Construction and Analysis of a Machine Learning Model for Risk Prediction of Essential Hypertension with Left Ventricular Hypertrophy Based on Pulse Chart Parameters
10.19879/j.cnki.1005-5304.202411097
- VernacularTitle:基于脉图参数的原发性高血压伴左心室肥厚风险预测机器学习模型构建及分析
- Author:
Siman WANG
1
;
Mengchu ZHANG
1
;
Wen LI
1
;
Ai XU
1
;
Minghui YAO
1
;
Jin XU
1
;
Rui GUO
1
;
Yiqin WANG
1
;
Haixia YAN
1
Author Information
1. 上海中医药大学上海市健康辨识与评估重点实验室,上海 201203
- Publication Type:Journal Article
- Keywords:
essential hypertension;
pulse wave parameters;
left ventricular hypertrophy;
machine learning;
prediction model
- From:
Chinese Journal of Information on Traditional Chinese Medicine
2025;32(7):134-141
- CountryChina
- Language:Chinese
-
Abstract:
Objective To construct a model for predicting the risk of essential hypertension accompanied by left ventricular hypertrophy using machine learning algorithms based on pulse diagram parameters;To explore its clinical application value.Methods A total of 295 patients with essential hypertension who were hospitalized in Shuguang Hospital Affiliated to Shanghai University of Traditional Chinese Medicine,Shanghai Hospital of Traditional Chinese Medicine and Shanghai Hospital of Integrated Traditional Chinese and Western Medicine were selected from July 2020 to May 2021 and July 2023 to July 2024.According to the echocardiographic results,the selected research subjects were divided into the essential hypertension with left ventricular hypertrophy group(referred to as the"LVH group")and the essential hypertension without left ventricular hypertrophy group(referred to as the"non-LVH group").The general data and clinical biochemical indicators were collected,and the pulse diagram parameters of the patients were detected using the SMART-I type TCM digital pulse analyzer.A clinical prediction model was constructed based on decision tree,support vector machine and extreme gradient boosting model algorithms.The predictive performance of the model was evaluated in terms of discrimination,calibration and clinical prediction ability by using the receiver operating characteristic(ROC)curve,calibration curve and decision curve analysis respectively.The influence of each predictive factor on the risk of LVH in essential hypertension was explained based on the SHAP algorithm.Results Compared with the non-LVH group,the BMI,the proportion of males,drinkers and smokers was lower in the LVH group,with statistical significance(P<0.05);the thickened ventricular wall,left ventricular internal dimension enlargement,left common carotid artery intima-media thickness and high density lipoprotein cholesterol were higher in the LVH group than in the non-LVH group(P<0.05);the left common carotid peak systolic velocity,left common carotid resistance index,serum uric acid and serum creatinine were lower in the LVH group than in the non-LVH group(P<0.05).The pulse diagram parameters T4,T,W1,W2,H3/H1 and H4/H1 were higher in the LVH group than in the non-LVH group(P<0.05).The areas of the ROC curves of the models constructed by the three types of machine learning algorithms were 0.887,0.962 and 0.873 respectively,indicating that the model had good discrimination and certain diagnostic efficacy.The calibration curve suggested that the prediction accuracy of the model was average;the clinical decision curve showed that XGBoost model has a higher net benefit.Conclusion The interpretable model constructed based on pulse diagram parameters and machine learning algorithms can be used as a reliable tool for predicting the risk of essential hypertension with LVH.
