data includes at least one of the target object's subjective feelings about the second reference object's emotions and the degree of self-reference; the behavioral layer data includes at least one of the target object's recognition efficiency, judgment accuracy, and reaction bias in the emotion judgment process; the physiological layer data includes at least one of the neural oscillation characteristics, nervous system activation level, and peripheral somatic physiological representation feedback in the emotion processing stage.
9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the depression severity index prediction method according to any one of claims 1-8.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the depression severity index prediction method according to any one of claims 1-8.