CN-122025140-A - Method for predicting coronary artery stenosis degree based on tongue picture of coronary heart disease patient

CN122025140ACN 122025140 ACN122025140 ACN 122025140ACN-122025140-A

Abstract

The invention provides a method for predicting coronary artery stenosis degree based on tongue picture of coronary heart disease patient. The method comprises the steps of obtaining a data set, wherein the data set comprises electronic medical record information, blood examination indexes, heart ultrasonic indexes, electrocardiograms and tongue picture of a coronary heart disease patient, dividing the preprocessed data set into a training set and a testing set according to a preset proportion, constructing a deep learning network model, inputting the training set into the deep learning network model for training and optimizing, storing the final optimized deep learning network model, and inputting the testing set into the final optimized deep learning network model to obtain a prediction result of coronary artery stenosis degree of the coronary heart disease patient. The invention uses tongue picture, electrocardiogram, heart ultrasonic index and blood examination index of coronary heart disease patient to judge coronary artery stenosis degree, and converts tongue picture feature extracted from the tongue picture into quantifiable coronary artery stenosis noninvasive screening target spot, thereby promoting the intellectualization of combined diagnosis and treatment of Chinese and Western medicine.

Inventors

YU KAIFENG
ZHENG XIAOFENG
ZHANG ZHE
YANG DAI
LI YUE
LIU NING
WANG TIANQI
WU XIZE
REN JIAQI
DUAN YINGJIE

Assignees

辽宁中医药大学附属医院

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (6)

1.A method for predicting coronary artery stenosis based on tongue picture of coronary heart disease patient, comprising: step S1, acquiring a data set, wherein the data set comprises electronic medical record information, blood examination indexes, heart ultrasonic indexes, electrocardiogram and tongue picture of a coronary heart disease patient, and the data set is subjected to data marking; Step S2, preprocessing the data set; step S3, dividing the preprocessed data set into a training set and a testing set according to a preset proportion; s4, building a deep learning network model; S5, inputting the training set into the deep learning network model for training and optimizing, and storing the finally optimized deep learning network model; and S6, inputting the test set into the final optimized deep learning network model to directly obtain a prediction result of the coronary artery stenosis degree of the coronary heart disease patient.
2. The method for predicting coronary artery stenosis degree based on tongue picture of coronary heart disease patient as recited in claim 1, wherein said electronic medical record information comprises one or a combination of tongue diagnostic information and prior medical history in the electronic medical record; The blood test index comprises one or a combination of white blood cell count, neutrophil percentage, lymphocyte percentage, neutrophil count, lymphocyte count, monocyte count, red blood cell count, hemoglobin, hematocrit, platelet count, alanine aminotransferase, aspartic acid aminotransferase, triglyceride, total cholesterol, high density lipoprotein cholesterol, low density lipoprotein cholesterol, potassium, sodium, chlorine, urea, creatinine, fasting glucose, glomerular filtration rate, troponin, D dimer assay; The cardiac ultrasound index comprises one or a combination of a left Fang Najing, a right chamber inner diameter, a chamber interval thickness, a left chamber back wall thickness, a left chamber end diastole inner diameter, a left chamber end systole inner diameter, a left chamber Shu Mo volume, a left chamber end systole volume, a stroke volume and a ejection fraction.
3. The method for predicting coronary artery stenosis degree based on tongue picture of coronary heart disease patient as recited in claim 2, wherein said step S2 specifically comprises: S2.1, mapping two unordered classified variables, namely tongue diagnosis information and past medical history, in the electronic medical record into two classified variables by adopting a dummy variable conversion method, wherein the two classified variables are 0 or 1; s2.2, scaling the two quantitative data of the blood examination index and the heart ultrasonic index to an interval [0,1] by adopting normalized mathematical transformation so as to eliminate the difference of different characteristic dimensions; s2.3, preprocessing the tongue picture to extract tongue picture characteristics corresponding to the tongue picture; S2.4, preprocessing the electrocardiogram to extract 12 lead characteristics corresponding to the electrocardiogram; and S2.5, carrying out fusion on tongue image diagnosis information converted by the dummy variable, past medical history, blood examination indexes and heart ultrasonic indexes after normalized mathematical transformation, and the tongue image characteristics and the 12-lead characteristics to form a multi-mode data set.
4. The method for predicting coronary artery stenosis degree based on tongue picture of coronary heart disease patient as recited in claim 3, wherein said step S2.3 specifically comprises: s2.31, carrying out saturation normalization processing on the tongue picture by an image preprocessing algorithm based on color space analysis; S2.32, carrying out low-frequency alignment on the tongue picture subjected to saturation normalization processing based on an image low-frequency characteristic algorithm of Fourier spectrum analysis; s2.33, carrying out tongue image segmentation on the tongue image after low-frequency alignment; And S2.34, extracting tongue picture characteristics of the tongue picture image after tongue picture segmentation.
5. The method for predicting coronary artery stenosis based on a tongue picture of a patient with coronary heart disease of claim 3, wherein the 12 lead features corresponding to the electrocardiogram comprise 6 limb leads and 6 chest leads.
6. The method for predicting coronary artery stenosis degree based on tongue picture of coronary heart disease patient as recited in any one of claims 1 to 5, wherein said deep learning network model is composed of a multi-layer perceptron composed of an input layer, a hidden layer and an output layer, said hidden layer being located between said input layer and said output layer, said multi-layer perceptron being abbreviated as MLP.

Description

Method for predicting coronary artery stenosis degree based on tongue picture of coronary heart disease patient Technical Field The invention relates to the technical field of intelligent recognition of coronary artery stenosis degree, in particular to a method for predicting the coronary artery stenosis degree based on tongue picture of a patient with coronary heart disease. Background Coronary artery stenosis or blockage is a major cause of death in cardiovascular disease. The noninvasive examination for diagnosing coronary artery stenosis mainly comprises coronary artery CTA, load electrocardiogram, load heart ultrasound, nuclide myocardial imaging and other technologies. They indirectly assess the likelihood of coronary stenosis and the extent of myocardial ischemia by scanning or inducing cardiac load in vitro, and have the common advantage of safety, convenience, and can be used as a primary screening tool. However, the key drawbacks are that the results are mostly indirect deductions, the stenosis degree cannot be precisely quantified, and false positive or false negative results are easily generated due to individual differences of patients (such as heart rate, calcification, etc.). Invasive examination for diagnosing coronary artery stenosis uses coronary angiography as a gold standard, and derives blood flow reserve fraction, intravascular ultrasound, optical coherence tomography, and other techniques. They can display the anatomy of the vessel, plaque properties, and can directly measure the effect of stenosis on blood flow function by inserting a catheter or probe directly into the vessel. However, these methods all belong to interventional procedures, and the common disadvantages are the risk of puncture injuries, contrast nephropathy and other related procedures, complex operation and high cost. Based on the above drawbacks, up to now, chinese scholars have proposed 6 improved approaches. The method disclosed in patent 1 (patent number: CN 109222980A) is based on a deep learning coronary angiography image vessel diameter measurement technology, and utilizes DSA image segmentation and vessel center line extraction algorithm to realize automatic quantitative analysis of coronary stenosis rate and avoid subjective difference of the traditional PVA method. Patent 2 (patent number: CN 120180194A) applies an electrocardio/heart sound bimodal signal recognition technology, and realizes noninvasive evaluation of coronary stenosis degree through multi-domain feature fusion of amplitude/interval sequences. Patent 3 (patent number: CN 117860210A) adopts an improved coronary artery stenosis recognition scheme of a combined recursion chart, adopts a fuzzy threshold value to optimize the oscillation of the recursion chart and the extraction of local binarization features, and improves the capturing capability of electric-mechanical coupling information. Patent 4 (patent number: CN 118614882A) discloses a coronary artery identification method based on a relative position matrix, which is used for mining dual-signal implicit association information through matrix joint modeling of an electrocardio/heart sound interval sequence. Patent 5 (patent number: CN 118824533A) uses N-glycoset non-invasive classification model technique, and uses blood sugar chain fingerprint and LASSO regression feature screening to avoid the risk of wound of traditional ICA. Patent 6 (patent number: CN 120182244A) adopts a magnetic resonance CNR value coronary artery analysis method, combines active contour model ROI extraction and dynamic adaptive CNR calculation, and improves the stenosis diagnosis efficiency. Current coronary stenosis identification techniques, while making progress objectively, are still limited by device dependency and operational complexity. Particularly for early screening and primary medical scenarios, a need exists for a lower cost, non-invasive and easily generalized aid. It is worth noting that innovative methods for predicting diseases through human surface features are attracting attention, and the medical community is currently working on developing intelligent coronary disease prediction systems based on human surface images (such as tongue and the like) to break through the limitations of the existing diagnostic methods. Disclosure of Invention The present invention aims to solve at least one of the technical problems existing in the prior art or related art. Therefore, the invention aims to provide a method for predicting the coronary artery stenosis degree based on tongue picture of a patient with coronary heart disease. In order to achieve the above purpose, the technical scheme of the invention provides a method for predicting the coronary artery stenosis degree based on tongue picture of coronary heart disease patients. The coronary heart disease patient tongue image picture-based coronary artery stenosis degree prediction method comprises the steps of S1, obtaining a data set, wherein the data se