Pain management method based on bioelectrical impedance data and machine learning algorithm

Abstract

The invention discloses a pain management method based on bioelectrical impedance data and a machine learning algorithm, and relates to the technical field of medical health, and the method comprises the steps: carrying out the phase delay analysis of an impedance phase spectrum matrix, calculating the change rate of a phase angle of each frequency point along with time, and obtaining the maximum phase lag time; according to the temperature field sequence and the maximum phase lag time, time-space registration is carried out through a multi-modal data fusion method and impedance modulus changes of the corresponding areas, and a pain biomarker feature library is generated; through a bidirectional LSTM and a cross attention fusion method, time sequence modeling and cross-modal correlation analysis are carried out on bioelectricity signal characteristics extracted by a sliding window and historical pain data, a dynamic evolution rule of pain development is captured, a pain management triple containing pain intensity, intervention response and prognosis evaluation is generated, and a pain management result is obtained. And a quantitative basis is provided for personalized treatment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical health, and particularly relates to a pain management method based on bioelectrical impedance data and a machine learning algorithm. BACKGROUND

[0002] In the field of pain management, bioelectrical impedance analysis (BIA) and infrared thermography have become important physiological monitoring means. Traditional methods usually use single modal data (such as bioelectrical impedance or temperature field) for pain assessment, which indirectly reflects pain-related physiological changes by measuring tissue impedance characteristics or body surface temperature distribution. Bioelectrical impedance technology measures the impedance amplitude and phase angle of human tissue through multi-frequency AC signals, and combines with human composition analysis algorithms to obtain the electrical property changes of muscle, fat and other tissues. Infrared thermography measures the body surface temperature distribution through non-contact measurement, and analyzes the local blood flow and metabolic state by using thermal radiation signals. Existing technologies have realized pain classification based on threshold determination, for example, by comparing the impedance phase angle shift or regional temperature difference with the threshold to infer the degree of pain.

[0003] However, the existing methods still have limitations in data fusion and dynamic assessment. On the one hand, the difference in spatial and temporal resolution between bioelectrical impedance and infrared temperature field leads to insufficient cross-modal registration accuracy, for example, the sampling rate of temperature field (usually 30 Hz) is much lower than that of high-frequency impedance measurement (usually 1 kHz), and direct interpolation will cause time series distortion. On the other hand, traditional pain assessment models rely on static features (such as mean or extreme value), which are difficult to capture dynamic electrical-thermal coupling effects such as phase lag time, and studies have shown that the sympathetic nervous response caused by pain can significantly affect the delay relationship between tissue electrical properties and temperature conduction. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a pain management method based on bioelectrical impedance data and a machine learning algorithm, which solves the problems of insufficient multi-modal data fusion accuracy and insufficient dynamic electrical-thermal coupling feature extraction in the prior art.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a pain management method based on bioelectrical impedance data and a machine learning algorithm, which includes,

[0008] Collecting bioelectrical impedance data and infrared temperature field data and preprocessing to generate timestamp-aligned impedance phase spectrum matrix and temperature field sequence;

[0009] The phase delay analysis is performed on the impedance phase spectrum matrix, the change rate of the phase angle of each frequency point with time is calculated, and the maximum phase lag time is obtained;

[0010] According to the temperature field sequence and the maximum phase lag time, the impedance modulus change of the corresponding region is spatiotemporally registered through the multi-modal data fusion method, and a pain biomarker feature library is generated;

[0011] The bioelectric signal features are extracted from the pain biomarker feature library through time sliding, and input into the machine learning pain assessment model for cross-attention analysis, and a pain management triple is generated.

[0012] The pain management triple is compared and analyzed with historical clinical pain cases, a treatment intervention scheme is generated, and a simulation treatment intervention is performed, the data of the pain management triple before and after the treatment intervention are compared, the effectiveness of the treatment intervention method is verified, and the treatment intervention scheme is dynamically optimized and a treatment suggestion is generated.

[0013] As a preferred scheme of the pain management method based on bioelectrical impedance data and machine learning algorithm, the bioelectrical impedance data includes impedance amplitude, phase angle and human body composition estimation value;

[0014] The infrared temperature field data includes human body surface temperature distribution, spatial resolution temperature distribution and time sequence temperature distribution;

[0015] The preprocessing includes outlier rejection, filtering and noise reduction, phase spectrum smoothing, temperature correction and non-uniformity correction;

[0016] The preprocessed bioelectrical impedance and infrared temperature field data are converted to a unified time reference, time alignment interpolation is performed under the constraint of a reference clock, and a timestamp-aligned impedance phase spectrum matrix and temperature field sequence are generated.

[0017] As a preferred scheme of the pain management method based on bioelectrical impedance data and machine learning algorithm, the phase delay analysis is performed on the impedance phase spectrum matrix, the change rate of the phase angle of each frequency point with time is calculated, and the maximum phase lag time is obtained, and the specific steps are as follows,

[0018] The impedance phase spectrum matrix is divided by sliding according to a fixed time, the impedance phase angle features are extracted, and the time is aligned, and a one-dimensional sequence of the phase angle change with time is generated;

[0019] According to the one-dimensional sequence of the phase angle change with time, the instantaneous phase change rate is calculated through the numerical differentiation method, and the delay difference of different frequency components in the time response is analyzed, and the maximum phase lag time is obtained.

[0020] As a preferred solution of the pain management method based on bioelectrical impedance data and machine learning algorithm, wherein: the pain biomarker feature library is generated by spatiotemporal registration of the temperature field sequence and the maximum phase lag time and the impedance modulus change of the corresponding region through a multi-modal data fusion method, and the specific steps are as follows,

[0021] The maximum phase lag time is used as the time offset compensation amount to linearly interpolate the temperature field sequence to correct the temperature field timing;

[0022] The corrected temperature field timing is spatiotemporally registered with the impedance modulus change of the corresponding region through a multi-modal data fusion method to identify the mapping relationship of the electrode coordinates to the temperature image;

[0023] According to the mapping relationship of the impedance coordinates to the temperature image, the calculation of the thermo-electric coupling coefficient is performed for each electrode region to obtain statistical features of each electrode region and integrate and analyze them to generate the pain biomarker feature library.

[0024] As a preferred solution of the pain management method based on bioelectrical impedance data and machine learning algorithm, wherein: the pain biomarker feature library is generated by spatiotemporal registration of the temperature field sequence and the maximum phase lag time and the impedance modulus change of the corresponding region through a multi-modal data fusion method, and the specific steps are as follows,

[0025] The time and step of the sliding window are defined, the electrode regions in the pain biomarker feature library are read in chronological order, and the bioelectric signal features are extracted through sliding segmentation;

[0026] The bioelectric signal features are input into the machine learning pain assessment model to perform linear transformation and generate a bioelectric signal vector;

[0027] The bioelectric signal vector and the historical pain data are fused and analyzed to generate high-interaction fusion features, and cross-attention analysis is performed to generate a pain management triple.

[0028] As a preferred solution of the pain management method based on bioelectrical impedance data and machine learning algorithm, wherein: the pain management triple is compared and analyzed with historical clinical pain cases to generate a treatment intervention plan, and the specific steps are as follows,

[0029] The pain management triple is encoded and mapped to generate a numerical vector, and the cosine similarity is calculated to compare the similarity of the current numerical vector with each historical clinical pain case to obtain case matching information;

[0030] According to the case matching information, the average pain relief amplitude is calculated through the efficacy weighted average method, and the treatment intervention plan is generated in combination with the patient's file.

[0031] As a preferred solution of the pain management method based on bioelectrical impedance data and machine learning algorithm, wherein: the simulation treatment intervention is performed, the data of the pain management triplets before and after the treatment intervention are compared, and the effectiveness of the treatment intervention method is verified, and the specific steps are as follows,

[0032] According to the treatment intervention scheme, the simulation treatment intervention is performed, and the data of the pain management triplets after the treatment intervention are generated;

[0033] The data of the pain management triplets before and after the treatment intervention are compared and analyzed, the change difference of the data of the pain management triplets is obtained, and the change difference is compared with the human safety and health coefficient threshold, and the effectiveness of the treatment intervention method is verified.

[0034] As a preferred solution of the pain management method based on bioelectrical impedance data and machine learning algorithm, wherein: the dynamic optimization treatment intervention scheme is generated while the treatment suggestion is generated, and the specific steps are as follows,

[0035] According to the effectiveness result of the treatment intervention method, the effectiveness quantitative index is calculated by the multi-dimensional comprehensive evaluation method, and the pain management triplets are re-analyzed by fusing the traditional Chinese medicine meridian theory, so as to generate the dynamic optimization treatment intervention scheme and the dynamic treatment suggestion.

[0036] In the second aspect, the application provides a computer device, including a memory and a processor, and the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the pain management method based on bioelectrical impedance data and machine learning algorithm according to the first aspect of the application is realized.

[0037] In the third aspect, the application provides a computer readable storage medium, which stores a computer program, wherein: when the computer program is executed by the processor, any step of the pain management method based on bioelectrical impedance data and machine learning algorithm according to the first aspect of the application is realized.

[0038] The application has the beneficial effects that: by using the bidirectional LSTM and cross attention fusion method, the bioelectric signal features extracted by the sliding window and the historical pain data are subjected to time series modeling and cross-modal correlation analysis, the dynamic evolution law of pain development is captured, the pain management triplets including pain intensity, intervention response and prognosis evaluation are generated, and a quantitative basis for personalized treatment is provided. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0040] Fig. 1 Flowchart for pain management method based on bioelectrical impedance data and machine learning algorithm.

[0041] Fig. 2 Flowchart for phase delay analysis and multimodal data fusion.

[0042] Fig. 3 Flowchart for pain management triad generation.

[0043] Fig. 4 Flowchart for treatment intervention plan optimization. DETAILED DESCRIPTION

[0044] In order to make the above objectives, features and advantages of the present application more apparent and comprehensible, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0045] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0046] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0047] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides a pain management method based on bioelectrical impedance data and machine learning algorithm, comprising the following steps:

[0048] S1, collecting bioelectrical impedance data and infrared temperature field data and preprocessing to generate timestamp-aligned impedance phase spectrum matrix and temperature field sequence;

[0049] The bioelectrical impedance data includes impedance amplitude, phase angle and human body composition estimation value;

[0050] It should be noted that the collection process of bioelectrical impedance data is realized by a multi-band bioelectrical impedance analyzer. The device arranges an electrode array on the surface of the human body, applies a multi-band alternating current signal (usually 50 kHz-1 MHz) within a safe range, synchronously measures the impedance amplitude and phase angle (reflecting the time difference between the current signal and the voltage signal) of each electrode pair, and estimates the human composition parameters (such as muscle mass and fat percentage) according to the impedance spectrum characteristics in combination with the built-in algorithm.

[0051] The infrared temperature field data includes human body surface temperature distribution, spatial resolution temperature distribution, and time series temperature distribution.

[0052] It should be noted that the collection of infrared temperature field data is realized by a non-contact infrared thermal imager. The device scans the surface of the human body through a high-sensitivity infrared detector array, captures thermal radiation signals at different parts at a sampling rate of more than 30 frames per second, generates a thermal image containing spatial temperature distribution after blackbody radiation calibration and temperature-voltage conversion, and records the temperature change curve over time. All infrared temperature field data ensures the accuracy and consistency of the temperature field data through high-precision optical lenses and temperature calibration algorithms.

[0053] The preprocessing includes outlier rejection, filtering and noise reduction, phase spectrum smoothing, temperature correction, and non-uniformity correction.

[0054] Further, the preprocessing process first identifies and rejects outliers in impedance amplitude and phase angle (such as sudden data caused by poor electrode contact) through the 3σ criterion, then applies wavelet transform to multi-scale decomposition of bioelectrical impedance data, retaining the effective frequency band components reflecting physiological characteristics; Savitzky-Golay filter is applied to local polynomial fitting and smoothing of bioelectrical impedance data, eliminating high-frequency noise while retaining the true phase change trend; the infrared temperature field data is first corrected by environmental temperature compensation based on the blackbody radiation law, then the response difference between each pixel of the infrared detector is eliminated by the non-uniformity correction algorithm (NUC), and finally the spatial continuity and temporal stability of the infrared temperature field data are ensured by the spatio-temporal bilateral filter to eliminate random thermal noise.

[0055] The preprocessed bioelectrical impedance and infrared temperature field data are converted to a unified time reference, time-aligned and interpolated under the constraint of a reference clock, generating an impedance phase spectrum matrix and a temperature field sequence with timestamp alignment.

[0056] Further, the pre-processed bioelectrical impedance data and infrared temperature field data are synchronized by a time reference signal provided by a reference clock. The time series of bioelectrical impedance data recorded at a specified hertz frequency sampling rate is matched with the thermal image sequence of infrared temperature field data collected at a low frame rate. The infrared temperature field data is upsampled to the specified hertz frequency using a cubic spline interpolation method, while the bioelectrical impedance data is time-shift compensated based on the least squares criterion (e.g., reverse offset correction when the phase lag time is greater than 5 ms). Finally, an impedance phase spectrum matrix and a temperature field sequence with strictly aligned timestamps are generated.

[0057] It should be noted that the reference clock refers to a high-precision clock source serving as a time synchronization reference, such as a GPS disciplined atomic clock or a precision time protocol master clock, which can provide a unified time signal with nanosecond-level precision.

[0058] S2, phase delay analysis of the impedance phase spectrum matrix, calculating the rate of change of phase angle with time at each frequency point, and obtaining the maximum phase lag time;

[0059] The impedance phase spectrum matrix is divided into sliding segments according to a fixed time, the impedance phase angle feature is extracted, and the time is aligned to generate a one-dimensional sequence of phase angle changes over time;

[0060] Further, the impedance phase spectrum matrix is divided into sliding segments according to a fixed time window, the window is set to slide by a fixed time at each step, and the median of the phase angle in the specified frequency band (e.g., 10 kHz) in the impedance phase spectrum matrix is extracted as the impedance phase angle feature in each window. The timestamp corresponding to the center point of the window is paired with the impedance phase angle feature, and a one-dimensional sequence of phase angle changes over time is generated after arranging in time order. The time interval between adjacent data points in the sequence is equal to the sliding step.

[0061] According to the one-dimensional sequence of phase angle changes over time, the instantaneous phase change rate is calculated by numerical differentiation method, and the delay difference of different frequency components in time response is analyzed to obtain the maximum phase lag time.

[0062] Further, the one-dimensional sequence of phase angle changes over time is calculated by central difference method to obtain the phase change rate sequence corresponding to the one-dimensional sequence of phase angle changes over time. The phase change rate sequence is analyzed by cross-correlation analysis method, the frequency point time delay of each frequency point relative to different frequency components is calculated, and the peak position of the frequency point time delay is determined by Gaussian fitting. The time value corresponding to the maximum value is recorded to generate the maximum phase lag time.

[0063] It should be pointed out that different frequency components refer to the response data corresponding to multiple discrete frequency alternating current signals applied in bioelectrical impedance measurement, such as impedance phase angle measurement values at preset frequency points of 1 kHz, 5 kHz, 10 kHz, 50 kHz, 100 kHz, etc.

[0064] S3, according to the temperature field sequence and the maximum phase lag time, the impedance modulus change of the corresponding region is spatiotemporally registered by a multi-modal data fusion method to generate a pain biomarker feature library;

[0065] The maximum phase lag time is taken as a time offset compensation amount, and the temperature field sequence is linearly interpolated to correct the temperature field timing;

[0066] Further, the maximum phase lag time is taken as a time compensation reference amount, and the temperature field sequence is weighted and fused with the maximum phase lag time value to form a new reference time axis; for the time on the new reference time axis between two sampling points, a linear interpolation algorithm is used to calculate the temperature value of the position, the time interval of adjacent original frames is obtained, and the time proportion coefficient of the interpolation position relative to the previous frame is calculated, then the temperature values of each spatial position in the adjacent two frames of temperature field are weighted and summed according to the time proportion coefficient, and finally the corrected temperature field sequence with strictly aligned time axis is generated.

[0067] The corrected temperature field timing is spatiotemporally registered with the impedance modulus change of the corresponding region by a multi-modal data fusion method to identify the mapping relationship of the electrode coordinates to the temperature image;

[0068] Further, first, the geometric position matrix of the electrode array is established in the three-dimensional space coordinate system by a stereo vision calibration method, then the electrode coordinates are mapped to the two-dimensional image plane of the temperature field sequence by Delaunay triangulation, and the projection area in the temperature image corresponding to each electrode is calculated; for the projection area in the temperature image corresponding to the electrode, the average gradient value of the temperature field timing in the corresponding ROI region is extracted by Sobel operator convolution method, and dot product operation is performed with the impedance modulus change rate at the same time to generate a spatial registration coefficient matrix; finally, the parameters of the two-dimensional image plane of the temperature field sequence are adjusted by an iterative optimization algorithm, so that the main diagonal elements (such as the correlation of the electrode itself region) of the spatial registration coefficient matrix all reach the coefficient safety threshold (usually the value range is 0.85~1), and a stable sub-pixel level mapping relationship of the electrode coordinates to the temperature image is obtained.

[0069] It should be pointed out that the impedance modulus change refers to the magnitude of the impedance absolute value fluctuation of biological tissue with time under the action of a specific frequency alternating current, which is usually expressed as a percentage change rate relative to the baseline value, for example, when the impedance modulus changes from 50Ω to 52Ω at a frequency of 10 kHz, the change rate is +4%.

[0070] According to the mapping relationship of impedance coordinates to temperature images, the calculation of the thermal-electric coupling coefficient is performed for each electrode region, the statistical characteristics of each electrode region are obtained, and principal component analysis and multi-modal feature level fusion are integrated and analyzed to generate a pain biomarker feature library.

[0071] Further, based on the mapping relationship of impedance coordinates to temperature images, the thermal-electric coupling coefficient is calculated for each electrode region by spatial weighted average method; according to the calculation of the thermal-electric coupling coefficient of the corresponding region of the electrode, the time domain statistical characteristics and frequency domain characteristics are extracted for each electrode region by sliding segmentation, and weighted fusion is performed to form an initial fusion matrix; the principal component analysis method is used to reduce the dimension of the initial fusion matrix to the principal component, and then the multi-modal feature level fusion is performed with the synchronous acquisition of the phase lag time by canonical correlation analysis, to generate a pain biomarker feature library containing spatio-temporal correlation characteristics, each data entry in the feature library contains a timestamp, spatial coordinates and standardized feature value.

[0072] S4, extract bioelectric signal features from the pain biomarker feature library by time sliding, and input into the machine learning pain assessment model for cross-attention analysis to generate a pain management triple;

[0073] Define the time and step length of the sliding window, read the electrode regions in the pain biomarker feature library in chronological order, and extract bioelectric signal features by sliding segmentation;

[0074] Further, define the time length of the sliding window, read the electrode regions in the pain biomarker feature library in chronological order, and extract bioelectric signal features by sliding segmentation. First, determine the time length and step length parameters of the sliding window, for example, set the time length to 30 seconds and the step length to 5 seconds. Then, according to the set time length and step length, select the data of a specific electrode region from the pain biomarker feature library, and start sliding the window from the earliest time point backward. In each sliding window, extract all related bioelectric signal features, including impedance modulus, phase change rate, temperature and its change slope, etc. Specifically, in the first window, extract all data from the start time to the 30th second; then slide the window forward (for example, the data from the 5th second to the 35th second as the data set of the next window), and so on, until the entire time sequence is covered.

[0075] Input the bioelectric signal features into the machine learning pain assessment model, perform linear transformation, and generate a bioelectric signal vector;

[0076] Further, the bioelectric signal features are normalized to form a fixed-dimension numerical sequence, such as an electromyogram signal amplitude containing 128 time points, and a brain electrical power spectrum feature containing 64 frequency bands. The linear transformation layer of the machine learning pain assessment model receives the fixed-dimension numerical sequence, performs linear transformation, and generates a weighted sum of bioelectric signal features; finally, the weighted sum of the electrical signal features is encoded and mapped by the linear transformation method to generate a bioelectric signal vector.

[0077] It should be noted that the training process of the machine learning pain assessment model first pairs the bioelectric signal (such as EEG, EMG) and the clinical pain label for sampling. The output fusion parameters are predicted by the cross-entropy loss function to generate a pain category probability distribution; based on the pain category probability distribution, the gradient of the fusion parameters in the machine learning pain assessment model is optimized through the back propagation process to complete the training process.

[0078] The bioelectric signal vector and the historical pain data are fused and analyzed to generate high-interaction fusion features, and cross-attention analysis is performed to generate a pain management triple.

[0079] Further, first, the bioelectric signal vector and the historical pain data are dimensionally matched by the multi-modal feature alignment method, and the dimensionally matched bioelectric signal vector is calculated by cross-modal attention to generate high-interaction fusion features. Then, the high-interaction fusion features are processed by cross-attention analysis through bidirectional LSTM time step sliding segmentation, and finally the pain management triple is generated.

[0080] S5, compare the pain management triple with the historical clinical pain case, generate a treatment intervention scheme, and perform a simulation treatment intervention. Compare the pain management triple data before and after the treatment intervention to verify the effectiveness of the treatment intervention method, dynamically optimize the treatment intervention scheme, and generate a treatment suggestion.

[0081] The pain management triple is encoded and mapped by the data encoding and mapping method to generate a numerical vector, and the cosine similarity is calculated to compare the similarity between the current numerical vector and each historical clinical pain case to obtain case matching information.

[0082] Furthermore, firstly, the three components of the pain management triplet (pain intensity prediction, drug response assessment, and intervention suggestion coding) are concatenated using a data encoding mapping method to form a unified feature vector. Then, based on this unified feature vector, a two-layer fully connected network using the ReLU activation function performs a nonlinear mapping. The first layer performs dimensionality reduction, generating a low-dimensional feature vector; the second layer assigns values ​​to this low-dimensional feature vector using a numerical method, generating a numerical vector. Each case in the historical clinical pain case database is pre-processed using the same data encoding mapping method to generate a pain feature vector. Subsequently, cosine similarity is calculated to obtain the cosine similarity, and the current numerical vector is compared with the similarity of each historical clinical pain case, outputting case matching information.

[0083] The specific expression for calculating cosine similarity is:

[0084] ;

[0085] in, Represents a numerical vector; Indicates dimension; Indicates the first Pain feature vectors for each case; Represents the first numeric vector Dimensional value; Indicates the first The first of the pain feature vectors of the cases Dimensional value; This represents the cosine similarity.

[0086] Based on case matching information, the average pain relief level is calculated using the efficacy weighted average method, and a treatment intervention plan is generated by combining the patient's records.

[0087] Furthermore, based on case matching information, pain relief records of Top-K historical clinical pain cases were first obtained through time-domain analysis. Then, the average pain relief magnitude was calculated using a weighted average method with cosine similarity as the weight. This average pain relief magnitude was then analyzed in conjunction with contraindication records in the patient files to screen for eligible cases, ultimately generating a treatment intervention plan.

[0088] Based on the treatment intervention plan, the treatment intervention was simulated using feature mapping and numerical simulation methods to generate data of pain management triplet after the treatment intervention.

[0089] Further, the intervention type and parameters in the treatment intervention scheme are converted into numerical representation, for example, the frequency of TENS intervention is set to 100 Hz, the pulse width is 200 μs, and the duration is 30 minutes. Then the feature mapping is performed on the pain management triplets in the current state, and the pain intensity, pain pattern, and recommended intervention suggestions are respectively encoded into the pain management vector, for example, the pain intensity is represented by the scalar value 7.5, and the pain pattern is represented by the one-hot vector [0, 1, 0] indicating neuropathic pain. Then the pain management vector is integrated with the pain history data, and the numerical simulation method is used to predict the physiological change trend after intervention, for example, it is assumed that the impedance modulus decreases by 5%, the temperature rises by 0.8℃, and the phase lag decreases by 1.2 seconds. The new bioelectric signal features are reconstructed from the physiological change trend after intervention, and input into the machine learning pain assessment model again, and the new pain intensity, pain pattern, and possible maintenance or adjustment of intervention suggestions are output, and finally the data of the pain management triplets containing the predicted state are formed.

[0090] The data of the pain management triplets before and after the treatment intervention are compared and analyzed, the change difference of the data of the pain management triplets is obtained, and compared with the human body safety and health coefficient threshold to verify the effectiveness of the treatment intervention method;

[0091] Further, first, the pain intensity value, pain pattern category, and intervention suggestion type in the pain management triplet data before treatment intervention are obtained by primary visual observation method, and the corresponding items in the pain management triplet data after treatment intervention are obtained, for example, the pain intensity before treatment intervention is 7.5, and the pain intensity after treatment intervention is 5.2, and the difference between the two is 2.3. For the pain pattern, the category label is converted into numerical encoding, and the difference between the encoding values before and after intervention or the category change is calculated. The difference between the encoding values before and after intervention or the category change is compared with the human body safety and health coefficient threshold (for the safety threshold of electrophysiological signal, the value range is usually 10-35 μA / cm²; for the value range of the safety threshold of temperature field, it is usually 36.5-40.0℃), for example, the human body safety and health coefficient threshold stipulates that the pain intensity should decrease by more than or equal to 1.5 without adverse reactions to be considered effective. If the change difference of the pain management triplet data meets the requirement of the human body safety and health coefficient threshold, the treatment intervention method is determined to be effective.

[0092] According to the effectiveness result of the treatment intervention method, the effectiveness quantitative index is calculated by multi-dimensional comprehensive evaluation method, and the pain management triplets are analyzed again by integrating traditional Chinese meridian theory to generate dynamic optimization treatment intervention scheme and dynamic treatment suggestion.

[0093] Further, first, the pain intensity change amount, the pain pattern transition level and the intervention response time are obtained from the data of the pain management triad before and after the therapeutic intervention. The pain intensity change amount, the pain pattern transition level and the intervention response time are normalized and given corresponding weights. Then, the comprehensive effectiveness quantitative index is calculated by weighted summation. The current comprehensive effectiveness quantitative index is mapped to the traditional Chinese meridian map by regional mapping method for diagnosis; for example, the region with reduced impedance and increased temperature is located on both sides of the waist, corresponding to the path of the Foot-Taiyang Bladder Meridian, and combined with the phase lag characteristic, it is judged as "Qi stagnation and blood stasis syndrome". Based on the diagnosis result, the intervention strategy is adjusted, and the dynamic optimization treatment intervention scheme and the dynamic treatment suggestion are generated combined with the patient's physical information.

[0094] It should be noted that the above content has been agreed by the user and is for legal purposes.

[0095] The embodiment also provides a computer device suitable for the pain management method based on bioelectrical impedance data and machine learning algorithm, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the pain management method based on bioelectrical impedance data and machine learning algorithm proposed in the above embodiment.

[0096] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be realized through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0097] The embodiment also provides a storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the pain management method based on bioelectrical impedance data and machine learning algorithm proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0098] In summary, the application captures the dynamic evolution law of pain development by modeling and cross-modal correlation analysis of the bioelectric signal features and historical pain data extracted by the sliding window through the bidirectional LSTM and cross-attention fusion method, and generates a pain management triplet including pain intensity, intervention response, and prognosis evaluation, thereby providing a quantitative basis for personalized treatment.

[0099] It should be noted that the above embodiments are only used to illustrate the technical solutions of the application rather than limit the application. Although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the application, and all of them should be covered in the scope of the claims of the application.

Claims

1. A pain management method based on bioelectrical impedance data and machine learning algorithms, characterized in that: include, Bioelectrical impedance data and infrared temperature field data were collected and preprocessed to generate a timestamp-aligned impedance phase spectrum matrix and temperature field sequence. Phase delay analysis is performed on the impedance phase spectrum matrix to calculate the rate of change of phase angle with time at each frequency point and obtain the maximum phase lag time. Based on the temperature field sequence and the maximum phase lag time, spatiotemporal registration with the impedance modulus change of the corresponding region is performed using a multimodal data fusion method to generate a pain biomarker feature library. Bioelectrical signal features are extracted from a pain biomarker feature library by time sliding and input into a machine learning pain assessment model for cross-attention analysis to generate pain management triples. By comparing and analyzing the pain management triad with historical clinical pain cases, a treatment intervention plan is generated. Simulated treatment intervention is then conducted, and the data of the pain management triad before and after the treatment intervention are compared to verify the effectiveness of the treatment intervention method. The treatment intervention plan is dynamically optimized, and treatment suggestions are generated.

2. The pain management method based on bioelectrical impedance data and machine learning algorithms as described in claim 1, characterized in that: The bioelectrical impedance data includes impedance amplitude, phase angle, and estimated human body composition. The infrared temperature field data includes human body surface temperature distribution, spatial resolution temperature distribution, and time series temperature distribution; The preprocessing includes outlier removal, filtering and noise reduction, phase spectrum smoothing, temperature correction, and non-uniformity correction. The preprocessed bioelectrical impedance and infrared temperature field data are converted into a unified time reference. Time-aligned interpolation is performed under the constraint of a reference clock to generate a timestamp-aligned impedance phase spectrum matrix and temperature field sequence.

3. The pain management method based on bioelectrical impedance data and machine learning algorithms as described in claim 2, characterized in that: The specific steps for performing phase delay analysis on the impedance phase spectrum matrix, calculating the rate of change of phase angle with time at each frequency point, and obtaining the maximum phase lag time are as follows. The impedance phase spectrum matrix is ​​slide-segmented at fixed time intervals to extract impedance phase angle features, which are then aligned with time to generate a one-dimensional sequence of phase angle changes over time. Based on a one-dimensional sequence of phase angle changes over time, the instantaneous phase change rate is calculated using numerical differentiation, and the delay differences in time response of different frequency components are analyzed to obtain the maximum phase lag time.

4. The pain management method based on bioelectrical impedance data and machine learning algorithms as described in claim 3, characterized in that: The process involves generating a pain biomarker feature library by spatiotemporally registering the temperature field sequence and maximum phase lag time with the impedance modulus changes in the corresponding region using a multimodal data fusion method. The specific steps are as follows: The maximum phase lag time is used as the time offset compensation amount to perform linear interpolation on the temperature field sequence and correct the temperature field time series. The corrected temperature field time series and the impedance modulus change in the corresponding region are spatiotemporally registered using a multimodal data fusion method to identify the mapping relationship between electrode coordinates and temperature images. Based on the mapping relationship between impedance coordinates and temperature images, the thermo-electric coupling coefficient is calculated for each electrode region, the statistical characteristics of each electrode region are obtained, and integrated analysis is performed to generate a pain biomarker feature library.

5. The pain management method based on bioelectrical impedance data and machine learning algorithms as described in claim 4, characterized in that: The process involves extracting bioelectrical signal features from a pain biomarker feature library via time-sliding, inputting them into a machine learning pain assessment model for cross-attention analysis, and generating pain management triples. The specific steps are as follows: Define the time and step size of the sliding window, read the electrode regions in the pain biomarker feature library in chronological order, and perform sliding segmentation to extract bioelectrical signal features; The bioelectrical signal features are input into a machine learning pain assessment model, and a linear transformation is performed to generate a bioelectrical signal vector. The bioelectrical signal vectors and historical pain data are fused and analyzed to generate highly interactive fusion features, and cross-attention analysis is performed to generate pain management triples.

6. The pain management method based on bioelectrical impedance data and machine learning algorithms as described in claim 5, characterized in that: The process of comparing and analyzing the pain management triad with historical clinical pain cases to generate a treatment intervention plan involves the following steps: The pain management triples are encoded and mapped to generate numerical vectors. Cosine similarity is calculated to compare the similarity between the current numerical vector and each historical clinical pain case to obtain case matching information. Based on case matching information, the average pain relief level is calculated using the efficacy weighted average method, and a treatment intervention plan is generated by combining the patient's records.

7. The pain management method based on bioelectrical impedance data and machine learning algorithms as described in claim 6, characterized in that: The simulated treatment intervention was conducted, and the data of the pain management triplet before and after the intervention were compared to verify the effectiveness of the treatment intervention method. The specific steps are as follows: Based on the treatment intervention plan, simulated treatment intervention is conducted to generate pain management ternary set data after the treatment intervention; The data of the pain management tripartite group before and after the treatment intervention were compared and analyzed to obtain the difference in the change of the pain management tripartite group data, and compared with the human safety and health coefficient threshold to verify the effectiveness of the treatment intervention method.

8. The pain management method based on bioelectrical impedance data and machine learning algorithms as described in claim 7, characterized in that: The dynamically optimized treatment intervention plan simultaneously generates treatment suggestions, and the specific steps are as follows. Based on the effectiveness results of the treatment intervention methods, a multi-dimensional comprehensive evaluation method was used to calculate the quantitative indicators of effectiveness. The pain management ternary reanalysis was then conducted by integrating the theory of meridians in traditional Chinese medicine to generate a dynamically optimized treatment intervention plan and dynamic treatment recommendations.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the pain management method based on bioelectrical impedance data and machine learning algorithms as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the pain management method based on bioelectrical impedance data and machine learning algorithms as described in any one of claims 1 to 8.

Images