A method and system for analysis of neurologic pathology features

Abstract

The application discloses a kind of neurology pathological feature analysis method and system, by collecting the multimodal time physiological information and symptom text diary of target individual;Multimodal time physiological information is preprocessed and feature fusion, extract time domain, frequency domain and nonlinear entropy feature, obtain cross-modal physiological state feature vector;Symptom text diary is processed by natural language, generate symptom intensity sequence and emotion label sequence;Based on cross-modal physiological state feature vector, construct supervised and unsupervised two-way analysis model, respectively output individualized symptom risk prediction value and group deviation degree;Through gate recurrent unit, the time consistency adjustment and feature attribution analysis of two-way output are carried out;Finally, integrated neural state assessment report is generated.The application realizes individual-group dual verification under the fusion of multimodal data, significantly improves the robustness, explanation consistency and personalized early warning accuracy of neural state assessment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical artificial intelligence and neurological data analysis, and particularly relates to a neurological pathological feature analysis method and system. BACKGROUND

[0002] Neurological pathological feature analysis is an important direction for evaluating and monitoring the functional state of the nervous system, especially in the early identification and long-term management of common psychosomatic diseases such as anxiety and depression. At present, the research and practice in this field mainly rely on a single type of data source, such as obtaining the subjective symptom description of the patient through a scale or collecting fragmentary physiological indicators using medical equipment. Although this method can reflect part of the state information, it is difficult to fully capture the dynamic interaction between physiological and psychological elements in the development process of the disease. In addition, due to significant individual differences, simply relying on group benchmarks or general models for analysis often cannot accurately adapt to the specific performance of different users, resulting in insufficient sensitivity and relevance of the evaluation results. In practical applications, how to effectively integrate multi-source, heterogeneous time series data and build a robust analysis model that can balance individualized prediction and group distribution reference is still a major challenge for current technology. Therefore, it is urgent to explore new technical approaches to improve the accuracy and practicality of neurological health status management. SUMMARY

[0003] In view of the above problems, the present application provides a neurological pathological feature analysis method and system, which realizes the complementary enhancement of individualized prediction and group benchmark verification by fusing multi-modal time series data and building a supervised and unsupervised dual-path collaborative analysis model, and solves the problem of insufficient reliability of existing neurological state evaluation results.

[0004] To achieve the above purpose, in a first aspect, the present application provides a neurological pathological feature analysis method, comprising:

[0005] Collecting multi-modal time series physiological information of a target individual, the multi-modal time series physiological information including autonomic nervous activity indicators and sleep structure indicators obtained through wearable devices, and simultaneously collecting symptom text diaries actively reported by the user as subjective annotation information;

[0006] Preprocessing and feature fusion of the multi-modal time series physiological information, using a dynamic time warping algorithm to align heterogeneous time series signals, and extracting time domain, frequency domain and nonlinear entropy features to obtain a cross-modal physiological state feature vector in units of individuals;

[0007] Natural language processing of the symptom text diaries, using a pre-training model based on Transformer for symptom entity recognition and sentiment polarity analysis to generate structured symptom intensity sequences and sentiment label sequences;

[0008] Based on the cross-modal physiological state feature vector, a supervised and unsupervised dual-channel analysis model is constructed, the dual-channel analysis model includes a first channel and a second channel, the first channel adopts a long short-term memory network to establish a dynamic mapping relationship between physiological indicators and symptom labels, and outputs an individualized symptom risk prediction value, which is recorded as a first channel prediction result; the second channel adopts a one-class support vector machine to construct an anomaly detection model based on the baseline distribution of the group, and calculates the deviation degree of the current physiological mode relative to the group norm, which is recorded as a second channel prediction result;

[0009] The output results of the dual-channel analysis model are adjusted for time sequence consistency by a gated recurrent unit, when the first channel prediction result and the second channel prediction result significantly differ, a feature attribution analysis based on attention weight is performed, the core physiological dimension causing the difference is identified, and an attribution analysis result is output;

[0010] An integrated neural state evaluation report is generated according to the first channel prediction result, the second channel prediction result and the attribution analysis result.

[0011] In some embodiments, the multi-modal time sequence physiological information is preprocessed and feature fused, a dynamic time warping algorithm is used to align the heterogeneous time sequence signals, and time domain, frequency domain and nonlinear entropy features are extracted, to obtain a cross-modal physiological state feature vector in units of individuals, including:

[0012] The autonomic nervous activity indicators are analyzed by heart rate variability, the SDNN and RMSSD indicators in the time domain features, the low frequency and high frequency power ratio indicators in the frequency domain features, and the autonomic nervous regulation stability indicators in the nonlinear features are extracted by a sample entropy algorithm;

[0013] The sleep structure indicators are analyzed by sleep staging feature extraction, the duration proportions of deep sleep and rapid eye movement sleep are obtained, and the sleep efficiency and the number of awakenings are calculated to obtain sleep quantitative indicators;

[0014] The dynamic time warping algorithm is used to time-align the above-mentioned autonomic nervous feature sequence and sleep feature sequence, and eliminate the time sequence deviation caused by the difference in collection time points;

[0015] The aligned multi-modal features and electroencephalogram rhythm power features are spliced by feature-level fusion, and principal component analysis is used for dimension reduction processing, to finally generate a cross-modal physiological state feature vector.

[0016] In some embodiments, the symptom text diary is subjected to natural language processing, a pre-training model based on Transformer is used for symptom entity recognition and sentiment polarity analysis, to generate structured symptom intensity sequences and sentiment label sequences, including:

[0017] construct an independent dictionary for the field of neurological symptoms, the independent dictionary including a somatization symptom descriptor set, an emotional state keyword set, and a degree adverb mapping table;

[0018] adopt a pre-training language model based on a Transformer architecture as a basic feature extractor to perform context semantic encoding on the symptom text diary and obtain a deep semantic representation vector for each word;

[0019] perform named entity recognition based on the encoded deep semantic representation vector using a conditional random field model to extract specific symptom entities and their modification relationships, the symptom entities including palpitations, headaches, and gastrointestinal discomfort;

[0020] perform sentiment polarity analysis on the identified symptom entities and their modification relationships, calculate a text sentiment tendency score through a multi-head attention mechanism, and generate a quantified symptom intensity value in combination with the degree adverb mapping table;

[0021] sort the symptom intensity values based on timing information, aggregate the symptom intensity values of the same time period into structured records, and generate a symptom intensity sequence and a sentiment label sequence including symptom types, intensity values, and sentiment polarities.

[0022] In some embodiments, based on the cross-modal physiological state feature vector, a supervised and unsupervised dual-path analysis model is constructed, including:

[0023] construct a supervised analysis model framework, use the cross-modal physiological state feature vector as the model input, use the symptom intensity sequence and the sentiment label sequence as the supervised signal, and adopt a deep neural network architecture to establish a mapping relationship from physiological features to symptom manifestations;

[0024] construct an unsupervised analysis model framework, use the cross-modal physiological state feature vector of a large-scale population sample as training data, adopt a density clustering algorithm to establish a feature distribution space of the population norm, and define an abnormal score mechanism based on distance measurement;

[0025] standardize and calibrate the numerical range output by the supervised analysis model and the unsupervised analysis model to ensure the comparability of the results of the two analysis paths;

[0026] When new multi-modal time-series physiological information is collected, an online learning algorithm is used to update the parameters of the supervised analysis model, and an incremental learning algorithm is used to adjust the feature distribution space of the population norm of the unsupervised analysis model.

[0027] In some embodiments, the first path adopts a long short-term memory network to establish a dynamic mapping relationship between physiological indicators and symptom labels, output an individualized symptom risk prediction value, denoted as the first path prediction result, including:

[0028] The cross-modal physiological state feature vector is taken as an input sequence, and a long short-term memory network model based on a gating mechanism is constructed. Through the cooperative control of the forgetting gate, the input gate and the output gate, the long-term time sequence dependence between physiological indicators is captured.

[0029] The symptom intensity sequence and the emotion label sequence are taken as a supervised signal. A time distribution fully connected layer is used to map the hidden state output by the long short-term memory network to a symptom risk space, generating an individualized symptom risk prediction value synchronized with the input sequence.

[0030] The weight distribution of the symptom intensity prediction task and the emotion polarity classification task is adjusted, and the long short-term memory network parameters are updated through the back propagation algorithm.

[0031] A sliding window mechanism is used to process continuous time series data, and the latest individualized symptom risk prediction value is output in real time, denoted as the first prediction result.

[0032] In some embodiments, the second path adopts a one-class support vector machine to build an anomaly detection model based on the baseline distribution of the group, and calculates the deviation of the current physiological mode relative to the group norm, denoted as the second prediction result, including:

[0033] The sample cross-modal physiological state feature vector of a preset group is collected as a training data set, and a physiological norm representing the relatively stable state of the preset group is constructed. The preset group is configured to be a group diagnosed as mild anxiety or somatization disorder through clinical evaluation and in the clinical stable period;

[0034] A one-class support vector machine algorithm is used to estimate the distribution density of the sample cross-modal physiological state feature vector, and a decision hyperplane is found through nonlinear kernel function mapping to define the main feature distribution space of the stable preset group;

[0035] The cross-modal physiological state feature vector of the current user is input into the trained anomaly detection model, and the relative distance of the cross-modal physiological state feature vector of the current user to the decision hyperplane is calculated. The relative distance is quantified as the deviation of the current physiological mode relative to the group stable state norm;

[0036] According to the distribution characteristics of the group stable state norm, the deviation is normalized to generate a second prediction result that can be compared across individuals.

[0037] In some embodiments, the output results of the two-path analysis model are adjusted for time sequence consistency through a gated recurrent unit, including:

[0038] The first prediction result and the second prediction result are aligned and spliced according to the time sequence to form a two-path prediction result sequence as an input feature.

[0039] A gating recurrent unit network is constructed to learn the temporal dependency and complementary characteristics between the two prediction results through update gate and reset gate mechanisms.

[0040] An attention mechanism is used to calculate the weight distribution of historical prediction results, enhancing the influence weight of historical information with high relevance to the current state.

[0041] The two prediction result sequences are adjusted for temporal consistency.

[0042] In some embodiments, when the first prediction result and the second prediction result significantly diverge, an attention weight-based feature attribution analysis is performed to identify the core physiological dimension causing the divergence, and the attribution analysis result is output, including:

[0043] Based on the two prediction result sequences adjusted for temporal consistency, the relative entropy values of the two prediction results in adjacent time windows are calculated, and when the relative entropy of three consecutive time windows all exceeds the adaptive threshold, it is determined to be a state of significant divergence.

[0044] A multi-level feature attribution analysis is started, and at the first level, an integrated gradient algorithm is used to calculate the contribution of each input feature dimension in the long short-term memory network to the current prediction result, generating a first feature importance distribution.

[0045] At the second level, a kernel function-based Shapley value estimation algorithm is used to analyze the marginal contribution of each feature dimension in the support vector machine model to the abnormal score, generating a second feature importance distribution.

[0046] A feature contribution difference matrix is constructed to compare the first feature importance distribution and the second feature importance distribution, and the top K feature dimensions with the largest contribution difference are identified as core divergence features, denoted as the identification result.

[0047] Based on the identification result, an attribution analysis result is generated, including the divergence time period, the core feature dimension list, the contribution difference value, and the clinical significance explanation.

[0048] In some embodiments, an integrated neural state assessment report is generated based on the first prediction result, the second prediction result, and the attribution analysis result, including:

[0049] The symptom risk trend of the first prediction result, the physiological abnormality degree of the second prediction result, and the core feature dimension of the attribution analysis result are weighted and fused.

[0050] According to the consistency degree of the two prediction results, a corresponding assessment mode is selected, including:

[0051] When the two prediction results are consistent, a collaborative confirmation mode is used.

[0052] An attribution explanation mode is adopted when the two-path prediction results are different;

[0053] A structured assessment result is generated, and the assessment result includes a symptom risk level, a physiological abnormality index, a key influencing factor analysis, and a confidence level;

[0054] A personalized intervention suggestion is generated according to the structured assessment result, and the personalized intervention suggestion includes a pre-warning prompt, a lifestyle adjustment suggestion, and professional medical guidance;

[0055] An integrated neural state assessment report is generated according to the structured assessment result and the personalized intervention suggestion.

[0056] In a second aspect, the present application also provides a neurological pathology feature analysis system suitable for the method of the first aspect.

[0057] Unlike the prior art, the above technical solution collects multi-modal time-series physiological information and symptom text diary of a target individual; pre-processes and fuses the multi-modal time-series physiological information, extracts time-domain, frequency-domain and nonlinear entropy features, and obtains a cross-modal physiological state feature vector; performs natural language processing on the symptom text diary to generate a symptom intensity sequence and an emotion label sequence; constructs a supervised and unsupervised two-path analysis model based on the cross-modal physiological state feature vector, and respectively outputs individualized symptom risk prediction values and group deviation; adjusts the two-path outputs for time-series consistency through a gated recurrent unit and performs feature attribution analysis; and finally generates an integrated neural state assessment report. Through the collaborative analysis and time-series consistency adjustment of the supervised and unsupervised two-path models, the present application realizes individual-group dual verification under multi-modal data fusion, and significantly improves the robustness, explanation consistency and personalized early warning accuracy of neural state assessment.

[0058] The above invention content is only a summary of the technical solution of the present application. In order to enable those skilled in the art to more clearly understand the technical solution of the present application, and then implement the content recorded in the specification and drawings, and in order to make the above-mentioned purposes and other purposes, features and advantages of the present application more easily understood, the following will be described in combination with the specific embodiments of the present application and the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0059] The accompanying drawings are only used to show the principles, implementation modes, applications, characteristics and effects of the specific embodiments and other related contents of the present application, and cannot be considered as a limitation of the present application.

[0060] In the drawings of the specification:

[0061] Figure 1 A step schematic diagram of steps S101 to S106 of the analysis method described in the specific embodiments;

[0062] Figure 2 Step S201 to step S204 of the analysis method described in the detailed description are schematically shown in the following figure;

[0063] Figure 3 Step S301 to step S305 of the analysis method described in the detailed description are schematically shown in the following figure. DETAILED DESCRIPTION

[0064] To explain the possible application scenarios, technical principles, specific schemes that can be implemented, purposes and effects that can be achieved, etc. of the present application in detail, the following specific embodiments are combined with the accompanying drawings. The embodiments described herein are only used to more clearly illustrate the technical solutions of the present application, and therefore only serve as examples, and cannot limit the protection scope of the present application.

[0065] In this paper, the term "embodiment" means that the specific features, structures or characteristics described in conjunction with the embodiment can be included in at least one embodiment of the present application. The term "embodiment" appearing at various places in the specification does not necessarily refer to the same embodiment, and does not particularly limit its independence or association with other embodiments. In principle, in the present application, as long as there is no technical contradiction or conflict, each technical feature mentioned in each embodiment can be combined in any way to form a corresponding implementable technical solution.

[0066] Unless otherwise defined, the meaning of the technical terms used herein is the same as that generally understood by those skilled in the art to which the present application belongs; the use of related terms in this paper is only for the purpose of describing specific embodiments, and is not intended to limit the present application.

[0067] In the description of the present application, the phrase "and / or" is a description of the logical relationship between the objects, which means that there can be three relationships, for example, A and / or B, which means that there are three cases: A exists, B exists, and A and B exist at the same time. In addition, the character " / " in this paper generally represents that the associated objects before and after are a "or" logical relationship.

[0068] In the present application, such as "first" and "second", the terms are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual quantity, primary and secondary or order relationship between the entities or operations.

[0069] In the absence of more limitations, in this application, the "include", "contain", "have" or other similar open expressions used in the statements are intended to cover the non-exclusive inclusion, and these expressions do not exclude the presence of other elements in the process, method or product including the elements, so that the process, method or product including a series of elements can not only include those limited elements, but also include other elements not explicitly listed, or also include the elements inherent in such process, method or product.

[0070] As the same understanding in the "Examination Guidelines", in this application, the expressions such as "greater than", "less than", "exceed" are understood as not including the number; the expressions such as "above", "below", "within" are understood as including the number. In addition, in the description of the embodiments of the application, the meaning of "multiple" is more than two (including two), and similar expressions related to "multiple" are also understood in this way, for example, "multiple groups", "multiple times" and the like, unless otherwise explicitly limited.

[0071] In the description of the embodiments of the application, the spatial-related expressions used, such as "center", "longitudinal", "transverse", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "vertical", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", and the like, indicate the orientation or positional relationship based on the orientation or positional relationship shown in the specific embodiments or the drawings, and are only for the convenience of describing the specific embodiments of the application or for the reader to understand, and do not indicate or imply that the indicated device or component must have a particular position, a particular orientation, or be constructed or operated in a particular orientation, therefore cannot be understood as a limitation of the embodiments of the application.

[0072] The processor of the embodiments of the present application can be implemented by hardware, firmware, software or a combination thereof, and can use at least one of circuit, single or multiple Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), Central Processing Unit (CPU), controller, microcontroller, microprocessor, and other physical, biological or chemical structures that can realize the same or equivalent functions as the above-mentioned processors, such as biological neurons, quantum computing units, DNA computing units, etc., so that the processor can execute part or all of the steps or any combination of the steps mentioned in the computer programs or methods of the embodiments of the present application.

[0073] The computer program involved in the embodiments can be stored in a computer device readable storage medium, including but not limited to magnetic disk, magnetic tape, magnetic card, floppy disk, flash memory, optical disc, optical card, read-only memory (ROM), random access memory (RAM), erasable programmable ROM (EPROM) and electrically erasable programmable ROM (EEPROM), etc., and also includes other biological, physical or chemical structures that can realize similar or equivalent functions as the above-mentioned storage media, such as DNA, RNA, protein and other units with information storage ability, etc. In specific embodiments, the storage medium can be one of the above-mentioned medium types, or a combination of the above-mentioned medium types. In different embodiments, the computer program involved in the embodiments can be stored in a single medium in a centralized manner, or stored in multiple media in a distributed manner. The storage medium containing the computer device readable storage medium can be a non-volatile memory or a random access memory. These computer device readable storage media can be built-in in the device, or connected with the device as an external device or part of the external device. In some embodiments, the storage medium with the computer device readable storage medium is deployed locally; in other embodiments, the storage medium can also be deployed remotely from the processor, such as network attached storage accessed via RF circuit or external port and communication network, wherein the communication network can be the Internet, one or more intranets, local area network (LAN), wide area network (WAN), storage area network (SAN) and the like, or appropriate combination thereof, as long as the computer device can access the storage medium. In addition, the computer program involved in the embodiments can be stored in plaintext / ciphertext form, or can be designed as training data, and integrated and reorganized by model training to be implicitly saved in the parameter state of the deep neural network or other machine learning model.

[0074] Please refer to Figure 1 In the first aspect, the embodiment provides a neurological pathology feature analysis method, comprising:

[0075] S101, collecting multi-modal time-series physiological information of a target individual, the multi-modal time-series physiological information comprising autonomic nervous activity indicators and sleep structure indicators obtained by a wearable device, and simultaneously collecting a symptom text diary actively reported by the user as subjective annotation information;

[0076] S102, pre-processing and feature fusion of the multi-modal time-series physiological information, aligning heterogeneous time-series signals by using a dynamic time warping algorithm, and extracting time domain, frequency domain and nonlinear entropy features to obtain a cross-modal physiological state feature vector in units of individuals;

[0077] S103, natural language processing is performed on the symptom text diary, symptom entity recognition and sentiment polarity analysis are performed by using a pre-training model based on a transformer, and a structured symptom intensity sequence and a sentiment label sequence are generated;

[0078] S104, based on the cross-modal physiological state feature vector, a supervised and unsupervised dual-channel analysis model is constructed, the dual-channel analysis model includes a first channel and a second channel, the first channel uses a long short-term memory network to establish a dynamic mapping relationship between physiological indicators and symptom labels, and outputs an individualized symptom risk prediction value, which is denoted as a first channel prediction result; the second channel uses a one-class support vector machine to construct an anomaly detection model based on a group baseline distribution, and calculates a deviation degree of the current physiological mode relative to the group norm, which is denoted as a second channel prediction result;

[0079] S105, the output results of the dual-channel analysis model are adjusted for time sequence consistency by using a gated recurrent unit, when the first channel prediction result and the second channel prediction result significantly differ, a feature attribution analysis based on attention weight is performed, a core physiological dimension causing the difference is identified, and an attribution analysis result is output;

[0080] S106, an integrated neural state evaluation report is generated according to the first channel prediction result, the second channel prediction result and the attribution analysis result.

[0081] In step S101, the target individual refers to a specific user who needs to be evaluated for the neural state. Multimodal time-series physiological information is continuously collected by a wearable device, wherein the autonomic nervous activity indicators can include parameters such as heart rate variability reflecting the balance of sympathetic and parasympathetic nerves, and the sleep structure indicators can include parameters such as sleep stage duration and wake-up frequency quantifying sleep quality. The wearable device is preferably a portable electronic device such as a smart bracelet, a smart watch, etc. capable of continuously monitoring physiological parameters such as heart rate and sleep. The symptom text diary is actively recorded by the user through a mobile terminal application, forming subjective annotation information synchronized in time with the physiological data.

[0082] In step S102, the dynamic time warping algorithm is used to solve the problem of time axis misalignment caused by different sampling frequencies of heterogeneous time-series signals, and the time mapping between signals is realized by finding the optimal path. Time domain feature extraction can include statistical quantities such as mean and variance, frequency domain features can be obtained by Fourier transform to obtain energy distribution in each frequency band, and nonlinear entropy features can be quantified by sample entropy algorithm to quantify the complexity of the signal. Optionally, the multimodal time-series physiological information can also include electroencephalogram rhythm power features obtained by electroencephalogram acquisition devices to further improve the completeness of the state representation. These features are fused to form a cross-modal physiological state feature vector, representing the comprehensive physiological state of the individual in a specific time period.

[0083] In step S103, a pre-trained model based on Transformer is used for semantic understanding of the symptom text. Symptom entity recognition is used to extract keywords describing specific symptoms, and sentiment polarity analysis is used to determine the emotional tendency expressed in the text. By combining degree adverbs and other modifiers, subjective descriptions are converted into quantified symptom intensity values, and finally standardized symptom records with a time sequence structure are generated.

[0084] In the dual-path analysis model constructed in step S104, the first path of the long short-term memory network learns the time sequence dependency between physiological characteristics and symptoms through its gating mechanism, achieving individualized prediction. The second path of the one-class support vector machine maps the features to a high-dimensional space through a kernel function, establishing an anomaly detection boundary based on population data. The two analysis paths are independent and complementary to each other. The first path focuses on individual internal change patterns, and the second path provides a group reference benchmark.

[0085] In step S105, the gating recurrent unit learns the time sequence dependency between the prediction results of the two paths through the update gate and reset gate mechanisms, and initiates feature attribution analysis when significant discrepancies are detected. The attribution method based on attention weights can identify the physiological dimensions that contribute most to the prediction discrepancy, providing an explanation for subsequent analysis. Preferably, the attention weights are calculated based on the correlation between historical prediction results and the current state, which can be achieved through a query key-value attention mechanism or a correlation score algorithm.

[0086] The integrated neural state assessment report generated in step S106 combines the prediction results of the two paths and the attribution analysis results. Preferably, the final assessment result is formed through weighted fusion, and the integrated neural state assessment report not only contains risk level determination, but also provides key factor analysis leading to the assessment, enhancing the explainability and clinical utility of the results.

[0087] The present embodiment fuses multi-modal time sequence physiological information and symptom text diaries, constructs a supervised and unsupervised dual-path analysis model, realizes the collaborative analysis of individualized symptom risk prediction and population anomaly detection, uses a gating recurrent unit for time sequence consistency adjustment, and identifies key physiological dimensions through feature attribution analysis when there is a discrepancy in the results, finally generating an integrated neural state assessment report. The dual-path analysis model of the present embodiment allows the system to consider both individual specificity and population regularity. When the conclusions of the two paths are consistent, they confirm each other, and when there is a discrepancy, they provide a reasonable explanation through attribution analysis, effectively improving the accuracy, robustness and explainability of neural state assessment, and providing reliable technical support for personalized health management.

[0088] Please refer to Figure 2In some embodiments, the multi-modal time physiological information is pre-processed and feature fused, the dynamic time warping algorithm is used to align the heterogeneous time signals, and the time domain, frequency domain and nonlinear entropy features are extracted to obtain the cross-modal physiological state feature vector in units of individuals, including:

[0089] S201, heart rate variability analysis is performed on the autonomic nervous activity index, SDNN and RMSSD indexes in the time domain feature, low frequency and high frequency power ratio indexes in the frequency domain feature, and the autonomic nervous regulation stability index in the nonlinear feature is calculated by the sample entropy algorithm;

[0090] S202, sleep structure index is used for sleep staging feature extraction, the duration ratio of deep sleep period and rapid eye movement period is obtained, and sleep efficiency and wake-up times are calculated to obtain sleep quantitative index;

[0091] S203, the dynamic time warping algorithm is used to time align the above autonomic nervous feature sequence and sleep feature sequence, and eliminate the time sequence deviation caused by the difference in collection time points;

[0092] S204, the aligned multi-modal features and electroencephalogram rhythm power features are spliced by feature level fusion, and principal component analysis is used for dimension reduction processing, and finally the cross-modal physiological state feature vector is generated.

[0093] In step S201, the heart rate variability analysis is realized by R wave detection and segment division on the electrocardiogram signal. The SDNN index is obtained by calculating the standard deviation of all normal sinus intervals, reflecting the overall regulation ability of the autonomic nervous system; the RMSSD index is obtained by calculating the root mean square of adjacent heart rate interval differences, mainly used for evaluating the parasympathetic nerve activity. The frequency domain analysis uses fast Fourier transform to decompose the signal into different frequency bands, the low frequency power (0.04-0.15Hz) is related to the sympathetic nervous activity, the high frequency power (0.15-0.4Hz) reflects the parasympathetic nervous activity, and the ratio of the two can quantify the sympathetic-parasympathetic nerve balance state. The sample entropy algorithm calculates the probability of new pattern generation in time series to evaluate the signal complexity, and the higher the entropy value, the more flexible the autonomic nervous regulation.

[0094] In step S202, sleep staging feature extraction is based on the international sleep medicine society standard, and a machine learning algorithm is used to automatically stage the sleep polygraph signal. The duration ratio of deep sleep period is obtained by calculating the ratio of the sleep time in this period to the total sleep time, reflecting the recovery function of sleep; the rapid eye movement period ratio is obtained by calculating the ratio of the duration of this period to the total sleep time, which is related to emotion regulation and memory consolidation. Sleep efficiency is calculated by the percentage of total sleep time to total bed time, and the number of awakenings can be quantified by detecting the number of micro-awakening events in the electroencephalogram to quantify the continuity of sleep.

[0095] In step S203, the dynamic time warping algorithm finds the minimum cumulative distance path by constructing the distance matrix between the autonomic nervous feature sequence and the sleep feature sequence, using dynamic programming to achieve nonlinear alignment of the heterogeneous time series signals. This algorithm can effectively handle the time series offset problem caused by different measurement devices or physiological process time scale differences, providing time synchronization guarantee for subsequent multi-modal feature fusion.

[0096] The feature-level fusion of step S204 combines the aligned autonomic nervous features, sleep quantitative indicators, and electroencephalogram rhythm power features into a high-dimensional feature vector using vector splicing. Preferably, principal component analysis selects the principal component corresponding to the largest eigenvalue as the new feature dimension by calculating the covariance matrix and its eigenvectors of the features, achieving dimension reduction processing while retaining more than 90% of the original data variance, and finally generating a cross-modal physiological state feature vector suitable for machine learning models.

[0097] This embodiment uses a systematic feature extraction and fusion method to deeply process multi-modal time series physiological information. Time domain indicators such as SDNN and RMSSD and frequency domain indicators such as low-to-high frequency power ratio are extracted using heart rate variability analysis, combined with sample entropy algorithm to quantify autonomic nervous regulation stability; sleep quantitative indicators such as deep sleep and rapid eye movement period proportion are extracted through sleep staging; dynamic time warping algorithm is used to achieve alignment of heterogeneous time series signals, and finally cross-modal physiological state feature vector is generated through feature-level fusion and principal component analysis, effectively solving the problem of multi-source physiological data heterogeneity and time series asynchronization, providing comprehensive, accurate and interpretable feature basis for neural state evaluation, significantly improving the input quality and reliability of subsequent analysis models.

[0098] Please refer to Figure 3 In some embodiments, the symptom text diary is subjected to natural language processing, and a pre-trained model based on Transformer is used for symptom entity recognition and sentiment polarity analysis to generate structured symptom intensity sequences and sentiment label sequences, including:

[0099] S301, an independent dictionary for neurological symptom field is constructed, which includes somatization symptom description word set, emotion state keyword set, and degree adverb mapping table, wherein the degree adverb mapping table is used to quantify description words such as "mild" and "severe" into numerical intensity levels;

[0100] S302, a pre-trained language model based on Transformer architecture is used as a basic feature extractor to encode the context semantics of the symptom text diary and obtain a deep semantic representation vector for each word;

[0101] S303, perform named entity recognition based on the encoded deep semantic representation vector using a conditional random field model, extract specific symptom entities and their modification relationships, including palpitation, headache, and gastrointestinal discomfort;

[0102] S304, perform sentiment polarity analysis on the identified symptom entities and their modification relationships, calculate a text sentiment tendency score through a multi-head attention mechanism, and generate a quantified symptom intensity value in combination with a degree adverb mapping table;

[0103] S305, sort the symptom intensity values based on time sequence information, aggregate the symptom intensity values in the same time period into structured records, and generate a symptom intensity sequence and a sentiment label sequence containing symptom type, intensity value, and sentiment polarity.

[0104] In step S301, an independent dictionary is specially constructed for the field of neurology, the somatization symptom description word set contains common somatic symptom expressions such as palpitation and headache, the emotional state keyword set covers emotional related words such as anxiety and depression, and the degree adverb mapping table maps descriptive words such as "mild", "moderate", and "severe" to numerical intensity levels, realizing the objective quantification of subjective description.

[0105] In step S302, the pre-trained language model learns medical semantic representation by performing self-supervised task such as mask language modeling on large medical text corpus. The model calculates the correlation weight between words through self-attention mechanism, dynamically aggregates context information, and encodes each word into a deep representation vector containing rich semantic relationships, providing high-quality semantic feature basis for subsequent symptom analysis.

[0106] In step S303, the conditional random field model preferably uses BIO tagging strategy, uses deep semantic representation vector as feature input, and through joint optimization of transition probability and state feature function, accurately identifies the boundary position of symptom entity, learns the grammatical dependency relationship between symptom words and modifier words, and realizes accurate symptom entity extraction.

[0107] In step S304, the multi-head attention mechanism can calculate the sentiment features of different levels of text in parallel, obtain a comprehensive sentiment tendency score by weighting and fusing the outputs of each attention head, and then convert the descriptive symptom intensity into a quantifiable numerical indicator in combination with the degree adverb mapping table.

[0108] In step S305, based on the timestamp information, the quantified symptom intensity value is sorted in time sequence, and multiple symptom records in the same time period are aggregated into unified structured data, and finally a standardized output sequence containing symptom type, intensity value and sentiment polarity is generated.

[0109] The embodiment realizes the structured processing of subjective symptom text by combining professional dictionaries with deep learning technology, retains the richness of natural language expression, and converts it into machine-processable quantitative indicators, providing standardized text feature input for subsequent multi-modal data analysis, and effectively solving the semantic gap problem between subjective reports and objective physiological data.

[0110] In some embodiments, based on the cross-modal physiological state feature vector, a supervised and unsupervised dual-path analysis model is constructed, including:

[0111] A supervised analysis model framework is constructed, taking the cross-modal physiological state feature vector as the model input, taking the symptom intensity sequence and the emotion label sequence as the supervised signal, and using a deep neural network architecture to establish the mapping relationship from physiological features to symptom performance.

[0112] An unsupervised analysis model framework is constructed, taking the cross-modal physiological state feature vector of a large-scale population sample as the training data, using a density clustering algorithm to establish the feature distribution space of the population norm, and defining an abnormal score mechanism based on distance measurement.

[0113] The numerical range of the output of the supervised analysis model and the unsupervised analysis model is standardized and calibrated to ensure the comparability of the results of the two analysis paths.

[0114] And when new multi-modal time-series physiological information is collected, the supervised analysis model uses an online learning algorithm to update the parameters, and the unsupervised analysis model uses an incremental learning algorithm to adjust the feature distribution space of the population norm.

[0115] In the embodiment, the supervised analysis model framework learns the complex mapping relationship between the cross-modal physiological state feature vector and the symptom intensity sequence and emotion label sequence through a deep neural network architecture. This framework can use recurrent neural networks or time convolutional networks to capture the temporal dependence of physiological features and establish a prediction model from multi-modal physiological data to subjective symptom performance.

[0116] The unsupervised analysis model framework uses a density clustering algorithm to model the distribution of the cross-modal physiological state feature vector of a large-scale population sample, and uses Gaussian mixture models or DBSCAN algorithms to establish the feature distribution space of the population norm. The abnormal score mechanism is based on Mahalanobis distance or local outlier factor algorithm to calculate the deviation of individual feature vector from the population distribution center.

[0117] The standardization and calibration process uses min-max scaling or z-score standardization method to map the prediction probability value output by the supervised analysis model and the abnormal score output by the unsupervised analysis model to a unified numerical interval, ensuring that the output results of the two analysis paths are directly comparable.

[0118] The online learning algorithm can use optimization methods such as stochastic gradient descent or adaptive M-estimation, so that the supervised analysis model can dynamically adjust the model parameters according to the newly collected multi-modal time-series physiological information; the incremental learning algorithm can make the unsupervised analysis model adapt to the gradual changes of the group norm characteristics by updating the clustering center or adjusting the distribution parameters.

[0119] In this embodiment, the individualized prediction and group benchmark analysis are organically combined through the collaborative construction of the supervised and unsupervised dual-path analysis models. The two analysis paths are independent of each other and complementary to each other. The supervised path focuses on learning the physiological-symptom mapping rules specific to the individual, and the unsupervised path provides an objective reference benchmark based on the group distribution, thereby providing a reliable technical foundation for the subsequent fusion and decision-making of the dual-path results.

[0120] In some embodiments, the first path uses a long short-term memory network to establish a dynamic mapping relationship between physiological indicators and symptom labels, and outputs individualized symptom risk prediction values, denoted as the first path prediction result, including:

[0121] The cross-modal physiological state feature vector is taken as an input sequence, and a long short-term memory network model based on a gating mechanism is constructed. Through the collaborative control of the forget gate, the input gate and the output gate, the long-term time-series dependency relationship between physiological indicators is captured.

[0122] The symptom intensity sequence and the emotion label sequence are taken as supervised signals, and a time distribution fully connected layer is used to map the hidden state output by the long short-term memory network to the symptom risk space, thereby generating individualized symptom risk prediction values synchronized with the input sequence.

[0123] The weight distribution of the symptom intensity prediction task and the emotion polarity classification task is adjusted, and the long short-term memory network parameters are updated through the back propagation algorithm.

[0124] A sliding window mechanism is used to process continuous time-series data, and the latest individualized symptom risk prediction value is output in real time, denoted as the first path prediction result.

[0125] In this embodiment, the long short-term memory network processes the cross-modal physiological state feature vector sequence through its unique gating mechanism. The forget gate controls the retention degree of historical information, the input gate adjusts the input amount of new information, and the output gate determines the output of the current hidden state. The three work together to capture the long-term time-series dependency relationship between physiological indicators.

[0126] The time distribution fully connected layer independently maps the hidden state of each time step of the long short-term memory network, converts high-dimensional features to the symptom risk space, generates individualized symptom risk prediction values synchronized with the input physiological data in time, and realizes end-to-end time-series prediction.

[0127] In the multi-task learning framework, the loss contribution ratio of the two tasks is automatically adjusted by learnable weight parameters. Specifically, the system calculates the mean square error loss of symptom intensity prediction and the cross-entropy loss of sentiment polarity classification, and obtains the total loss function by weighted summation. The back propagation algorithm calculates the gradient layer by layer from the output layer to the input layer based on the chain rule, first calculates the partial derivative of the total loss with respect to the network output node, then propagates the gradient along the time dimension by the time reverse propagation algorithm, and finally updates the weight parameters of all gate units and fully connected layers in the long short-term memory network using the optimizer, realizing the collaborative optimization of the two tasks and improving the representation ability of the model for complex symptom patterns.

[0128] The sliding window mechanism adopts a fixed length time window to slide on continuous time series data, and only processes the data segment within the window each time, ensuring that the model can process streaming data in real time and output the latest prediction results, meeting the needs of real-time clinical monitoring.

[0129] This embodiment effectively captures the dynamic mapping relationship between physiological indicators and symptom manifestations through the excellent time series modeling capability of the long short-term memory network. Combined with multi-task learning and real-time processing mechanism, the accuracy of individualized prediction is improved, and the timeliness of the system in actual application is guaranteed, providing reliable technical support for personalized health management.

[0130] In some embodiments, the second path adopts a one-class support vector machine to build an abnormality detection model based on the group baseline distribution, calculates the deviation of the current physiological pattern relative to the group norm, denoted as the second path prediction result, including:

[0131] The sample cross-modal physiological state feature vector of a preset group is collected as the training data set, and a physiological norm representing the relatively stable state of the preset group is constructed. The preset group is configured to be a group diagnosed as mild anxiety or somatization disorder through clinical evaluation and in the clinical stable period;

[0132] A one-class support vector machine algorithm is used to estimate the distribution density of the sample cross-modal physiological state feature vector, and a decision hyperplane is found through nonlinear kernel function mapping to define the main feature distribution space of the stable preset group;

[0133] The cross-modal physiological state feature vector of the current user is input into the trained abnormality detection model, the relative distance of the cross-modal physiological state feature vector of the current user to the decision hyperplane is calculated, and the relative distance is quantified as the deviation of the current physiological pattern relative to the group stable norm;

[0134] The deviation is normalized according to the distribution characteristics of the group stable norm to generate the second path prediction result that can be compared across individuals.

[0135] In this embodiment, by continuously collecting multi-modal physiological data of the preset group in a stable state, after feature extraction and fusion processing, a clustering algorithm (such as K-means) or a Gaussian mixture model is used to model the feature distribution, and a physiological norm representing the stable state of the group is constructed. The norm can accurately depict the typical physiological feature distribution pattern of the stable state of the patient.

[0136] One class support vector machine maps samples to a high-dimensional feature space through a kernel trick, and finds an optimal hyperplane in the space. Specifically, the similarity between samples is calculated by using a Gaussian kernel function, and the distance from the hyperplane to the origin is maximized to define the boundary of normal data. The decision hyperplane divides the feature space into normal and abnormal regions, and the normal samples are as far as possible contained in the hyperplane, while the abnormal samples fall outside the hyperplane, thereby accurately defining the feature distribution space of the stable group.

[0137] The calculation of relative distance is based on the decision function of support vector machine. After mapping the current user features to a high-dimensional space through the kernel function, the perpendicular distance between the mapping point and the optimal hyperplane is calculated. The distance value is processed by a sign function, and the positive sign indicates that it is on the normal side of the hyperplane, and the negative sign indicates that it is on the abnormal side. The greater the absolute value of the distance, the greater the degree of deviation from the group norm, thereby realizing the quantitative conversion from geometric distance to abnormality degree.

[0138] The normalization processing adopts a standardization method based on the statistical characteristics of the group distribution, such as z-score standardization or normalization based on quantile, to convert the original distance value into a deviation score with uniform dimension and comparability, ensuring that the results of different individuals can be compared fairly.

[0139] The unsupervised anomaly detection model constructed by one class support vector machine in this embodiment provides an objective evaluation dimension based on group benchmark for the system. This model does not rely on individual labeled data, but learns the feature distribution pattern of the stable group to identify anomalies, effectively supplementing the individual bias problem that may exist in the supervised path, and enhancing the robustness and reliability of the overall system.

[0140] In some embodiments, the output results of the two-path analysis model are adjusted for temporal consistency by a gated recurrent unit, including:

[0141] Aligning and splicing the first path prediction result and the second path prediction result according to the time sequence to form a two-path prediction result sequence as an input feature;

[0142] A gated recurrent unit network is constructed to learn the temporal dependence and complementary characteristics between the two-path prediction results through update gate and reset gate mechanisms;

[0143] The attention mechanism calculates the weight distribution of the historical prediction results to enhance the influence weight of the historical information with high relevance to the current state.

[0144] The output is a two-path prediction result sequence adjusted for temporal consistency.

[0145] In this embodiment, the two-path prediction result sequence is formed by aligning the timestamps of the first and second prediction results, ensuring that the outputs of the two analysis paths are completely synchronized in the time dimension, providing an accurate data foundation for subsequent temporal analysis.

[0146] The gated recurrent unit network updates the retention degree of historical information and resets the input proportion of new information regulated by the gate, both of which work together to learn the complex temporal dependency between the two-path prediction results, effectively capturing the dynamic association patterns and complementary characteristics between the outputs of the two analysis paths.

[0147] The attention mechanism dynamically evaluates the importance of prediction results at different historical time steps by calculating the attention weight of each historical time step. The higher the weight, the greater the contribution of historical information to the current adjustment process, thereby focusing and enhancing key historical information.

[0148] The temporal consistency adjustment process integrates the network state of the gated recurrent unit and the attention weight to smooth and correct the two-path prediction results, eliminating outliers caused by data noise or model fluctuations, and finally outputting an adjusted sequence with better temporal consistency and reliability.

[0149] This embodiment combines gated recurrent units and attention mechanisms to achieve intelligent temporal integration of two-path prediction results, not only preserving the advantages of each analysis path, but also enhancing the stability and reliability of the results through temporal consistency processing, providing a more accurate and consistent input foundation for subsequent decision analysis.

[0150] In some embodiments, when the first and second prediction results significantly diverge, an attention weight-based feature attribution analysis is performed to identify the core physiological dimensions causing the divergence, and the attribution analysis results are output, including:

[0151] Based on the two-path prediction result sequence adjusted for temporal consistency, the relative entropy values of the two-path prediction results within adjacent time windows are calculated, and when the relative entropy of three consecutive time windows exceeds the adaptive threshold, it is determined to be in a state of significant divergence.

[0152] Start multi-level feature attribution analysis, use integrated gradient algorithm to calculate the contribution of each input feature dimension in long short-term memory network to the current prediction result at the first level, and generate the first feature importance distribution.

[0153] At the second level, the Shapley value estimation algorithm based on the kernel function is adopted to analyze the marginal contribution of each feature dimension to the abnormal score in a support vector machine model, and a second feature importance distribution is generated;

[0154] A feature contribution difference matrix is constructed to compare the first feature importance distribution and the second feature importance distribution, and the top K feature dimensions with the largest contribution difference are identified as core divergent features, which are denoted as the identification result;

[0155] Based on the identification result, an attribution analysis result is generated, which includes the divergence time period, the list of core feature dimensions, the contribution difference value, and the clinical significance explanation.

[0156] In this embodiment, first, the double-path prediction results of adjacent time windows are converted into probability distributions, and then the relative entropy values between the two distributions are calculated. The adaptive threshold is dynamically adjusted according to the statistical characteristics of the historical relative entropy values. When the entropy values of three consecutive time windows exceed the threshold, it indicates that there is a persistent and significant divergence between the double-path prediction results.

[0157] The integrated gradient algorithm first constructs a linear interpolation path between the input features and the baseline, then calculates the prediction output gradient of each point on the path, and finally integrates these gradients to obtain the contribution score of each feature dimension. The integrated gradient algorithm quantifies the contribution of each physiological feature dimension to the long short-term memory network prediction result by calculating the gradient integral of the input features with respect to the prediction output, and eliminates the gradient saturation problem through path integration, ensuring the accuracy of feature importance evaluation.

[0158] The Shapley value estimation algorithm based on the kernel function generates feature subsets by random sampling, calculates the marginal contribution of each feature value using kernel function weighting, and finally obtains the Shapley value estimate by weighted averaging. Through the kernel technique, the exponential computational complexity is reduced to polynomial, realizing efficient calculation in large-scale feature space.

[0159] Preferably, the construction of the feature contribution difference matrix adopts the Euclidean distance measurement, and the difference between the first feature importance distribution vector and the second feature importance distribution vector is calculated, and then the difference vector is standardized. The top K features with the largest difference are selected by sorting, which represent the most significant divergence sources between the two-path models.

[0160] The clinical significance explanation can be realized through the knowledge graph, which associates and maps the identified core features with the pathophysiological mechanisms in the medical knowledge base, and generates targeted explanations and instructions combining clinical guidelines and expert experience, providing data-driven analysis results while ensuring the clinical rationality and operability of the conclusions.

[0161] The embodiment can not only identify the key feature dimensions leading to the divergence, but also provide clinical explanations, enhance interpretability and practicality, and provide important support for precision medicine by analyzing the reasons for the divergence of the two paths through multi-level feature attribution analysis method from the two dimensions of mathematical calculation and clinical interpretation.

[0162] In some embodiments, an integrated neural state assessment report is generated according to the first path prediction result, the second path prediction result, and the attribution analysis result, including:

[0163] The symptom risk trend of the first path prediction result, the physiological abnormality degree of the second path prediction result, and the core feature dimension of the attribution analysis result are weighted and fused;

[0164] According to the consistency degree of the two-path prediction results, a corresponding assessment mode is selected, including:

[0165] When the two-path prediction results are consistent, a synergistic confirmation mode is adopted;

[0166] When the two-path prediction results are divergent, an attribution explanation mode is adopted;

[0167] A structured assessment result is generated, and the assessment result includes a symptom risk level, a physiological abnormality index, a key influencing factor analysis, and a confidence level;

[0168] According to the structured assessment result, individualized intervention suggestions are generated, including early warning prompts, lifestyle adjustment suggestions, and professional medical guidance;

[0169] According to the structured assessment result and the individualized intervention suggestions, an integrated neural state assessment report is generated.

[0170] In the embodiment, the weighted fusion process adopts an adaptive weight distribution mechanism, which dynamically adjusts the weight proportion of each input item according to the confidence and historical accuracy of the two-path prediction results. The symptom risk trend mainly reflects the change of subjective feeling, the physiological abnormality degree reflects the deviation of objective indicators, and the core feature dimension provides in-depth analysis of the reasons for the divergence. The three are weighted and fused to form the basis of comprehensive assessment.

[0171] The assessment mode selection is based on the consistency measurement result of the two-path prediction results. When the two-path prediction results are consistent at the statistical significance level, the synergistic confirmation mode is adopted, emphasizing the credibility and stability of the results. When there is a significant divergence, the attribution explanation mode is enabled, focusing on analyzing the reasons for the divergence and providing reasonable explanations.

[0172] The generation of the structured assessment result adopts a multi-dimensional scoring system, the symptom risk level is determined according to the size and change trend of the predicted value, the physiological abnormality index is calculated based on the degree of deviation from the population norm, the key influencing factor analysis extracts the core characteristics in the attribution result, and the confidence is comprehensively considered in terms of data quality, model performance and result consistency.

[0173] The generation of the personalized intervention suggestion combines clinical guidelines and individual characteristics, the early warning prompt provides immediate warning for high-risk states, the lifestyle adjustment suggestion provides specific improvement schemes based on the identified influencing factors, and the professional medical guidance recommends corresponding medical intervention measures according to the severity and clinical indications.

[0174] The integrated neural state assessment report adopts a standardized template, organically integrates the structured assessment result and the personalized suggestion, ensures the scientificity and integrity of the report, and ensures the clinical practicability and operability. The report generation process also considers the understanding ability and information demand of different user groups, and adopts a hierarchical presentation mode to enhance readability.

[0175] The embodiment realizes the complete transformation from multi-source data to personalized report through systematic information integration and intelligent decision mechanism, not only provides comprehensive state assessment, but also gives specific action guidelines, truly realizes the seamless connection of assessment and intervention, and provides effective technical support for precision health management.

[0176] In the second aspect, the embodiment also provides a neurological pathological feature analysis system suitable for the method of the first aspect.

[0177] By adopting the above technical solutions, the present application is different from the prior art and has the following beneficial effects: by fusing multi-modal time-series physiological information and symptom text diary, a supervised and unsupervised dual-path analysis model is constructed, the supervised analysis path uses a long short-term memory network to establish individualized symptom risk prediction, the unsupervised analysis path uses a one-class support vector machine to construct a group anomaly detection model, and the dual-path cooperation realizes dual consideration of individual specificity and group regularity; time-series consistency adjustment is performed through a gated recurrent unit, and when the results are different, feature attribution analysis based on attention weight is started, which significantly improves the accuracy and robustness of state assessment. The finally generated integrated neural state assessment report deeply integrates the symptom risk trend, physiological abnormality degree and attribution analysis result, and provides personalized intervention suggestions, realizes a complete technical closed loop from multi-source data collection to precision health management, and provides reliable technical support for early screening and individualized intervention of neurological diseases.

[0178] Finally, it should be noted that the above embodiments have been described in the specification and drawings of the application, but this does not limit the patent protection scope of the application. Any equivalent structure or equivalent process replacement or modification based on the essential concept of the application, using the content described in the specification and drawings of the application, and directly or indirectly implementing the technical solutions of the above embodiments in other related technical fields, etc., are all included in the patent protection scope of the application.

Claims

1. A method of analyzing a neurological pathology profile, characterized by, The method comprises the following steps: Collecting multi-modal time-series physiological information of a target individual, the multi-modal time-series physiological information comprising autonomic nervous activity indicators and sleep structure indicators acquired through a wearable device, and simultaneously collecting a symptom text diary actively reported by the user as subjective annotation information; Preprocessing and feature fusion of the multi-modal time-series physiological information, alignment of heterogeneous time-series signals by using a dynamic time warping algorithm, extraction of time-domain, frequency-domain and nonlinear entropy features, and obtaining a cross-modal physiological state feature vector in units of individuals; Natural language processing of the symptom text diary, symptom entity recognition and sentiment polarity analysis by using a pre-training model based on a Transformer, and generation of a structured symptom intensity sequence and a sentiment label sequence; Based on the cross-modal physiological state feature vector, a supervised and unsupervised dual-path analysis model is constructed, the dual-path analysis model comprising a first path and a second path, the first path using a long short-term memory network to establish a dynamic mapping relationship between physiological indicators and symptom labels, and outputting an individualized symptom risk prediction value, denoted as a first path prediction result; The second path uses a one-class support vector machine to construct an abnormality detection model based on a group baseline distribution, and calculates a deviation degree of the current physiological mode relative to the group norm, denoted as a second path prediction result; The output results of the dual-path analysis model are adjusted for time-series consistency by using a gated recurrent unit, when the first path prediction result and the second path prediction result significantly differ, a feature attribution analysis based on attention weights is performed to identify the core physiological dimension causing the difference, and an attribution analysis result is outputted; An integrated neural state assessment report is generated according to the first path prediction result, the second path prediction result and the attribution analysis result; Wherein, based on the cross-modal physiological state feature vector, a supervised and unsupervised dual-path analysis model is constructed, comprising: A supervised analysis model framework is constructed, the cross-modal physiological state feature vector is taken as the model input, the symptom intensity sequence and the sentiment label sequence are taken as the supervision signal, and a deep neural network architecture is used to establish a mapping relationship from physiological features to symptom manifestations; An unsupervised analysis model framework is constructed, the cross-modal physiological state feature vector of a large-scale group sample is taken as the training data, a density clustering algorithm is used to establish a feature distribution space of the group norm, and a distance measurement-based anomaly score mechanism is defined; The numerical ranges outputted by the supervised analysis model and the unsupervised analysis model are standardized and calibrated to ensure the comparability of the results of the two analysis paths; When new multi-modal time-series physiological information is collected, an online learning algorithm is used to update the parameters of the supervised analysis model, and an incremental learning algorithm is used to adjust the feature distribution space of the group norm of the unsupervised analysis model.

2. The method of analysis of neurologic pathologic features according to claim 1, characterized in that, The preprocessing and feature fusion of the multi-modal time-series physiological information, the alignment of heterogeneous time-series signals by using a dynamic time warping algorithm, the extraction of time-domain, frequency-domain and nonlinear entropy features, and the obtaining of a cross-modal physiological state feature vector in units of individuals, comprise: The heart rate variability analysis is performed on the autonomic nervous activity indicators, SDNN and RMSSD indicators in the time domain characteristics, low-frequency and high-frequency power ratio indicators in the frequency domain characteristics, and the autonomic nervous regulation stability indicators in the non-linear characteristics are calculated through the sample entropy algorithm; The sleep structure indicators are subjected to sleep staging feature extraction, the duration proportions of deep sleep and rapid eye movement sleep are obtained, and the sleep efficiency and the number of awakenings are calculated to obtain sleep quantitative indicators; The dynamic time warping algorithm is used to time-align the autonomic nervous feature sequence and the sleep feature sequence, so as to eliminate the time sequence deviation caused by the difference in collection time points; The aligned multi-modal features and the electroencephalogram rhythm power features are spliced through feature-level fusion, and principal component analysis is used for dimension reduction processing, and finally the cross-modal physiological state feature vector is generated.

3. The method of analysis of neurologic pathologic features according to claim 1, wherein, The symptom text diary is subjected to natural language processing, a pre-training model based on Transformer is used for symptom entity recognition and sentiment polarity analysis, and structured symptom intensity sequences and sentiment label sequences are generated, including: An independent dictionary for the field of neurological symptoms is constructed, which includes a somatization symptom description word set, an emotional state keyword set, and a degree adverb mapping table; A pre-training language model based on the Transformer architecture is used as a basic feature extractor to encode the context semantics of the symptom text diary and obtain a deep semantic representation vector for each word; Based on the encoded deep semantic representation vector, a conditional random field model is used for named entity recognition to extract specific symptom entities and their modification relationships, including palpitations, headaches, and gastrointestinal discomfort; The sentiment polarity analysis is performed on the recognized symptom entities and their modification relationships, the text sentiment tendency score is calculated through a multi-head attention mechanism, and the quantitative symptom intensity value is generated in combination with the degree adverb mapping table; Based on the timing information, the symptom intensity values are sorted, the symptom intensity values in the same time period are aggregated into structured records, and the symptom intensity sequences and sentiment label sequences containing symptom types, intensity values, and sentiment polarities are generated.

4. The method of analysis of neurologic pathologic features according to claim 1, characterized in that, The first path adopts a long short-term memory network to establish a dynamic mapping relationship between physiological indicators and symptom labels, outputs individualized symptom risk prediction values, and is recorded as the first path prediction result, including: A long short-term memory network model based on a gating mechanism is constructed with the cross-modal physiological state feature vector as the input sequence, and through the cooperative control of the forget gate, the input gate and the output gate, the long-term time sequence dependence relationship between physiological indicators is captured; The symptom intensity sequence and the sentiment label sequence are used as a supervision signal, and a time-distributed fully connected layer is used to map the hidden state output by the long short-term memory network to the symptom risk space, generating individualized symptom risk prediction values synchronized with the input sequence; The weight distribution of the symptom intensity prediction task and the sentiment polarity classification task is adjusted, and the long short-term memory network parameters are updated through the back propagation algorithm; The sliding window mechanism is used to process continuous time series data, and the latest individualized symptom risk prediction value is output in real time, which is recorded as the first path prediction result.

5. The method of analysis of neuropathological features according to claim 1, characterized in that, The second path adopts a one-class support vector machine to build an anomaly detection model based on a group baseline distribution, calculates a deviation degree of a current physiological pattern relative to a group norm, and records the deviation degree as a second path prediction result, including: Collect sample cross-modal physiological state feature vectors of a preset group as training data sets, build a physiological norm representing a relatively stable state of the preset group, and the preset group is configured to be a group diagnosed as mild anxiety or somatization disorder through clinical evaluation and in a clinical stable period; Distribute density estimation is performed on the sample cross-modal physiological state feature vectors by using a one-class support vector machine algorithm, a decision hyperplane is found by mapping through a nonlinear kernel function, and the main feature distribution space of the preset group is defined; The cross-modal physiological state feature vector of the current user is input into the trained anomaly detection model, the relative distance of the cross-modal physiological state feature vector of the current user to the decision hyperplane is calculated, and the relative distance is quantified as a deviation degree of the current physiological pattern relative to the group stable state norm; The deviation degree is normalized according to the distribution characteristics of the group stable state norm to generate a second path prediction result that can be compared across individuals.

6. The method of analysis of neurologic pathologic features according to claim 1, wherein, The output results of the double-path analysis model are adjusted for time sequence consistency by using a gated recurrent unit, including: The first path prediction result and the second path prediction result are aligned and spliced according to the time sequence to form a double-path prediction result sequence as an input feature; A gated recurrent unit network is constructed to learn the time sequence dependence and complementary characteristics between the double-path prediction results through update gates and reset gates; An attention mechanism is used to calculate the weight distribution of the historical prediction results to enhance the influence weight of the historical information with high relevance to the current state; The double-path prediction result sequence adjusted for time sequence consistency is output.

7. The method of analysis of neuropathological features according to claim 1, characterized in that, When the first path prediction result and the second path prediction result significantly differ, feature attribution analysis based on attention weights is performed to identify the core physiological dimensions causing the difference, and an attribution analysis result is output, including: Based on the double-path prediction result sequence adjusted for time sequence consistency, the relative entropy values of the double-path prediction results in adjacent time windows are calculated, and when the relative entropy values of three consecutive time windows all exceed an adaptive threshold, it is determined that a significant difference state exists; A multi-level feature attribution analysis is started, an integrated gradient algorithm is used at the first level to calculate the contribution of each input feature dimension in the long short-term memory network to the current prediction result, and a first path feature importance distribution is generated; At the second level, a kernel function-based Shapley value estimation algorithm is used to analyze the marginal contribution of each feature dimension in the one-class support vector machine model to the anomaly score, and a second path feature importance distribution is generated; A feature contribution difference matrix is constructed to compare the first path feature importance distribution and the second path feature importance distribution, identify the top K feature dimensions with the largest contribution difference as core difference features, and record the top K feature dimensions as an identification result; An attribution analysis result is generated based on the identification result, and the attribution analysis result includes a difference time period, a core feature dimension list, a contribution difference value, and a clinical significance explanation.

8. The method of analysis of neuropathological features according to claim 1, characterized in that, An integrated neural state assessment report is generated based on the first path prediction result, the second path prediction result, and the attribution analysis result, including: The symptom risk trend of the first path prediction result, the physiological abnormality degree of the second path prediction result, and the core feature dimension of the attribution analysis result are weighted and fused; According to the consistency degree of the double-path prediction result, a corresponding evaluation mode is selected, including: When the double-path prediction results are consistent, a collaborative confirmation mode is adopted; When the double-path prediction results are inconsistent, an attribution explanation mode is adopted; A structured evaluation result is generated, including a symptom risk level, a physiological abnormality index, a key influencing factor analysis, and a confidence level; According to the structured evaluation result, a personalized intervention suggestion is generated, including a warning prompt, a lifestyle adjustment suggestion, and professional medical guidance; According to the structured evaluation result and the personalized intervention suggestion, the integrated neural state evaluation report is generated.

9. A system for analyzing neurologic pathology features, the system comprising: Suitable for the method of any one of claims 1 to 8.

Images