Children myopia traditional Chinese medicine constitution typing method and system based on RPPG technology

Abstract

The invention discloses a children myopia traditional Chinese medicine constitution typing method and system based on an RPPG technology. The method comprises the following steps: acquiring basic data and face video data of a myopia patient; extracting an RPPG signal from the face video data; extracting a heart rate variability (HRV) feature based on the processed RPPG signal; performing physique identification on the child myopia patient through a physique identification tool to obtain a traditional Chinese medicine physique identification result; combining the basic data, the HRV characteristics and the traditional Chinese medicine constitution identification result to construct an intelligent typing model; and classifying the traditional Chinese medicine constitution of the child myopia patient to be tested by using the intelligent typing model to obtain a traditional Chinese medicine constitution typing result. Facial videos are collected through an RPPG technology to extract HRV features, a model is constructed in combination with basic data and traditional Chinese medicine constitution identification, intelligent classification of children's myopia traditional Chinese medicine constitution is achieved, objective typing basis is provided for clinic, precise prevention and control are assisted, and constitution identification efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of TCM constitution identification, and in particular to a TCM constitution typing method and system for children's myopia based on RPPG technology. Background Art

[0002] Traditional Chinese Medicine believes that physical characteristics influence disease susceptibility, and different physical constitutions manifest differently during health and disease. Childhood myopia is a common eye disease in my country. In 2018, the overall myopia rate among children and adolescents reached 53.6%. Without timely intervention, serious complications can occur. Current predictive applications for the classification and determination of children's TCM constitutions primarily rely on literature research, interviews with insiders, and expert questionnaires to determine constitution types, compile identification tables, and establish identification methods. However, these methods have significant shortcomings: the results are highly subjective, the operations are cumbersome and complex, and they lack the support of objective, quantitative physiological indicators. This makes it difficult to accurately reflect the physical characteristics of children with myopia, limiting the application of TCM constitution classification in the prevention and control of childhood myopia. How to leverage objective physiological signal acquisition and intelligent analysis technology to achieve automated and precise TCM constitution classification for children with myopia, providing a scientific basis for clinical dialectical treatment, has become a technical issue that needs to be addressed urgently. Summary of the Invention

[0003] In view of this, the present invention proposes a method and system for TCM constitution classification of children with myopia based on RPPG technology, which can analyze the TCM constitution classification of children with myopia through RPPG signals combined with HRV characteristics. The present invention provides the following technical solutions:

[0004] A method for TCM constitution classification of children's myopia based on RPPG technology, comprising:

[0005] Obtaining basic data and facial video data of children with myopia, wherein the basic data includes age, gender, myopia condition, and refractive power data;

[0006] Extracting an RPPG signal from the facial video data, and performing noise reduction processing on the RPPG signal to obtain a processed RPPG signal;

[0007] extracting heart rate variability (HRV) features based on the processed RPPG signal;

[0008] Using a constitution identification tool to identify the constitution of the child myopic patient, and obtaining a TCM constitution identification result;

[0009] Building an intelligent typing model by combining the basic data, the HRV characteristics and the TCM constitution identification results;

[0010] The intelligent classification model is used to classify the TCM constitution of the tested child myopia patients to obtain a TCM constitution classification result.

[0011] Optionally, extracting the RPPG signal from the facial video data and performing noise reduction processing on the RPPG signal to obtain the processed RPPG signal includes:

[0012] Denoising the single-channel pixels of the facial video to obtain a single-channel pixel light change curve of the face;

[0013] Based on the single-channel pixel light change curve, the FastICA algorithm is used to separate the source signal in the signal, and the source signal with the highest Pearson coefficient with the green channel is extracted as the RPPG signal;

[0014] The high-frequency noise and baseline drift of the RPPG signal are removed by wavelet transform, the pulse wave frequency in the signal is obtained based on fast Fourier transform, and a narrow bandpass filter is set to reconstruct the signal to obtain a processed RPPG signal.

[0015] Optionally, extracting heart rate variability (HRV) features based on the processed RPPG signal includes:

[0016] Performing cubic spline interpolation on the processed RPPG signal and up-sampling it to a preset frequency, extracting the peak point to obtain the RR interval sequence;

[0017] After removing outliers from the RR interval sequence, an HRV time domain curve is drawn, and time domain characteristic parameters of the time domain curve are extracted using a statistical method;

[0018] The RPPG signal is analyzed in the frequency domain by Welch power spectrum diagram to extract the frequency domain characteristic parameters;

[0019] The nonlinear characteristic parameters of the processed RPPG signal are extracted based on the Poincare scatter plot analysis method.

[0020] Optionally, the time domain characteristic parameter includes one of the standard deviation of the RR interval sequence, the proportion of adjacent RR interval differences exceeding a preset threshold, the root mean square of adjacent RR interval differences, the standard deviation of the RR interval differences, the coefficient of variation of the continuous difference, the coefficient of variation of the RR interval, and the standard deviation of the heart rate;

[0021] The frequency domain characteristic parameters include at least one of low frequency power (LF), high frequency power (HF), and LF / HF ratio;

[0022] The nonlinear characteristic parameter includes at least one of the major axis, minor axis, and sample entropy of the scatter plot ellipse.

[0023] Optionally, the constitution identification tool is the Children's Traditional Chinese Medicine Body Mass Scale;

[0024] The constitution identification tool is used to identify the constitution of the child myopic patient, and the TCM constitution identification result is obtained, which includes:

[0025] Based on the "Children's Traditional Chinese Medicine Body Mass Scale", a questionnaire survey or data collection was conducted on children with myopia to identify the nine major body constitution types in Traditional Chinese Medicine, including balanced body constitution, qi deficiency body constitution, yang deficiency body constitution, yin deficiency body constitution, phlegm-damp body constitution, damp-heat body constitution, blood stasis body constitution, qi stagnation body constitution and special body constitution;

[0026] The TCM constitution identification result includes the category of each constitution type and the corresponding identification score or probability value.

[0027] Optionally, the constructing of an intelligent typing model by combining the basic data, the HRV characteristics and the TCM constitution identification results includes:

[0028] The basic data and HRV characteristics were used as dependent variables, and the results of TCM constitution identification were used as independent variables, and the Logistic regression method was used to construct a discriminant model;

[0029] Based on the discriminant model, characteristic variables were screened by stepwise regression method, and the partial regression coefficient and odds ratio of each variable were calculated;

[0030] A constitution classification probability function is constructed based on the characteristic variables, partial regression coefficients and odds ratios to construct an intelligent typing model.

[0031] The present invention further discloses a TCM constitution classification system for children with myopia based on RPPG technology, comprising:

[0032] A data processing module is used to obtain basic data and facial video data of children with myopia, wherein the basic data includes age, gender, myopia condition and refractive power data;

[0033] a signal extraction module, configured to extract an RPPG signal from the facial video data and perform noise reduction processing on the RPPG signal to obtain a processed RPPG signal;

[0034] a feature extraction module, configured to extract heart rate variability (HRV) features based on the processed RPPG signal;

[0035] A model building module is used to identify the constitution of the child myopic patient using a constitution identification tool to obtain a TCM constitution identification result;

[0036] The model building module is also used to build an intelligent typing model by combining the basic data, the HRV characteristics and the TCM constitution identification results;

[0037] The constitution classification module is used to classify the TCM constitution of the child myopia patient to be tested using the intelligent classification model to obtain the TCM constitution classification result.

[0038] The present invention further discloses a computer-readable storage medium, characterized in that the storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0039] The present invention further discloses an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the program.

[0040] The present invention further discloses a computer program product, comprising a computer program, which implements the above method when executed by a processor.

[0041] According to the technical solution of the present invention, by obtaining basic data and facial video data of child myopia patients including age, gender, myopia status and refractive power, multi-dimensional information support is provided for subsequent analysis, RPPG signals are extracted and noise-reduced from facial videos to ensure the purity of physiological signals to accurately reflect the characteristics of cardiac activity, HRV features are extracted based on the processed signals, and the functional state of the autonomic nervous system is comprehensively characterized through time domain, frequency domain and nonlinear analysis. Combined with the TCM constitution identification results of the constitution identification tool, an intelligent classification model integrating objective physiological indicators and TCM constitution theory is constructed using methods such as Logistic regression, ultimately realizing the automatic classification of the TCM constitution of the tested patients, overcoming the defects of traditional TCM constitution classification, which is highly subjective and cumbersome to operate. Through RPPG non-contact signal acquisition and quantitative analysis of HRV features, an objective and scientific basis for constitution classification is provided for child myopia patients, which is conducive to dialectical treatment and precise intervention in myopia prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] For purposes of illustration and not limitation, the present invention will now be described with reference to embodiments thereof and the accompanying drawings, in which:

[0043] Figure 1 1 is a flow chart of a method for TCM constitution classification of children with myopia based on RPPG technology in an embodiment of the present invention;

[0044] Figure 2 Schematic diagram of the composition structure of the TCM constitution classification system for children with myopia based on RPPG technology in an embodiment of the present invention;

[0045] Figure 3 Schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of this application.

[0047] It should be noted that, in the absence of conflict, the embodiments of the present application and the features thereof can be combined with each other. The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0048] refer to Figure 1 This embodiment discloses a method for classifying children's myopia constitution in Traditional Chinese Medicine based on RPPG technology, which includes the following steps:

[0049] S100: Obtain basic data and facial video data of a child myopic patient, wherein the basic data includes age, gender, myopia condition, and refractive power data.

[0050] In this embodiment, basic data is first collected. The child's date of birth and gender are obtained through a questionnaire or electronic medical record system. Uncorrected and corrected visual acuity are measured using a standard logarithmic visual acuity chart. Myopia duration, progression rate, and family history are recorded. Initial refraction is performed using a fully automated computerized ophthalmometer. Children of different age groups are treated with appropriate mydriatic medications and re-refractions are performed to obtain true refractive power data and calculate spherical equivalent power. Subsequently, facial video data collection is performed to obtain facial video data. For example, a high-definition camera with a resolution of at least 1920×1080 and a frame rate of at least 30 fps is used in a uniform natural lighting environment of 500-1000 lux. Three videos, each lasting at least 30 seconds, are recorded simultaneously with environmental parameters. After collection, the basic data is stored in a structured database and linked. The facial videos are saved in a lossless format and metadata is added. The videos are then quality-assessed to remove problematic segments. This ensures the accuracy of basic data and the high quality of facial videos through standardized operating procedures. Combined with medical professional methods, the authenticity of refractive data is guaranteed, providing a reliable data source for subsequent RPPG signal extraction and improving the reliability of model input.

[0051] S200: Extracting an RPPG signal from the facial video data and performing noise reduction on the RPPG signal to obtain a processed RPPG signal. Specifically, the Haar cascade classifier in the OpenCV library is first used to perform face detection on the captured facial video, locating key feature points such as the eyes and nose tip. This allows the extraction of a facial region of interest (ROI) encompassing the forehead, cheeks, and other areas to reduce background interference. After converting the ROI to YCbCr color space, the green channel is selected and the pixel values ​​of the green channel are subjected to joint spatiotemporal denoising. A 3×3 median filter is first used to remove salt and pepper noise, followed by a 5-frame sliding average filter to eliminate temporal jitter, resulting in a preliminary single-channel pixel light change curve. The FastICA algorithm is then used to perform blind source separation on the curve. By setting the iterative convergence threshold to 1e-6, the signal is decomposed into multiple independent source signals. The Pearson correlation coefficient between each source signal and the original green channel is calculated, and the source signal with the largest absolute value of the correlation coefficient is selected as the initial RPPG signal. To further purify the signal, the initial RPPG signal was subjected to a five-layer wavelet decomposition using the Daubechies wavelet (db4). Soft thresholds were set within each layer's coefficients to remove high-frequency noise components, and baseline drift was removed through polynomial fitting. The denoised signal was spectrally analyzed using a fast Fourier transform. The peak frequency within the 0.5-4 Hz range was identified as the pulse wave fundamental frequency. A fourth-order Butterworth narrow-bandpass filter was designed based on this fundamental frequency, and the signal was reconstructed to obtain a pure RPPG signal. This multi-stage noise reduction process effectively suppressed interfering factors such as ambient light fluctuations and facial micro-expressions, increasing the signal-to-noise ratio of the RPPG signal to over 15 dB, providing a high-quality physiological signal for the subsequent accurate extraction of HRV features.

[0052] S300: Extract heart rate variability (HRV) features based on the processed RPPG signal. First, the processed RPPG signal is upsampled using a cubic spline interpolation algorithm, and the sampling rate is increased to 240Hz to meet the requirements of high-frequency feature extraction. The interpolation process is implemented through the spline function of MATLAB and a continuous and smooth signal curve is generated. Subsequently, the Pan-Tompkins algorithm is used to detect the peak point, and the dynamic threshold coefficient is set to 1.6 times the average amplitude to accurately identify the R wave peak of each cardiac cycle. The time interval between adjacent peaks is calculated to obtain the RR interval sequence. For the RR interval sequence, the Z-score method is used to eliminate outliers that deviate from the mean by more than 3 times the standard deviation. The outlier detection and correction are implemented through the scipy.stats module of Python. After ensuring the reliability of the sequence, the HRV time domain curve is drawn.

[0053] In the time domain feature extraction stage, based on the corrected RR interval sequence, multiple feature parameters are calculated using the numpy library function: the standard deviation function is used to calculate SDNN (standard deviation of normal sinus beat intervals) to characterize the degree to which heart rate variability deviates from the mean value. The proportion of adjacent RR interval differences exceeding 50ms is calculated to obtain PNN50, and the root mean square function is used to solve RMSSD (root mean square of adjacent RR interval differences) to characterize the overall change in HRV. The standard deviation of the difference sequence is calculated to obtain SDSD, and CVSD is further obtained by the ratio of SDSD to the average RR interval. CVnni is calculated by the ratio of SDNN to the average RR interval. At the same time, the RR interval is converted into a heart rate sequence and std_hr (heart rate standard deviation) is calculated. All parameter calculations retain four decimal places to ensure accuracy. Among them, the standard deviation (SDNN) of the RR interval sequence is calculated to indicate the degree to which heart rate variability deviates from the mean value. The formula is: in, is the average value of the RR interval of N heartbeats, RR i is the i-th RR interval;

[0054] The proportion of adjacent RR interval differences exceeding the preset threshold (PNN) was calculated to reflect the activity of the parasympathetic nerves. The formula is: Where NN50 is the number of adjacent RR interval differences exceeding the preset threshold of 50ms, and N is the total number of heartbeats;

[0055] The root mean square difference (RMSSD) of adjacent RR intervals was calculated to measure the regulatory effect of the parasympathetic nervous system on heart rate. The formula is:

[0056] The standard deviation of the RR interval differences (SDSD) was calculated to provide a more detailed representation of the overall change in HRV using the formula: Among them, RR′ i =RR i -RR i+1 ,

[0057] The coefficient of variation of consecutive differences (CVSD) was calculated using the formula:

[0058] The coefficient of variation of the RR interval (CVnni) was calculated using the following formula:

[0059] Calculate the standard deviation of heart rate (std_hr) using the formula: Among them, HR i is the i-th early heart rate,

[0060] For frequency domain analysis, after converting the RR interval series into a heart rate time series, the scipy.signal.welch function was used to estimate the power spectrum. The Hanning window length was set to 256 points, and the overlap rate of adjacent segments was 50%. The low-frequency power (LF) and high-frequency power (HF) were calculated in the frequency range of 0.04-0.4Hz. The LF / HF parameter was obtained by the ratio of the two, where LF corresponds to the 0.04-0.15Hz frequency band and HF corresponds to the 0.15-0.4Hz frequency band. The spectrum was normalized to 1 to eliminate the influence of individual differences. Frequency domain analysis can reflect the activities of the sympathetic and parasympathetic nerves, as well as the balance of the autonomic nervous system. The formula is as follows: In this formula, f(λ) is the power spectrum of the RR interval, is the ratio of the low-frequency to high-frequency power spectra.

[0061] When extracting nonlinear features, a Poincare scatter plot is constructed based on the RR interval sequence. Scatter plots are drawn with the current RR interval as the horizontal axis and the next RR interval as the vertical axis. The major axis SD2 and minor axis SD1 are calculated using an ellipse fitting algorithm to represent the long-term trend and short-term fluctuation of heart rate variability, respectively. The calculation formula is: The sample entropy algorithm is used to calculate the complexity of the sequence. The embedding dimension m = 2 and the threshold r = 0.15 times the standard deviation are set. The vector distance and probability value are calculated cyclically to obtain the sample entropy value. The entire calculation process is implemented through iterative operations using a custom Python function to ensure the accuracy of nonlinear features. The calculation of sample entropy includes: for a time series composed of N known data, denoted as {x(n)}, an m-dimensional vector X is constructed according to the sequence number. m (i): X(i) = {x(i), x(i+1), ..., x(i+m-1)}, 1≤i≤N-m+1, representing the m consecutive values ​​of x starting from i. Further calculation of vector X m (i) and X m (j), denoted as d[X m (i),X m (j)]:d[X m (i), X m (j)]=max k＝0，...，m-1 {|x(i+k)-x(j+k)|}. Further, the statistical distance d[X m (i),X m The number of (j)]≤r is denoted as N m (i), where r is the set threshold. The probability is calculated as follows: Calculate the mean: Finally, increase the dimension to m+1 and calculate the probability in the above steps and mean B m+1(r). Sample entropy is defined as: If N is a finite value, it can be expressed as

[0062] S400: Identifying the constitution of the child myopic patient using a constitution identification tool to obtain a Traditional Chinese Medicine constitution identification result.

[0063] The above-mentioned children were identified using the "Children's Traditional Chinese Medicine Body Type Scale" to identify their constitutions, resulting in nine TCM body type identification results and categories. For example, 5,969 children aged 6 to 18 were selected as research subjects. The "Children's Traditional Chinese Medicine Body Type Scale" was used to identify their constitutions, and visual acuity and refraction tests were performed. The results showed that the frequency of TCM constitutions among the 5,969 children was as follows: balanced constitution (39.81%), qi deficiency constitution (20.15%), yin deficiency constitution (17.88%), yang deficiency constitution (7.49%), qi stagnation constitution (7.19%), damp-heat constitution (3.03%), phlegm-damp constitution (2.48%), blood stasis constitution (1.21%), and special constitution (0.77%). Significant differences in visual acuity and refraction were observed among samples with different TCM body types (P < 0.01). The constitution with the highest visual acuity is the balanced constitution, with an average visual acuity of 4.78±0.32; the constitution with the lowest visual acuity is the yang deficiency constitution, with an average visual acuity of 4.59±0.38; the constitution with the lowest refractive power is the balanced constitution, with an average refractive power of -1.23±1.82D; the constitution with the highest refractive power is the yang deficiency constitution, with an average refractive power of -2.25±2.26D. Children with a balanced constitution have better visual acuity and lower myopia, while children with a yang deficiency constitution have worse visual acuity and higher myopia. The degree of myopia varies among children with different constitutions, and improving constitution may be considered to prevent and control myopia progression. For example, using an epidemiological case-control design, a questionnaire survey was conducted among 492 myopic patients and 450 non-myopic patients, and their constitutions were classified. The results showed that there were significant differences in the constitution ratios between the two groups (p<0.01). The myopic group had significantly higher rates of yang deficiency, phlegm-dampness, and damp-heat constitutions than the non-myopic group, while the non-myopic group had higher rates of yin deficiency than the myopic group. There were also significant differences in the constitution ratios between different degrees of myopia (p<0.001). The high myopia group had significantly higher rates of yin deficiency and blood stasis constitutions than the moderate and mild myopia groups. There was also a significant difference in the heritability of myopia between the two groups (p<0.001). The results showed that the constitution characteristics of myopic patients are yang deficiency, phlegm-dampness, and damp-heat constitutions, which are consistent with the pathogenesis of myopia: the constitution of myopic patients has a significant impact on the occurrence, development, and outcome of the disease.

[0064] S500: Construct an intelligent typing model in combination with the basic data, the HRV characteristics and the TCM constitution identification results. Specifically, first pre-process the basic data, HRV characteristics and TCM constitution identification results, convert the age into a numerical variable, encode the gender into 0-1 (0 for female, 1 for male), keep the myopia and refractive power data in numerical form, retain three decimal places for HRV characteristics, and use TCM constitution type as a categorical variable. Divide the pre-processed data set into a training set and a test set in a ratio of 7:3, and use stratified sampling to ensure that the distribution ratio of each TCM constitution type in the training set and the test set is consistent. Use the basic data and HRV characteristics as the dependent variable matrix X, and the TCM constitution type as the independent variable vector Y to construct a Logistic regression model. During the training process, the stepwise regression method was used for feature screening. First, all features were included in the model. The significance of each feature was evaluated through the Wald test. Insignificant features were gradually eliminated. Finally, the optimal subset of 12 key features including age, equivalent spherical power, SDNN, RMSSD, LF / HF, SD2 / SD1, and sample entropy was determined. The model parameters were estimated using the training set data, and the maximum likelihood method was used to solve the regression coefficient to obtain the partial regression coefficient β of each feature. i And odds ratio (OR value). Further construct the probability function of constitution classification: where X i is the characteristic variable value, N is the sum of characteristic variables, and the intelligent classification model is constructed based on this.

[0065] S600: Using the intelligent classification model to classify the TCM constitution of the child myopia patient to be tested, to obtain a TCM constitution classification result.

[0066] First, basic data collection and facial video recording were performed on the children undergoing the test. The process was consistent with the training phase, acquiring basic data such as age, gender, and diopter, along with facial videos. The same RPPG signal extraction and HRV feature extraction processes were then used to obtain a parameter matrix containing time-domain, frequency-domain, and nonlinear features. The basic data and HRV features were preprocessed, and numerical variables were normalized using the same standardization methods used in the training phase. Categorical variables, such as gender, were converted to 0-1 encoding to form a feature vector that met the model input requirements.

[0067] The standardized eigenvector is input into the intelligent classification model, and the model's built-in probability function is used to calculate the probability of the sample belonging to one of the nine TCM constitution types. For example, after inputting the eigenvector, the model outputs [0.12, 0.08, 0.25, 0.15, 0.07, 0.06, 0.04, 0.11, 0.12], corresponding to the probabilities of belonging to the neutral constitution to the special constitution, respectively. The constitution type corresponding to the highest probability is selected as the preliminary judgment result, and the probability values ​​of each constitution are recorded to form a confidence matrix.

[0068] The preliminary results are verified using clinical rules. If the probability of a balanced constitution is ≥0.6 and all other constitutions are <0.4, the patient is directly diagnosed as a balanced constitution. If there are multiple constitutions with probabilities ≥0.4, they are ranked from high to low by probability, and supplemented by clinical experience (for example, when the probabilities of yang deficiency and qi deficiency are similar, a comprehensive judgment is made based on refractive data and SDNN features). A classification report is ultimately generated, including the constitution type, the probability values ​​of each constitution, and the influence weights of key features. For example, "Determination result: Yang deficiency constitution (probability 0.38), ranking of influencing factors: LF / HF (weight 32%) > equivalent spherical power (27%) > sample entropy (19%)."

[0069] In order to ensure the reliability of the classification results, a model credibility threshold is set. When the highest probability is less than 0.35, it is marked as "uncertain constitution". It is recommended to re-collect data or combine it with a face-to-face consultation with traditional Chinese medicine for review. The system records the input features and output results of each classification in real time, accumulates them into a new data set, and regularly uses incremental learning algorithms to update the model. Through ten-fold cross-validation, it is ensured that the accuracy of the model on the newly added data remains above 80%. This implementation method realizes the automated mapping from physiological signals to traditional Chinese medicine constitution types through standardized data processing procedures and model application mechanisms. The single classification processing time is less than 2 minutes, which meets the clinical rapid diagnosis needs and provides a quantitative reference basis for the dialectical treatment of childhood myopia.

[0070] refer to Figure 2 This embodiment further discloses a TCM constitution classification system for children with myopia based on RPPG technology, including a data processing module, a signal extraction module, a feature extraction module, a model building module and a constitution classification module, which are described in detail below:

[0071] The data processing module 21 is used to obtain basic data and facial video data of child myopia patients, wherein the basic data includes age, gender, myopia condition and refractive power data.

[0072] The signal extraction module 22 is used to extract the RPPG signal from the facial video data and perform noise reduction processing on the RPPG signal to obtain a processed RPPG signal, including: performing noise reduction on the single-channel pixels of the facial video to obtain a single-channel pixel light change curve of the face; based on the single-channel pixel light change curve, using the FastICA algorithm to separate the source signal in the signal, and extracting the source signal with the highest Pearson coefficient with the green channel as the RPPG signal; removing high-frequency noise and baseline drift from the RPPG signal through wavelet transform, obtaining the pulse wave frequency in the signal based on fast Fourier transform, and setting a narrow bandpass filter to reconstruct the signal to obtain the processed RPPG signal.

[0073] The feature extraction module 23 is used to extract heart rate variability (HRV) features based on the processed RPPG signal, including: performing cubic spline interpolation on the processed RPPG signal and upsampling it to a preset frequency, extracting peak points to obtain an RR interval sequence; drawing an HRV time domain curve after removing outliers from the RR interval sequence, and extracting time domain feature parameters of the time domain curve using statistical methods; performing frequency domain analysis on the RPPG signal through a Welch power spectrum diagram to extract frequency domain feature parameters; and extracting nonlinear feature parameters of the processed RPPG signal based on the Poincare scatter plot analysis method.

[0074] The model building module 24 is used to identify the constitution of the child myopia patient through the constitution identification tool to obtain the TCM constitution identification result, including: conducting a questionnaire survey or data collection on the child myopia patient based on the "Child TCM Constitution Scale" to identify the nine major TCM constitution types, the nine major constitution types including balanced constitution, qi deficiency constitution, yang deficiency constitution, yin deficiency constitution, phlegm-damp constitution, damp-heat constitution, blood stasis constitution, qi stagnation constitution and special constitution; the TCM constitution identification result includes the category of each constitution type and the corresponding identification score or probability value.

[0075] The model construction module 24 is also used to construct an intelligent typing model in combination with the basic data, the HRV characteristics and the TCM constitution identification results, including: using the basic data and HRV characteristics as dependent variables and the TCM constitution identification results as independent variables, and using the Logistic regression method to construct a discriminant model; based on the discriminant model, screening characteristic variables through the stepwise regression method, calculating the partial regression coefficient and odds ratio of each variable; constructing a constitution classification probability function based on the characteristic variables, partial regression coefficients and odds ratios to construct an intelligent typing model.

[0076] The constitution classification module 25 is used to classify the TCM constitution of the child myopia patient to be tested using the intelligent classification model to obtain a TCM constitution classification result.

[0077] Figure 3 A schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, such as Figure 3 As shown, the electronic device 50 includes: a processor 501 (processor), a memory 502 (memory) and a bus 503;

[0078] The processor 501 and the memory 502 communicate with each other via the bus 503 ; the processor 501 is used to call program instructions in the memory 502 to execute the methods provided by the above-mentioned method implementation methods.

[0079] This embodiment provides a non-transitory computer-readable storage medium, which stores computer instructions. The computer instructions enable a computer to execute the methods provided by the above-mentioned method embodiments.

[0080] Those skilled in the art will understand that all or part of the steps for implementing the above-mentioned method implementation method can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method implementation method; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk, etc. Various storage media that can store program codes.

[0081] The device embodiments described above are merely illustrative. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of these modules may be selected based on actual needs to achieve the objectives of this embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0082] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or certain parts of the embodiment.

[0083] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for classifying children's myopia in Traditional Chinese Medicine based on RPPG technology, characterized in that: include: Obtaining basic data and facial video data of children with myopia, wherein the basic data includes age, gender, myopia condition, and refractive power data; Extracting an RPPG signal from the facial video data, and performing noise reduction processing on the RPPG signal to obtain a processed RPPG signal; extracting heart rate variability (HRV) features based on the processed RPPG signal; Using a constitution identification tool to identify the constitution of the child myopic patient, and obtaining a TCM constitution identification result; Building an intelligent typing model by combining the basic data, the HRV characteristics and the TCM constitution identification results; The intelligent classification model is used to classify the TCM constitution of the tested child myopia patients to obtain a TCM constitution classification result.

2. The method for Chinese medicine constitution typing of children's myopia based on RPPG technology according to claim 1, characterized in that: The extracting the RPPG signal from the facial video data and performing noise reduction processing on the RPPG signal to obtain the processed RPPG signal includes: Denoising the single-channel pixels of the facial video to obtain a single-channel pixel light change curve of the face; Based on the single-channel pixel light change curve, the FastICA algorithm is used to separate the source signal in the signal, and the source signal with the highest Pearson coefficient with the green channel is extracted as the RPPG signal; The high-frequency noise and baseline drift of the RPPG signal are removed by wavelet transform, the pulse wave frequency in the signal is obtained based on fast Fourier transform, and a narrow bandpass filter is set to reconstruct the signal to obtain a processed RPPG signal.

3. The method for TCM constitution typing of children with myopia based on RPPG technology according to claim 1, characterized in that: The extracting of heart rate variability (HRV) features based on the processed RPPG signal includes: Performing cubic spline interpolation on the processed RPPG signal and up-sampling it to a preset frequency, extracting the peak point to obtain the RR interval sequence; After removing outliers from the RR interval sequence, an HRV time domain curve is drawn, and time domain characteristic parameters of the time domain curve are extracted using a statistical method; The RPPG signal is analyzed in the frequency domain by Welch power spectrum diagram to extract the frequency domain characteristic parameters; The nonlinear characteristic parameters of the processed RPPG signal are extracted based on the Poincare scatter plot analysis method.

4. The method for TCM constitution typing of children with myopia based on RPPG technology according to claim 3, characterized in that: The time domain characteristic parameters include one of the standard deviation of the RR interval sequence, the proportion of adjacent RR interval differences exceeding a preset threshold, the root mean square of adjacent RR interval differences, the standard deviation of the RR interval differences, the coefficient of variation of the continuous difference, the coefficient of variation of the RR interval, and the standard deviation of the heart rate; The frequency domain characteristic parameters include at least one of low frequency power (LF), high frequency power (HF), and LF / HF ratio; The nonlinear characteristic parameter includes at least one of the major axis, minor axis, and sample entropy of the scatter plot ellipse.

5. The method for TCM constitution typing of children with myopia based on RPPG technology according to claim 1, characterized in that: The constitution identification tool is the Traditional Chinese Medicine Body Mass Scale for Children; The constitution identification tool is used to identify the constitution of the child myopic patient, and the TCM constitution identification result is obtained, which includes: Based on the "Children's Traditional Chinese Medicine Body Mass Scale", a questionnaire survey or data collection was conducted on children with myopia to identify the nine major body constitution types in Traditional Chinese Medicine, including balanced body constitution, qi deficiency body constitution, yang deficiency body constitution, yin deficiency body constitution, phlegm-damp body constitution, damp-heat body constitution, blood stasis body constitution, qi stagnation body constitution and special body constitution; The TCM constitution identification result includes the category of each constitution type and the corresponding identification score or probability value.

6. The method for TCM constitution typing of children with myopia based on RPPG technology according to claim 1, characterized in that: The construction of the intelligent typing model by combining the basic data, the HRV characteristics and the TCM constitution identification results includes: The basic data and HRV characteristics were used as dependent variables, and the results of TCM constitution identification were used as independent variables, and the Logistic regression method was used to construct a discriminant model; Based on the discriminant model, characteristic variables were screened by stepwise regression method, and the partial regression coefficient and odds ratio of each variable were calculated; A constitution classification probability function is constructed based on the characteristic variables, partial regression coefficients and odds ratios to construct an intelligent typing model.

7. A TCM constitution classification system for children with myopia based on RPPG technology, characterized by: include: A data processing module is used to obtain basic data and facial video data of children with myopia, wherein the basic data includes age, gender, myopia condition and refractive power data; a signal extraction module, configured to extract an RPPG signal from the facial video data and perform noise reduction processing on the RPPG signal to obtain a processed RPPG signal; a feature extraction module, configured to extract heart rate variability (HRV) features based on the processed RPPG signal; A model building module is used to identify the constitution of the child myopic patient using a constitution identification tool to obtain a TCM constitution identification result; The model building module is also used to build an intelligent typing model by combining the basic data, the HRV characteristics and the TCM constitution identification results; The constitution classification module is used to classify the TCM constitution of the child myopia patient to be tested using the intelligent classification model to obtain the TCM constitution classification result.

8. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images