Electromagnetic personalized treatment parameter generation method and system

Abstract

The invention belongs to the technical field of electromagnetic stimulation treatment, and discloses an electromagnetic personalized treatment parameter generation method and system, and the method comprises the steps: obtaining a multi-source physiological signal of a patient in an electromagnetic stimulation treatment process, obtaining the personal information and medical history information of the patient, carrying out the preprocessing of the multi-source physiological signal, and carrying out the processing of the multi-source physiological signal; extracting time-domain features, frequency-domain features and nonlinear features, combining the time-domain features, the frequency-domain features and the nonlinear features with personal information and medical history information of the patient to form a physiological response feature vector, inputting the physiological response feature vector into a preset physiological response evaluation model, and calculating to obtain a current physiological state evaluation result of the patient. Inputting the current physiological state evaluation result and the current treatment stage information of the patient into a preset parameter adjustment model to generate an electromagnetic parameter adjustment scheme; through the treatment strategy chain selected based on the data information of the to-be-migrated data and the treatment demand information, data treatment is carried out while data migration is carried out on the to-be-migrated data, and the treatment efficiency during data migration is improved.

Description

Technical Field

[0001] The present application relates to the technical field of electromagnetic stimulation therapy, and in particular, to a method and system for generating electromagnetic personalized therapy parameters. Background Art

[0002] Electromagnetic stimulation therapy, particularly transcranial magnetic stimulation (TMS), is a non-invasive physical therapy that is widely used clinically to treat a variety of neuropsychiatric disorders, including depression. In traditional hospital or clinic treatment environments, electromagnetic stimulation therapy typically follows a standardized treatment plan, or the clinician sets the stimulation parameters (such as stimulation intensity, frequency, pulse pattern, etc.) based on the patient's initial assessment and their own experience. These parameters often remain fixed throughout the treatment course or are only adjusted to a limited extent at different stages of treatment.

[0003] In recent years, in order to improve the personalization and efficacy of electromagnetic stimulation therapy, researchers have begun to explore the combination of patients' physiological information to guide treatment. In the existing technology, even if physiological data is collected, the adjustment of parameters often depends on the pre-treatment evaluation or the results of staged clinical evaluations, and fails to fully utilize the real-time dynamic changes in the patient's status during the treatment process. Because the patient's physiological response may change rapidly and complexly during electromagnetic stimulation treatment, the existing fixed or lagged adjustment mode is difficult to capture these changes, thereby limiting the personalization of the treatment and further improvement of the effect. In addition, the existing parameter adjustment method lacks sufficient sophistication and personalization, and fails to integrate the individual differences of patients, thereby limiting the further optimization of the effect of electromagnetic stimulation treatment.

[0004] Therefore, in order to solve the technical problems that the existing electromagnetic parameter adjustment method fails to fully utilize the real-time dynamic change information of the patient's status during treatment and fails to integrate the individual differences of patients, an electromagnetic personalized treatment parameter generation method and system is urgently needed. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for generating electromagnetic personalized treatment parameters. After evaluating the patient's current physiological state through the patient's multi-source physiological signals, personal information and medical history information, an electromagnetic parameter adjustment plan is generated in combination with the patient's current treatment stage information. This solves the problem that the existing electromagnetic parameter adjustment method fails to fully utilize the real-time dynamic change information of the patient's state during the treatment process and fails to integrate the individual differences of the patients. It adapts to the changes in the patient's physiological state and the individual differences of the patients during the treatment process, improves the level of treatment personalization, and optimizes the treatment effect.

[0006] In a first aspect, the present application provides a method for generating electromagnetic personalized treatment parameters, comprising:

[0007] Acquiring multi-source physiological signals of a patient during electromagnetic stimulation treatment, as well as obtaining personal information and medical history information of the patient;

[0008] Preprocessing the multi-source physiological signals to extract time domain features, frequency domain features, and nonlinear features, and combining them with the patient's personal information and medical history information to form a physiological response feature vector;

[0009] Inputting the physiological response feature vector into a preset physiological response evaluation model to calculate and obtain an evaluation result of the patient's current physiological state;

[0010] The current physiological state assessment result and the patient's current treatment stage information are input into a preset parameter adjustment model to generate an electromagnetic parameter adjustment plan.

[0011] The electromagnetic personalized treatment parameter generation method provided in this application can generate electromagnetic treatment parameters. After evaluating the patient's current physiological state through the patient's multi-source physiological signals, personal information and medical history information, an electromagnetic parameter adjustment plan is generated in combination with the patient's current treatment stage information. This solves the problem that the existing electromagnetic parameter adjustment method fails to fully utilize the real-time dynamic change information of the patient's state during the treatment process and fails to integrate the individual differences of the patients. It adapts to the changes in the patient's physiological state and the individual differences of the patients during the treatment process, improves the level of treatment personalization, and optimizes the treatment effect.

[0012] Optionally, preprocessing the multi-source physiological signals to extract time domain features, frequency domain features, and nonlinear features, which are combined with the patient's personal information and medical history information to form a physiological response feature vector, including:

[0013] performing filtering processing on the multi-source physiological signals to obtain filtered multi-source physiological signals;

[0014] Extracting features from the filtered multi-source physiological signals to obtain time domain feature data, frequency domain feature data, and nonlinear feature data;

[0015] According to the personal information and medical history information of the patient, the time domain feature data, the frequency domain feature data and the nonlinear feature data are normalized and combined to form a physiological response feature vector.

[0016] The electromagnetic personalized treatment parameter generation method provided in this application can generate electromagnetic treatment parameters, combine the normalized feature data with personal information and medical history information, and construct a comprehensive feature vector. This vector not only contains the objective characteristics of the physiological signal, but also incorporates the patient's individual background information, thereby more comprehensively and accurately characterizing the patient's current physiological state and providing high-quality input for subsequent physiological state evaluation models.

[0017] Optionally, filtering the multi-source physiological signal to obtain a filtered multi-source physiological signal includes:

[0018] Performing filtering on the multi-source physiological signals to obtain preliminarily filtered multi-source physiological signals;

[0019] An independent component analysis method is used to remove artifacts from the multi-source physiological signal after the preliminary filtering to obtain a filtered multi-source physiological signal.

[0020] Optionally, feature extraction is performed on the filtered multi-source physiological signal to obtain time domain feature data, frequency domain feature data and nonlinear feature data, including:

[0021] Dividing the filtered multi-source physiological signal into a plurality of signal segments according to a preset time window and sliding step size;

[0022] Using a dynamic time warping algorithm, the time axes of the different signal segments are aligned to obtain time-aligned multi-source physiological signal segments;

[0023] Feature extraction is performed on the time-aligned multi-source physiological signal segments to obtain time domain feature data, frequency domain feature data and nonlinear feature data.

[0024] Optionally, feature extraction is performed on the time-aligned multi-source physiological signal segments to obtain time domain feature data, frequency domain feature data, and nonlinear feature data, including:

[0025] Extracting time domain features from the time-aligned multi-source physiological signal segments and combining them to obtain time domain feature data;

[0026] Extracting frequency domain features from the time-aligned multi-source physiological signal segments and combining them to obtain frequency domain feature data;

[0027] Nonlinear features are extracted from the time-aligned multi-source physiological signal segments and combined to obtain nonlinear feature data.

[0028] Optionally, the time domain feature data, the frequency domain feature data, and the nonlinear feature data are normalized according to the patient's personal information and medical history information, and combined to form a physiological response feature vector, including:

[0029] constructing corresponding feature weight vectors based on the degree of influence of the patient's personal information and medical history information on the time domain feature data, the frequency domain feature data, and the nonlinear feature data respectively;

[0030] performing weighted normalization processing on the time domain feature data, the frequency domain feature data, and the nonlinear feature data according to the feature weight vector to obtain normalized feature vector data;

[0031] The normalized feature vector data is fused and spliced with the patient's personal information and medical history information through one-hot encoding to obtain a physiological response feature vector.

[0032] The method for generating personalized electromagnetic therapy parameters provided in this application can generate electromagnetic therapy parameters by weighting the impact of personal and medical history information on physiological characteristics, generating a physiological response feature vector that better reflects individual patient differences and their regulatory effects on physiological responses. This vector provides richer, more individual-specific input for subsequent physiological state assessment models, thereby supporting the generation of more accurate and personalized electromagnetic therapy parameter adjustment plans.

[0033] Optionally, based on the degree of influence of the patient's personal information and medical history information on the time domain feature data, the frequency domain feature data, and the nonlinear feature data, a corresponding feature weight vector is constructed, including:

[0034] Determining the preliminary influence of each item of personal information and medical history information on the time domain feature data, the frequency domain feature data, and the nonlinear feature data based on expert experience or historical data statistics to construct an initial weight matrix;

[0035] According to preset weight adjustment rules, based on the patient's personal information and medical history information, a fuzzy logic reasoning algorithm is used to calculate the weight correction coefficient of each weight parameter in the initial weight matrix;

[0036] The initial weight matrix is updated by the weight correction coefficient to obtain the corresponding feature weight vector.

[0037] Optionally, the normalized feature vector data is fused and concatenated with the patient's personal information and medical history information by one-hot encoding to obtain a physiological response feature vector, including:

[0038] The patient's personal information and medical history information are converted into coded feature data through one-hot encoding;

[0039] The encoded feature data and the normalized feature vector data are fused and spliced to obtain a physiological response feature vector.

[0040] Optionally, the method for constructing the physiological response assessment model is specifically as follows:

[0041] Obtain historical multi-source physiological signals, personal information and medical history information of multiple patients and corresponding historical physiological status assessment results from a database;

[0042] generating a corresponding historical physiological response feature vector based on the historical multi-source physiological signals, personal information, and medical history information of each patient;

[0043] The physiological response evaluation model is constructed using a machine learning algorithm based on the historical physiological response feature vector and the historical physiological state evaluation results.

[0044] In a second aspect, the present application provides an electromagnetic personalized treatment parameter generation system, comprising:

[0045] an acquisition module, configured to acquire multi-source physiological signals of a patient during electromagnetic stimulation treatment, and to acquire personal information and medical history information of the patient;

[0046] a combination module, configured to pre-process the multi-source physiological signals, extract time domain features, frequency domain features, and nonlinear features, and combine them with the patient's personal information and medical history information to form a physiological response feature vector;

[0047] a calculation module, configured to input the physiological response feature vector into a preset physiological response evaluation model to calculate an evaluation result of the patient's current physiological state;

[0048] The updating module is used to input the current physiological state assessment result and the current treatment stage information of the patient into a preset parameter adjustment model to generate an electromagnetic parameter adjustment plan.

[0049] This electromagnetic personalized treatment parameter generation system uses the patient's multi-source physiological signals, personal information and medical history information to evaluate the patient's current physiological state, and then generates an electromagnetic parameter adjustment plan based on the patient's current treatment stage information. This solves the problem that existing electromagnetic parameter adjustment methods fail to fully utilize the real-time dynamic changes in the patient's state during treatment and fail to integrate the individual differences of patients. It adapts to the changes in the patient's physiological state and individual differences during treatment, improves the level of treatment personalization, and optimizes the treatment effect.

[0050] Beneficial effects: The electromagnetic personalized treatment parameter generation method and system provided in this application evaluate the patient's current physiological state through the patient's multi-source physiological signals, personal information and medical history information, and then generate an electromagnetic parameter adjustment plan in combination with the patient's current treatment stage information. This solves the problem that the existing electromagnetic parameter adjustment method fails to fully utilize the real-time dynamic change information of the patient's state during treatment and fails to integrate the individual differences of the patients. It adapts to the changes in the patient's physiological state and the individual differences of the patients during treatment, improves the level of treatment personalization, and optimizes the treatment effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of the method for generating electromagnetic personalized treatment parameters provided in an embodiment of the present application.

[0052] Figure 2 This is a schematic diagram of the structure of the electromagnetic personalized treatment parameter generation system provided in an embodiment of the present application.

[0053] Explanation of reference numerals: 1. Acquisition module; 2. Combination module; 3. Calculation module; 4. Update module; 301. Processor; 302. Memory; 303. Communication bus. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of the present application.

[0055] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0056] Please refer to Figure 1 , Figure 1 A method for generating electromagnetic personalized treatment parameters in some embodiments of the present application is used to generate electromagnetic treatment parameters, including the steps of:

[0057] Step S101, obtaining multi-source physiological signals of the patient during electromagnetic stimulation treatment, as well as obtaining the patient's personal information and medical history information;

[0058] Step S102 , preprocessing the multi-source physiological signals to extract time domain features, frequency domain features, and nonlinear features, which are combined with the patient's personal information and medical history information to form a physiological response feature vector;

[0059] Step S103, inputting the physiological response feature vector into a preset physiological response evaluation model to calculate and obtain the patient's current physiological state evaluation result;

[0060] Step S104: input the current physiological state assessment result and the patient's current treatment stage information into a preset parameter adjustment model to generate an electromagnetic parameter adjustment plan.

[0061] This electromagnetic personalized treatment parameter generation method uses the patient's multi-source physiological signals, personal information and medical history information to evaluate the patient's current physiological state, and then generates an electromagnetic parameter adjustment plan in combination with the patient's current treatment stage information. This solves the problems that existing electromagnetic parameter adjustment methods fail to fully utilize the real-time dynamic changes in the patient's state during treatment and fail to integrate the individual differences of patients. It adapts to the changes in the patient's physiological state and individual differences during treatment, improves the level of treatment personalization, and optimizes the treatment effect.

[0062] Specifically, in step S101, multi-source physiological signals of the patient during the electromagnetic stimulation treatment are obtained, as well as the patient's personal information and medical history information. The multi-source physiological signals include one or more of the following signals: EEG signals, ECG signals, and EMG signals. The personal information includes information such as the patient's age and gender. The medical history information includes information such as the patient's historical disease type, historical course of disease, and historical medication.

[0063] Specifically, in step S102, multi-source physiological signals are preprocessed to extract time domain features, frequency domain features, and nonlinear features, which are combined with the patient's personal information and medical history information to form a physiological response feature vector, including:

[0064] performing filtering processing on the multi-source physiological signals to obtain filtered multi-source physiological signals;

[0065] Extract features from the filtered multi-source physiological signals to obtain time domain feature data, frequency domain feature data, and nonlinear feature data;

[0066] According to the patient's personal information and medical history information, the time domain feature data, frequency domain feature data and nonlinear feature data are normalized and combined to form a physiological response feature vector.

[0067] Specifically, in step S102, filtering is performed on the multi-source physiological signals to obtain filtered multi-source physiological signals, including:

[0068] Performing filtering processing on the multi-source physiological signals to obtain preliminarily filtered multi-source physiological signals;

[0069] The independent component analysis method is used to remove artifacts from the multi-source physiological signals after preliminary filtering to obtain the filtered multi-source physiological signals.

[0070] In step S102, the original multi-source physiological signal is first subjected to preliminary filtering using existing filtering methods. This preliminary filtering process can reduce the overall noise level of the signal and remove interference from some known frequency bands, resulting in a preliminarily purified signal. However, complex artifacts such as motion artifacts and electrooculogram artifacts present in the physiological signal are difficult to completely remove through simple filtering. To address this issue, independent component analysis (ICA) is used to process the filtered signal after the preliminary filtering. ICA can separate the mixed physiological and artifact signals into multiple independent components. By analyzing the characteristics of these independent components, it is possible to distinguish between components representing true physiological activity and those representing artifacts. Subsequently, the identified artifact components are removed from the signal, and the remaining components are reconstructed to obtain the artifact-removed physiological signal. This artifact-removed signal is the filtered multi-source physiological signal. This combined method of preliminary filtering and ICA can more effectively purify the physiological signal, reduce the interference of artifacts on subsequent feature extraction and physiological state assessment, and thus improve the accuracy of the physiological response feature vector.

[0071] For example, a bandpass filter is used to filter the multi-source physiological signals, such as performing bandpass filtering on the multi-source physiological signals from 0.5Hz to 45Hz to obtain a preliminarily filtered multi-source physiological signal. The preliminarily filtered EEG signal is input into the FastICA algorithm (independent component analysis method) for processing to obtain multiple independent components. By examining the time waveforms, spectral characteristics or topographic maps of these independent components, components related to electrooculography, electromyography or power frequency interference are identified. For example, components with typical electrooculography artifact waveforms or components mainly concentrated in power frequency (such as 50Hz or 60Hz) are marked as artifact components. Subsequently, the weights of these artifact components are set to zero, and the remaining components are used to reconstruct the signal to obtain the EEG signal after artifact removal.

[0072] Specifically, in step S102, feature extraction is performed on the filtered multi-source physiological signals to obtain time domain feature data, frequency domain feature data, and nonlinear feature data, including:

[0073] Divide the filtered multi-source physiological signal into multiple signal segments according to a preset time window and sliding step size;

[0074] Using the dynamic time warping algorithm, the time axes of different signal segments are aligned to obtain time-aligned multi-source physiological signal segments;

[0075] Feature extraction is performed on the time-aligned multi-source physiological signal segments to obtain time domain feature data, frequency domain feature data and nonlinear feature data.

[0076] In step S102, when extracting features from the filtered multi-source physiological signals, the waveforms or patterns of the physiological signals may be nonlinearly stretched or compressed in time due to individual differences between patients, dynamic changes in physiological states, or minor disturbances during the acquisition process. This means that when the same physiological event or pattern occurs at different time points or in different patients, its duration or the temporal relationship of its internal structure may be slightly different. If feature extraction is simply performed according to a fixed time window or sampling point, this temporal misalignment will cause the extracted feature data to lack consistency and comparability, thereby affecting the accuracy of subsequent physiological state assessment. Therefore, directly extracting features from the filtered signals may not effectively handle the temporal variability of the signals and the nonlinear time distortion between different signal segments, resulting in a lack of consistency and comparability in the extracted feature data, affecting the accuracy of subsequent physiological state assessment. By dividing the filtered multi-source physiological signals into multiple signal segments according to a preset time window and sliding step size (the preset time window and sliding step size can be set according to actual needs), the long-term signal is decomposed into easily processable units. A dynamic time warping algorithm is used to align the time axes of different signal segments. This algorithm calculates the temporal correspondence between the segments, corrects for any temporal deviations, and ensures comparable temporal structures for feature extraction. Time, frequency, and nonlinear features are extracted from the time-aligned multi-source physiological signal segments. Feature extraction based on time alignment yields more stable and representative feature data. The dynamic time warping algorithm is an existing one and will not be described in detail here.

[0077] For example, a patient's EEG signal is collected as one of the multi-source physiological signals. The time window is set to 5 seconds, and the sliding step is 1 second. The filtered EEG signal is segmented into multiple 5-second signal segments, with a new segment taken every 1 second. Two adjacent signal segments are selected. A dynamic time warping algorithm is used to calculate the optimal time path between the two segments. For example, Euclidean distance is used as the local distance metric and the Sakoe-Chiba global constrained search space is employed. Based on the calculated optimal path, one or both segments are adjusted on the time axis to align them with the other segment. Features are extracted from the aligned EEG signal segments. For example, time domain features such as mean and standard deviation, frequency domain features such as power spectral density in the Delta, Theta, Alpha, Beta, and Gamma bands, and nonlinear features such as sample entropy and Lyapunov exponents are calculated. This ensures that EEG segments collected at different time points have comparable temporal structures before feature extraction, thereby improving feature stability.

[0078] Specifically, in step S102, feature extraction is performed on the time-aligned multi-source physiological signal segments to obtain time domain feature data, frequency domain feature data, and nonlinear feature data, including:

[0079] Extracting time domain features from the time-aligned multi-source physiological signal segments and combining them to obtain time domain feature data;

[0080] Extract frequency domain features from the time-aligned multi-source physiological signal segments and combine them to obtain frequency domain feature data;

[0081] The nonlinear features in the time-aligned multi-source physiological signal segments are extracted and combined to obtain nonlinear feature data.

[0082] In step S102, by explicitly extracting three types of features: time domain, frequency domain, and nonlinearity, signal information of different dimensions is provided. Time domain features, such as the signal's mean, variance, peak value, and RMS value, are calculated from the time-aligned signal segments. These features reflect the signal's amplitude and variation along the time axis, and are combined to obtain time domain feature data. Frequency domain analysis is performed on the time-aligned signal segments, for example, by calculating the power spectral density through Fourier transform to extract the energy or dominant frequency of specific frequency bands. These features reflect the signal's periodicity or oscillation characteristics, and are combined to obtain frequency domain feature data. Nonlinear features, such as sample entropy, Lyapunov exponent, and correlation dimension, are calculated from the time-aligned signal segments. These features reflect the signal's complexity, degree of chaos, or characteristics of the dynamic system, and are combined to obtain nonlinear feature data. By extracting and combining these three types of features, a comprehensive and multi-faceted description of the physiological signal is obtained. This feature data is used in subsequent physiological response assessment, improving the accuracy and robustness of the assessment. The extraction and calculation methods for time domain features, frequency domain features, and nonlinear features are based on existing algorithms and will not be described in detail here.

[0083] Specifically, in step S102, the time domain feature data, frequency domain feature data, and nonlinear feature data are normalized based on the patient's personal information and medical history information, and combined to form a physiological response feature vector, including:

[0084] Based on the degree of influence of the patient's personal information and medical history information on the time domain feature data, frequency domain feature data and nonlinear feature data, the corresponding feature weight vector is constructed;

[0085] According to the feature weight vector, the time domain feature data, the frequency domain feature data and the nonlinear feature data are weighted and normalized to obtain normalized feature vector data;

[0086] Through one-hot encoding, the normalized feature vector data is fused and spliced with the patient's personal information and medical history information to obtain the physiological response feature vector.

[0087] Specifically, in step S102, based on the degree of influence of the patient's personal information and medical history information on the time domain feature data, frequency domain feature data, and nonlinear feature data, a corresponding feature weight vector is constructed, including:

[0088] Determine the initial impact of each item of personal information and medical history information on the time domain feature data, frequency domain feature data, and nonlinear feature data based on expert experience or historical data statistics to construct an initial weight matrix;

[0089] According to the preset weight adjustment rules, based on the patient's personal information and medical history information, the fuzzy logic reasoning algorithm is used to calculate the weight correction coefficient of each weight parameter in the initial weight matrix;

[0090] The initial weight matrix is updated through the weight correction coefficient to obtain the corresponding feature weight vector.

[0091] In step S102, the initial weight matrix can be constructed by collecting a large amount of historical patient data and analyzing the statistical correlations between different personal information (e.g., age, gender) and medical history information (e.g., disease course, previous treatment history) and physiological characteristics (time domain, frequency domain, nonlinearity) to determine the initial impact. Alternatively, clinical experts can be interviewed and surveyed to quantify their judgments on these impact relationships, thereby providing a universal impact relationship baseline. This initial matrix reflects the average impact of different personal and medical history information on various physiological characteristics under general circumstances.

[0092] However, individual patient conditions vary widely, and simply using this universal baseline for normalization may not be accurate enough. To address this issue, a weight adjustment mechanism based on individual patient information is introduced. Based on pre-defined weight adjustment rules and the patient's specific personal and medical history information, a fuzzy logic inference algorithm is used to calculate weight correction coefficients. The pre-defined weight adjustment rules can be defined as a series of fuzzy rules, for example, "If the patient's age is 'elderly' and the disease duration is 'long,' then the weight correction coefficient for the time-domain features of heart rate variability is 'moderate increase.'" The fuzzy logic inference algorithm can handle the ambiguity and uncertainty in personal and medical history information, such as age group and disease duration, and derive specific adjustments to the initial weights based on this information and the pre-defined weight adjustment rules. These weight correction coefficients are then used to update the initial weight matrix, generating a feature weight vector personalized for the patient. This personalized feature weight vector more accurately reflects the degree to which the patient's personal and medical history information influences their time-domain, frequency-domain, and nonlinear physiological feature data. Therefore, the subsequent weighted normalization processing using this individualized weight vector can more accurately adjust the physiological characteristic data so that it can better reflect the individual physiological response characteristics of the patient and provide more accurate input for subsequent physiological status assessment.

[0093] For example, the initial weight matrix can be an M x N matrix, where M is the total number of personal and medical history information items, and N is the total number of time-domain, frequency-domain, and nonlinear features. Each element W(i, j) in the matrix represents the initial degree of influence of the i-th personal / medical history information item on the j-th physiological feature, with the numerical value range set, for example, between 0 and 1 (determined based on expert experience or historical data statistics). Weight adjustment rules can be stored in a rule base containing multiple fuzzy rules of the "IF-THEN" form, such as "IF age is elderly AND disease duration is long THEN increase the weight correction coefficient is for HRV_SDNN (standard deviation of heart rate variability) by 0.2." The fuzzy logic reasoning process can utilize the Mamdanifuzzy inference method. The patient's age (e.g., 70 years) and disease duration (e.g., 5 years) are fuzzified, and their membership in the corresponding fuzzy sets (e.g., "elderly" and "long") is calculated. Based on the rule base and membership, fuzzy inference is used to obtain the output strength of each rule. Aggregation and defuzzification are then performed to calculate the precise correction coefficient for each physiological feature weight (for example, the correction coefficient for the HRV_SDNN weight is calculated to be 0.18). The final feature weight vector is obtained by adding the calculated correction coefficient to the value of the corresponding element in the initial weight matrix. For example, if the initial influence weight of age on HRV_SDNN in the initial weight matrix is 0.6 and the calculated correction coefficient is 0.18, then the corresponding weight in the final individualized weight vector becomes 0.6 + 0.18 = 0.78.

[0094] Specifically, in step S102, by constructing a feature weight vector, the degree of influence of the patient's personal information and medical history information on the time domain, frequency domain and nonlinear physiological characteristics is quantified, and how individual factors regulate the performance of physiological responses in different feature dimensions is identified. According to the constructed feature weight vector, the extracted physiological feature data (time domain feature data, frequency domain feature data and nonlinear feature data) are weighted and normalized, that is, the extracted physiological feature data are multiplied by the corresponding feature weight vector, and then all feature data are normalized (such as minimum-maximum normalization or Z-score normalization) to obtain normalized feature vector data. This makes the normalized feature data more able to reflect the true performance of the individual under a specific physiological state, and enhances the individual representativeness of the feature.

[0095] Specifically, in step S102, the normalized feature vector data is fused and spliced with the patient's personal information and medical history information through one-hot encoding to obtain a physiological response feature vector, including:

[0096] Through one-hot encoding, the patient's personal information and medical history information are converted into coded feature data;

[0097] The encoded feature data and the normalized feature vector data are fused and spliced to obtain the physiological response feature vector.

[0098] In step S102, a one-hot encoding method is used to convert each classification information in the patient's personal information and medical history information into a binary vector. The length of the vector is equal to the number of possible values of the classification feature, where only one position is 1 and the rest are 0. Thus, the patient's personal information and medical history information are converted into a numerical encoding feature data vector. At the same time, the normalized time domain, frequency domain and nonlinear physiological feature data that have been obtained form a normalized feature vector data. Subsequently, this encoding feature data vector is spliced with the normalized feature vector data. The splicing operation is to simply connect two vectors together to form a longer vector. Thus, a physiological response feature vector containing the patient's physiological characteristics and individual background information is obtained. This physiological response feature vector serves as the input of the subsequent physiological response evaluation model, so that the model can comprehensively consider the immediate state reflected by the physiological signal and the individual differences of the patient when evaluating the patient's current physiological state, thereby improving the accuracy and personalization level of the evaluation.

[0099] For example, suppose a patient's personal information includes gender (male / female) and age group (young / middle-aged / elderly), and their medical history includes diagnosis (depression / anxiety) and previous treatment history (yes / no). Using one-hot encoding, gender "male" might be encoded as [1,0], age group "middle-aged" as [0,1,0], diagnosis "depression" as [1,0], and previous treatment history "yes" as [1,0]. Concatenating these encoding vectors yields encoded feature data, for example, [1,0,0,1,0,1,0,1,0]. Assume the normalized physiological feature vector data is [0.75,0.62,0.88] (representing the normalized time domain, frequency domain, and nonlinear features, respectively). The encoded feature data [1, 0, 0, 1, 0, 1, 0, 1, 0] is concatenated with the normalized feature vector data [0.75, 0.62, 0.88] to obtain the final physiological response feature vector [0.75, 0.62, 0.88, 1, 0, 0, 1, 0, 1, 0, 1, 0]. This vector contains the patient's physiological state information and individual background information and is used as input to the physiological response assessment model.

[0100] Specifically, in step S103, the physiological response feature vector is input into a preset physiological response assessment model, which maps complex physiological response features into quantifiable state indicators, calculates the patient's current physiological state assessment result, and realizes an accurate assessment of the patient's real-time state. The current physiological state assessment result includes the patient's reaction degree to the current stimulus, tolerance or emotional state, and the corresponding score. The method for constructing the physiological response assessment model is specifically as follows:

[0101] Obtain historical multi-source physiological signals, personal information and medical history information of multiple patients and corresponding historical physiological status assessment results from a database;

[0102] Based on each patient's historical multi-source physiological signals, personal information, and medical history information, a corresponding historical physiological response feature vector is generated;

[0103] Based on the historical physiological response feature vectors and historical physiological state assessment results, a physiological response assessment model was constructed using a machine learning algorithm.

[0104] When constructing a physiological response assessment model, historical physiological signals, personal information, medical history, and physiological state assessment results corresponding to each treatment session of different patients during electromagnetic stimulation treatment are collected from a database. This historical data provides the foundational samples needed for model training. The collected historical multi-source physiological signals are preprocessed and feature extracted. Combined with the patient's historical personal information and medical history, historical physiological response feature vectors are generated according to the aforementioned method for combining physiological response feature vectors. This process ensures that the data format used for model training is consistent with the input data format received during actual model operation. Thus, using these historical physiological response feature vectors as input data and the corresponding historical physiological state assessment results as target outputs, machine learning algorithms (such as existing algorithms such as support vector machines (SVMs), neural networks, and random forests) are used for training. Machine learning algorithms can learn the complex mapping relationship between input features and physiological state assessment results from a large number of historical input-output pairs, thereby constructing a model that can predict the patient's current physiological state based on the new physiological response feature vectors. By training on diverse historical data, the constructed model can better capture individual differences between patients and the dynamic changes in physiological status during treatment, thereby improving the accuracy of the patient's current physiological status assessment and providing a more reliable basis for subsequent electromagnetic parameter adjustments. The machine learning algorithm is currently available and will not be described in detail here.

[0105] Specifically, in step S104, the calculated current physiological state assessment results and the patient's current treatment stage (e.g., whether it is early, mid-stage, or late-stage treatment) are input into another preset parameter adjustment model. This parameter adjustment model dynamically generates an optimized electromagnetic parameter adjustment plan based on the patient's real-time physiological state assessment results and treatment stage. The electromagnetic parameter adjustment plan can be a recommendation for increasing or decreasing the electromagnetic parameters corresponding to the current treatment stage (e.g., a recommendation to adjust stimulation intensity, frequency, pulse pattern, or treatment duration), or a new parameter combination.

[0106] The parameter adjustment model is constructed using a machine learning algorithm, capable of learning patterns by analyzing a large amount of historical treatment data. This historical data includes physiological responses, treatment phases, and final treatment outcomes or patient feedback from different patients under different electromagnetic stimulation parameter settings. By learning from this data, the model establishes a correlation between physiological state, treatment phase, and electromagnetic stimulation parameter adjustment plans that can improve treatment outcomes, thereby outputting electromagnetic parameter adjustment plans.

[0107] As can be seen from the above, the electromagnetic personalized treatment parameter generation method obtains the patient's multi-source physiological signals during the electromagnetic stimulation treatment process, as well as the patient's personal information and medical history information, pre-processes the multi-source physiological signals, extracts time domain features, frequency domain features and nonlinear features, and combines the patient's personal information and medical history information to form a physiological response feature vector. The physiological response feature vector is input into a preset physiological response evaluation model to calculate the patient's current physiological state evaluation result. The current physiological state evaluation result and the patient's current treatment stage information are input into a preset parameter adjustment model to generate an electromagnetic parameter adjustment plan. Thus, after evaluating the patient's current physiological state evaluation result through the patient's multi-source physiological signals, personal information and medical history information, an electromagnetic parameter adjustment plan is generated in combination with the patient's current treatment stage information, thereby solving the problem that the existing electromagnetic parameter adjustment method fails to fully utilize the real-time dynamic change information of the patient's state during the treatment process and fails to integrate the individual differences of the patient. The method adapts to the changes in the patient's physiological state and the individual differences of the patient during the treatment process, improves the level of treatment personalization, and optimizes the treatment effect.

[0108] refer to Figure 2 The present application provides an electromagnetic personalized treatment parameter generation system for generating electromagnetic treatment parameters, including:

[0109] Acquisition module 1, used to acquire multi-source physiological signals of the patient during electromagnetic stimulation treatment, as well as the patient's personal information and medical history information;

[0110] Combination module 2 is used to pre-process multi-source physiological signals, extract time domain features, frequency domain features and nonlinear features, and combine them with the patient's personal information and medical history information to form a physiological response feature vector;

[0111] Calculation module 3, used to input the physiological response feature vector into a preset physiological response evaluation model to calculate the patient's current physiological state evaluation result;

[0112] The updating module 4 is used to input the current physiological state assessment result and the patient's current treatment stage information into a preset parameter adjustment model to generate an electromagnetic parameter adjustment plan.

[0113] This electromagnetic personalized treatment parameter generation system uses the patient's multi-source physiological signals, personal information and medical history information to evaluate the patient's current physiological state, and then generates an electromagnetic parameter adjustment plan based on the patient's current treatment stage information. This solves the problem that existing electromagnetic parameter adjustment methods fail to fully utilize the real-time dynamic changes in the patient's state during treatment and fail to integrate the individual differences of patients. It adapts to the changes in the patient's physiological state and individual differences during treatment, improves the level of treatment personalization, and optimizes the treatment effect.

[0114] Specifically, when the acquisition module 1 is executed, it obtains the multi-source physiological signals of the patient during the electromagnetic stimulation treatment, as well as the patient's personal information and medical history information, wherein the multi-source physiological signals include one or more of the following signals: EEG signals, ECG signals, EMG signals, etc.; the personal information includes information such as the patient's age and gender; and the medical history information includes information such as the patient's historical disease type, historical course of disease, historical medication status, etc.

[0115] Specifically, when the combination module 2 preprocesses the multi-source physiological signals and extracts the time domain features, frequency domain features, and nonlinear features to combine the patient's personal information and medical history information to form a physiological response feature vector, it executes:

[0116] performing filtering processing on the multi-source physiological signals to obtain filtered multi-source physiological signals;

[0117] Extract features from the filtered multi-source physiological signals to obtain time domain feature data, frequency domain feature data, and nonlinear feature data;

[0118] According to the patient's personal information and medical history information, the time domain feature data, frequency domain feature data and nonlinear feature data are normalized and combined to form a physiological response feature vector.

[0119] Specifically, when the combination module 2 performs filtering processing on the multi-source physiological signals to obtain the filtered multi-source physiological signals, it executes:

[0120] Performing filtering processing on the multi-source physiological signals to obtain preliminarily filtered multi-source physiological signals;

[0121] The independent component analysis method is used to remove artifacts from the multi-source physiological signals after preliminary filtering to obtain the filtered multi-source physiological signals.

[0122] When executing, combination module 2 first performs preliminary filtering on the original multi-source physiological signals using existing filtering methods. This preliminary filtering can reduce the overall noise level of the signal and remove interference from some known frequency bands, resulting in a preliminarily purified signal. However, complex artifacts such as motion artifacts and electrooculogram artifacts present in physiological signals are difficult to completely remove through simple filtering. To address this issue, after the preliminary filtering, independent component analysis (ICA) is used to process the preliminarily filtered signals. ICA can separate the mixed physiological and artifact signals into multiple independent components. By analyzing the characteristics of these independent components, it is possible to distinguish between components representing true physiological activity and those representing artifacts. Subsequently, the identified artifact components are removed from the signal, and the remaining components are reconstructed to obtain the artifact-removed physiological signal. This artifact-removed signal is the filtered multi-source physiological signal. This combined method of preliminary filtering and ICA can more effectively purify physiological signals, reduce the interference of artifacts on subsequent feature extraction and physiological state assessment, and thus improve the accuracy of the physiological response feature vector.

[0123] For example, a bandpass filter is used to filter the multi-source physiological signals, such as performing bandpass filtering on the multi-source physiological signals from 0.5Hz to 45Hz to obtain a preliminarily filtered multi-source physiological signal. The preliminarily filtered EEG signal is input into the FastICA algorithm (independent component analysis method) for processing to obtain multiple independent components. By examining the time waveforms, spectral characteristics or topographic maps of these independent components, components related to electrooculography, electromyography or power frequency interference are identified. For example, components with typical electrooculography artifact waveforms or components mainly concentrated in power frequency (such as 50Hz or 60Hz) are marked as artifact components. Subsequently, the weights of these artifact components are set to zero, and the remaining components are used to reconstruct the signal to obtain the EEG signal after artifact removal.

[0124] Specifically, when the combination module 2 extracts features from the filtered multi-source physiological signals to obtain time domain feature data, frequency domain feature data, and nonlinear feature data, it executes:

[0125] Divide the filtered multi-source physiological signal into multiple signal segments according to a preset time window and sliding step size;

[0126] Using the dynamic time warping algorithm, the time axes of different signal segments are aligned to obtain time-aligned multi-source physiological signal segments;

[0127] Feature extraction is performed on the time-aligned multi-source physiological signal segments to obtain time domain feature data, frequency domain feature data and nonlinear feature data.

[0128] When the combination module 2 is executed and features are extracted from the filtered multi-source physiological signals, the waveform or pattern of the physiological signals may be stretched or compressed nonlinearly in time due to individual differences in patients, dynamic changes in physiological state, or small disturbances during the acquisition process. This means that when the same physiological event or pattern occurs at different time points or in different patients, its duration or the temporal relationship of its internal structure may be slightly different. If feature extraction is simply performed according to a fixed time window or sampling point, this temporal misalignment will lead to a lack of consistency and comparability in the extracted feature data, thereby affecting the accuracy of subsequent physiological state assessment. Therefore, directly extracting features from the filtered signals may not effectively handle the temporal variability of the signals and the nonlinear time distortion between different signal segments, resulting in a lack of consistency and comparability in the extracted feature data, affecting the accuracy of subsequent physiological state assessment. By dividing the filtered multi-source physiological signals into multiple signal segments according to a preset time window and sliding step size (the preset time window and sliding step size can be set according to actual needs), the long-term signal is decomposed into easily processable units. A dynamic time warping algorithm is used to align the time axes of different signal segments. This algorithm calculates the temporal correspondence between the segments, corrects for any temporal deviations, and ensures comparable temporal structures for feature extraction. Time, frequency, and nonlinear features are extracted from the time-aligned multi-source physiological signal segments. Feature extraction based on time alignment yields more stable and representative feature data. The dynamic time warping algorithm is an existing one and will not be described in detail here.

[0129] For example, a patient's EEG signal is collected as one of the multi-source physiological signals. The time window is set to 5 seconds, and the sliding step is 1 second. The filtered EEG signal is segmented into multiple 5-second signal segments, with a new segment taken every 1 second. Two adjacent signal segments are selected. A dynamic time warping algorithm is used to calculate the optimal time path between the two segments. For example, Euclidean distance is used as the local distance metric and the Sakoe-Chiba global constrained search space is employed. Based on the calculated optimal path, one or both segments are adjusted on the time axis to align them with the other segment. Features are extracted from the aligned EEG signal segments. For example, time domain features such as mean and standard deviation, frequency domain features such as power spectral density in the Delta, Theta, Alpha, Beta, and Gamma bands, and nonlinear features such as sample entropy and Lyapunov exponents are calculated. This ensures that EEG segments collected at different time points have comparable temporal structures before feature extraction, thereby improving feature stability.

[0130] Specifically, when the combination module 2 extracts features from the time-aligned multi-source physiological signal segments to obtain time domain feature data, frequency domain feature data, and nonlinear feature data, it executes:

[0131] Extracting time domain features from the time-aligned multi-source physiological signal segments and combining them to obtain time domain feature data;

[0132] Extract frequency domain features from the time-aligned multi-source physiological signal segments and combine them to obtain frequency domain feature data;

[0133] The nonlinear features in the time-aligned multi-source physiological signal segments are extracted and combined to obtain nonlinear feature data.

[0134] When executed, Combination Module 2 provides signal information of different dimensions by explicitly extracting three types of features: time domain, frequency domain, and nonlinearity. Time domain features, such as the signal's mean, variance, peak value, and RMS value, are calculated from the time-aligned signal segments. These features reflect the signal's amplitude and variation along the time axis, and are combined to produce time domain feature data. Frequency domain analysis is performed on the time-aligned signal segments, such as by calculating the power spectral density through Fourier transform to extract the energy or dominant frequency in specific frequency bands. These features reflect the signal's periodicity or oscillation characteristics, and are combined to produce frequency domain feature data. Nonlinear features, such as sample entropy, Lyapunov exponent, and correlation dimension, are calculated from the time-aligned signal segments. These features reflect the signal's complexity, degree of chaos, or characteristics of the dynamic system, and are combined to produce nonlinear feature data. By extracting and combining these three types of features, a comprehensive and multi-faceted description of the physiological signal is obtained. This feature data is used in subsequent physiological response assessment, improving the accuracy and robustness of the assessment. The extraction and calculation methods for time domain, frequency domain, and nonlinear features are based on existing algorithms and will not be described in detail here.

[0135] Specifically, when the combining module 2 normalizes the time domain feature data, the frequency domain feature data, and the nonlinear feature data according to the patient's personal information and medical history information and combines them to form a physiological response feature vector, it executes:

[0136] Based on the degree of influence of the patient's personal information and medical history information on the time domain feature data, frequency domain feature data and nonlinear feature data, the corresponding feature weight vector is constructed;

[0137] According to the feature weight vector, the time domain feature data, the frequency domain feature data and the nonlinear feature data are weighted and normalized to obtain normalized feature vector data;

[0138] Through one-hot encoding, the normalized feature vector data is fused and spliced with the patient's personal information and medical history information to obtain the physiological response feature vector.

[0139] Specifically, when constructing corresponding feature weight vectors based on the degree of influence of the patient's personal information and medical history information on the time domain feature data, the frequency domain feature data, and the nonlinear feature data, the combination module 2 executes:

[0140] Determine the initial impact of each item of personal information and medical history information on the time domain feature data, frequency domain feature data, and nonlinear feature data based on expert experience or historical data statistics to construct an initial weight matrix;

[0141] According to the preset weight adjustment rules, based on the patient's personal information and medical history information, the fuzzy logic reasoning algorithm is used to calculate the weight correction coefficient of each weight parameter in the initial weight matrix;

[0142] The initial weight matrix is updated through the weight correction coefficient to obtain the corresponding feature weight vector.

[0143] When Combination Module 2 is executed, the initial weight matrix can be constructed by collecting a large amount of historical patient data and analyzing the statistical correlations between various personal information (such as age and gender) and medical history information (such as disease course and previous treatment history) and physiological characteristics (time domain, frequency domain, and nonlinearity) to determine the initial impact level. Alternatively, clinical experts can be interviewed and surveyed to quantify their judgments on these impact relationships, thereby providing a universal impact relationship baseline. This initial matrix reflects the average impact of different personal and medical history information on various physiological characteristics under general circumstances.

[0144] However, individual patient conditions vary widely, and simply using this universal baseline for normalization may not be accurate enough. To address this issue, a weight adjustment mechanism based on individual patient information is introduced. Based on pre-defined weight adjustment rules and the patient's specific personal and medical history information, a fuzzy logic inference algorithm is used to calculate weight correction coefficients. The pre-defined weight adjustment rules can be defined as a series of fuzzy rules, for example, "If the patient's age is 'elderly' and the disease duration is 'long,' then the weight correction coefficient for the time-domain features of heart rate variability is 'moderate increase.'" The fuzzy logic inference algorithm can handle the ambiguity and uncertainty in personal and medical history information, such as age group and disease duration, and derive specific adjustments to the initial weights based on this information and the pre-defined weight adjustment rules. These weight correction coefficients are then used to update the initial weight matrix, generating a feature weight vector personalized for the patient. This personalized feature weight vector more accurately reflects the degree to which the patient's personal and medical history information influences their time-domain, frequency-domain, and nonlinear physiological feature data. Therefore, the subsequent weighted normalization processing using this individualized weight vector can more accurately adjust the physiological characteristic data so that it can better reflect the individual physiological response characteristics of the patient and provide more accurate input for subsequent physiological status assessment.

[0145] For example, the initial weight matrix can be an M x N matrix, where M is the total number of personal and medical history information items, and N is the total number of time-domain, frequency-domain, and nonlinear features. Each element W(i, j) in the matrix represents the initial degree of influence of the i-th personal / medical history information item on the j-th physiological feature, with the numerical value range set, for example, between 0 and 1 (determined based on expert experience or historical data statistics). Weight adjustment rules can be stored in a rule base containing multiple fuzzy rules of the "IF-THEN" form, such as "IF age is elderly AND disease duration is long THEN increase the weight correction coefficient is for HRV_SDNN (standard deviation of heart rate variability) by 0.2." The fuzzy logic reasoning process can utilize the Mamdanifuzzy inference method. The patient's age (e.g., 70 years) and disease duration (e.g., 5 years) are fuzzified, and their membership in the corresponding fuzzy sets (e.g., "elderly" and "long") is calculated. Based on the rule base and membership, fuzzy inference is used to obtain the output strength of each rule. Aggregation and defuzzification are then performed to calculate the precise correction coefficient for each physiological feature weight (for example, the correction coefficient for the HRV_SDNN weight is calculated to be 0.18). The final feature weight vector is obtained by adding the calculated correction coefficient to the value of the corresponding element in the initial weight matrix. For example, if the initial influence weight of age on HRV_SDNN in the initial weight matrix is 0.6 and the calculated correction coefficient is 0.18, then the corresponding weight in the final individualized weight vector becomes 0.6 + 0.18 = 0.78.

[0146] Specifically, when the combination module 2 is executed, it constructs a feature weight vector to quantify the degree of influence of the patient's personal information and medical history information on the time domain, frequency domain and nonlinear physiological characteristics, and identifies how individual factors regulate the performance of physiological responses in different feature dimensions. According to the constructed feature weight vector, the extracted physiological feature data (time domain feature data, frequency domain feature data and nonlinear feature data) are weighted and normalized, that is, the extracted physiological feature data are multiplied by the corresponding feature weight vector, and then all feature data are normalized (such as minimum-maximum normalization or Z-score normalization) to obtain normalized feature vector data. This makes the normalized feature data more able to reflect the true performance of the individual in a specific physiological state, and enhances the individual representativeness of the feature.

[0147] Specifically, when the combination module 2 fuses the normalized feature vector data with the patient's personal information and medical history information through one-hot encoding to obtain the physiological response feature vector, it executes:

[0148] Through one-hot encoding, the patient's personal information and medical history information are converted into coded feature data;

[0149] The encoded feature data and the normalized feature vector data are fused and spliced to obtain the physiological response feature vector.

[0150] When the combination module 2 is executed, the one-hot encoding method is used to convert each classification information in the patient's personal information and medical history information into a binary vector. The length of the vector is equal to the number of possible values of the classification feature, where only one position is 1 and the rest are 0. Thus, the patient's personal information and medical history information are converted into a numerical encoding feature data vector. At the same time, the normalized time domain, frequency domain and nonlinear physiological feature data that have been obtained form a normalized feature vector data. Subsequently, this encoding feature data vector is spliced with the normalized feature vector data. The splicing operation is to simply connect two vectors together to form a longer vector. Thus, a physiological response feature vector containing the patient's physiological characteristics and individual background information is obtained. This physiological response feature vector serves as the input of the subsequent physiological response evaluation model, so that the model can comprehensively consider the immediate state reflected by the physiological signal and the individual differences of the patient when evaluating the patient's current physiological state, thereby improving the accuracy and personalization level of the evaluation.

[0151] For example, suppose a patient's personal information includes gender (male / female) and age group (young / middle-aged / elderly), and their medical history includes diagnosis (depression / anxiety) and previous treatment history (yes / no). Using one-hot encoding, gender "male" might be encoded as [1,0], age group "middle-aged" as [0,1,0], diagnosis "depression" as [1,0], and previous treatment history "yes" as [1,0]. Concatenating these encoding vectors yields encoded feature data, for example, [1,0,0,1,0,1,0,1,0]. Assume the normalized physiological feature vector data is [0.75,0.62,0.88] (representing the normalized time domain, frequency domain, and nonlinear features, respectively). The encoded feature data [1, 0, 0, 1, 0, 1, 0, 1, 0] is concatenated with the normalized feature vector data [0.75, 0.62, 0.88] to obtain the final physiological response feature vector [0.75, 0.62, 0.88, 1, 0, 0, 1, 0, 1, 0, 1, 0]. This vector contains the patient's physiological state information and individual background information and is used as input to the physiological response assessment model.

[0152] Specifically, when the calculation module 3 is executed, the physiological response feature vector is input into a preset physiological response evaluation model. The model maps the complex physiological response features into quantifiable state indicators, calculates the patient's current physiological state evaluation result, and realizes an accurate assessment of the patient's real-time state. The current physiological state evaluation result includes the patient's reaction degree to the current stimulus, tolerance or emotional state, and the corresponding score. The specific method for constructing the physiological response evaluation model is as follows:

[0153] Obtain historical multi-source physiological signals, personal information and medical history information of multiple patients and corresponding historical physiological status assessment results from a database;

[0154] Generate a corresponding historical physiological response feature vector based on each patient's historical multi-source physiological signals, personal information, and medical history information;

[0155] Based on the historical physiological response feature vectors and historical physiological state assessment results, a physiological response assessment model was constructed using a machine learning algorithm.

[0156] When constructing a physiological response assessment model, historical physiological signals, personal information, medical history, and physiological state assessment results corresponding to each treatment session of different patients during electromagnetic stimulation treatment are collected from a database. This historical data provides the foundational samples needed for model training. The collected historical multi-source physiological signals are preprocessed and feature extracted. Combined with the patient's historical personal information and medical history, historical physiological response feature vectors are generated according to the aforementioned method for combining physiological response feature vectors. This process ensures that the data format used for model training is consistent with the input data format received during actual model operation. Thus, using these historical physiological response feature vectors as input data and the corresponding historical physiological state assessment results as target outputs, machine learning algorithms (such as existing algorithms such as support vector machines (SVMs), neural networks, and random forests) are used for training. Machine learning algorithms can learn the complex mapping relationship between input features and physiological state assessment results from a large number of historical input-output pairs, thereby constructing a model that can predict the patient's current physiological state based on the new physiological response feature vectors. By training on diverse historical data, the constructed model can better capture individual differences between patients and the dynamic changes in physiological status during treatment, thereby improving the accuracy of the patient's current physiological status assessment and providing a more reliable basis for subsequent electromagnetic parameter adjustments. The machine learning algorithm is currently available and will not be described in detail here.

[0157] Specifically, during execution, generation module 4 inputs the calculated current physiological state assessment results and the patient's current treatment stage information (e.g., whether it is early, mid-stage, or late-stage treatment) into another preset parameter adjustment model. This parameter adjustment model dynamically generates an optimized electromagnetic parameter adjustment plan based on the patient's real-time physiological state assessment results and treatment stage. The electromagnetic parameter adjustment plan can be a recommendation to increase or decrease the electromagnetic parameters corresponding to the current treatment stage (e.g., a recommendation to adjust the stimulation intensity, frequency, pulse mode, or treatment duration), or a new parameter combination.

[0158] The parameter adjustment model is constructed using a machine learning algorithm, capable of learning patterns by analyzing a large amount of historical treatment data. This historical data includes physiological responses, treatment phases, and final treatment outcomes or patient feedback from different patients under different electromagnetic stimulation parameter settings. By learning from this data, the model establishes a correlation between physiological state, treatment phase, and electromagnetic stimulation parameter adjustment plans that can improve treatment outcomes, thereby outputting electromagnetic parameter adjustment plans.

[0159] As can be seen from the above, the electromagnetic personalized treatment parameter generation system obtains the patient's multi-source physiological signals during the electromagnetic stimulation treatment process, as well as the patient's personal information and medical history information, pre-processes the multi-source physiological signals, extracts time domain features, frequency domain features and nonlinear features, and combines the patient's personal information and medical history information to form a physiological response feature vector. The physiological response feature vector is input into a preset physiological response evaluation model to calculate the patient's current physiological state evaluation result. The current physiological state evaluation result and the patient's current treatment stage information are input into a preset parameter adjustment model to generate an electromagnetic parameter adjustment plan. Thus, after evaluating the patient's current physiological state evaluation result through the patient's multi-source physiological signals, personal information and medical history information, an electromagnetic parameter adjustment plan is generated in combination with the patient's current treatment stage information, thereby solving the problem that the existing electromagnetic parameter adjustment method fails to fully utilize the real-time dynamic change information of the patient's state during the treatment process and fails to integrate the individual differences of the patient. The system adapts to the changes in the patient's physiological state and the individual differences of the patient during the treatment process, improves the level of treatment personalization, and optimizes the treatment effect.

[0160] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. There may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some communication interface, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0161] In addition, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected based on actual needs to achieve the purpose of the solution of this embodiment.

[0162] Furthermore, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0163] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any actual relationship or order between these entities or operations.

[0164] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for generating electromagnetic personalized treatment parameters, for generating electromagnetic treatment parameters, characterized in that: Including steps: Acquiring multi-source physiological signals from a patient during electromagnetic stimulation treatment, as well as obtaining the patient's personal information and medical history information; preprocessing the multi-source physiological signals to extract time domain features, frequency domain features, and nonlinear features, which are combined with the patient's personal information and medical history information to form a physiological response feature vector; Inputting the physiological response feature vector into a preset physiological response evaluation model to calculate and obtain an evaluation result of the patient's current physiological state; The current physiological state assessment result and the patient's current treatment stage information are input into a preset parameter adjustment model to generate an electromagnetic parameter adjustment plan.

2. The method for generating electromagnetic personalized treatment parameters according to claim 1, characterized in that: Preprocessing the multi-source physiological signals to extract time domain features, frequency domain features, and nonlinear features, which are combined with the patient's personal information and medical history information to form a physiological response feature vector, including: performing filtering processing on the multi-source physiological signals to obtain filtered multi-source physiological signals; Extracting features from the filtered multi-source physiological signals to obtain time domain feature data, frequency domain feature data, and nonlinear feature data; According to the personal information and medical history information of the patient, the time domain feature data, the frequency domain feature data and the nonlinear feature data are normalized and combined to form a physiological response feature vector.

3. The method for generating electromagnetic personalized treatment parameters according to claim 2, characterized in that: Filtering the multi-source physiological signals to obtain filtered multi-source physiological signals includes: performing filtering processing on the multi-source physiological signals to obtain preliminarily filtered multi-source physiological signals; An independent component analysis method is used to remove artifacts from the multi-source physiological signal after preliminary filtering to obtain a filtered multi-source physiological signal.

4. The method for generating electromagnetic personalized treatment parameters according to claim 2, characterized in that: Performing feature extraction on the filtered multi-source physiological signals to obtain time domain feature data, frequency domain feature data, and nonlinear feature data, including: Dividing the filtered multi-source physiological signal into a plurality of signal segments according to a preset time window and sliding step size; Using a dynamic time warping algorithm, the time axes of the different signal segments are aligned to obtain time-aligned multi-source physiological signal segments; Feature extraction is performed on the time-aligned multi-source physiological signal segments to obtain time domain feature data, frequency domain feature data and nonlinear feature data.

5. The method for generating electromagnetic personalized treatment parameters according to claim 4, characterized in that: Performing feature extraction on the time-aligned multi-source physiological signal segments to obtain time domain feature data, frequency domain feature data, and nonlinear feature data, including: Extracting time domain features from the time-aligned multi-source physiological signal segments and combining them to obtain time domain feature data; Extracting frequency domain features from the time-aligned multi-source physiological signal segments and combining them to obtain frequency domain feature data; Nonlinear features are extracted from the time-aligned multi-source physiological signal segments and combined to obtain nonlinear feature data.

6. The method for generating electromagnetic personalized treatment parameters according to claim 2, characterized in that: According to the personal information and medical history information of the patient, the time domain feature data, the frequency domain feature data, and the nonlinear feature data are normalized and combined to form a physiological response feature vector, including: constructing corresponding feature weight vectors based on the degree of influence of the patient's personal information and medical history information on the time domain feature data, the frequency domain feature data, and the nonlinear feature data respectively; performing weighted normalization processing on the time domain feature data, the frequency domain feature data, and the nonlinear feature data according to the feature weight vector to obtain normalized feature vector data; The normalized feature vector data is fused and spliced with the patient's personal information and medical history information through one-hot encoding to obtain a physiological response feature vector.

7. The method for generating electromagnetic personalized treatment parameters according to claim 6, characterized in that: Based on the degree of influence of the patient's personal information and medical history information on the time domain feature data, the frequency domain feature data, and the nonlinear feature data, a corresponding feature weight vector is constructed, including: Determining the preliminary influence of each item of personal information and medical history information on the time domain feature data, the frequency domain feature data, and the nonlinear feature data based on expert experience or historical data statistics to construct an initial weight matrix; According to preset weight adjustment rules, based on the patient's personal information and medical history information, a fuzzy logic reasoning algorithm is used to calculate the weight correction coefficient of each weight parameter in the initial weight matrix; The initial weight matrix is updated by the weight correction coefficient to obtain the corresponding feature weight vector.

8. The method for generating electromagnetic personalized treatment parameters according to claim 6, characterized in that: The normalized feature vector data is fused and spliced with the patient's personal information and medical history information through one-hot encoding to obtain a physiological response feature vector, including: The patient's personal information and medical history information are converted into coded feature data through one-hot encoding; The encoded feature data and the normalized feature vector data are fused and spliced to obtain a physiological response feature vector.

9. The method for generating electromagnetic personalized treatment parameters according to claim 1, characterized in that: The method for constructing the physiological response assessment model is specifically as follows: Obtain historical multi-source physiological signals, personal information and medical history information of multiple patients and corresponding historical physiological status assessment results from a database; generating a corresponding historical physiological response feature vector based on the historical multi-source physiological signals, personal information, and medical history information of each patient; The physiological response evaluation model is constructed using a machine learning algorithm based on the historical physiological response feature vector and the historical physiological state evaluation results.

10. An electromagnetic personalized treatment parameter generation system for generating electromagnetic treatment parameters, characterized in that: include: an acquisition module, configured to acquire multi-source physiological signals of a patient during electromagnetic stimulation treatment, and to acquire personal information and medical history information of the patient; a combination module, configured to pre-process the multi-source physiological signals, extract time domain features, frequency domain features, and nonlinear features, and combine them with the patient's personal information and medical history information to form a physiological response feature vector; a calculation module, configured to input the physiological response feature vector into a preset physiological response evaluation model to calculate an evaluation result of the patient's current physiological state; The generation module is used to input the current physiological state assessment result and the current treatment stage information of the patient into a preset parameter adjustment model to generate an electromagnetic parameter adjustment plan.

Images