郑州大学学报（医学版）2026，Vol.61Issue(3)：86-90,5.DOI:10.13705/j.issn.1671-6825.2025.03.078
超声心动图影像组学数据对肥厚性心肌病心脏重构的预测价值
Prediction of cardiac remodeling in patients with hypertrophic cardiomy-opathy using ultrasound radiomics
摘要
Abstract
Aim:To explore the predictive value of radiomics from ultrasonic cardiography data for cardiac remodeling in patients with hypertrophic cardiomyopathy(HCM).Methods:A total of 265 patients with HCM who were treated in the Affiliated Hospital of Chengdu University and West China Fourth Hospital From January 1,2021 to December 31,2024,were selected,and 92 patients had cardiac remodeling.Logistic regression was used to screen factors influencing cardiac re-modeling and construct a clinical model.Apical four-chamber cardiac images were retained to extract shape,first-order,tex-ture,and wavelet features,and LASSO regression and 10-fold cross-validation were used to adjust parameters and select the optimal features;based on these optimal features,a radiomics model was constructed using linear regression.Using the opti-mal features as predictors,models were separately built with the k-nearest neighbor algorithm,multilayer neural network al-gorithm,and support vector machine(SVM)algorithm,and the model with the best performance was selected as the final deep learning model through ROC curve analysis.Based on the 3 models,a fusion model employing a weighted integration strategy was constructed.The performance of each model was evaluated using ROC curves combined with the Bootstrap 1 000 method.Results:The clinical model included 6 predictive factors:LAD,LVPWd,E/A,LVRI,MVCF,and LVMI.The radiomics model included 5 optimal features:wavelet_HLH_gray-level size zone matrix_zone percentage,wavelet_HHH_gray-level size zone matrix_zone size non-uniformity,wavelet_LHL_gray-level dependence matrix_dependence variance,gra-dient_gray-level size zone matrix_zone variance,and wavelet_LLH_gray-level dependence matrix_dependence non-uniformi-ty.The SVM model was selected as the deep learning model.The weights of the 3 models in the fusion model were 0.281,0.320,and 0.399,respectively.The fusion model showed the best performance in predicting cardiac remodeling,with an AUC(95％CI)of 0.967(0.847-0.983).Conclusion:The fusion model based on ultrasound radiomics can accurately predict cardiac remodeling in patients with HCM.关键词
肥厚性心肌病/心脏重构/超声影像组学/超声心动图Key words
hypertrophic cardiomyopathy/cardiac remodeling/ultrasound radiomics/ultrasonic cardiography分类
医药卫生引用本文复制引用
李建利,杨淑娟,曾红莲,杨波,冯坤..超声心动图影像组学数据对肥厚性心肌病心脏重构的预测价值[J].郑州大学学报（医学版）,2026,61(3):86-90,5.基金项目
四川省科技计划项目(2023YFS0251) （2023YFS0251）
