Assessment of Myocardial Microcirculation and Its Relationship with Cardiac Remodeling and Myocardial Fibrosis in Athletes Based on IVIM-DWI Images

To assess myocardial microcirculation in athletes using intravoxel incoherent motion (IVIM) diffusion-weighted imaging and its relationship with cardiac remodeling (CR) and myocardial fibrosis (MF).Continuous 174 athletes and 54 healthy sedentary controls were prospectively recruited for cardiac magnetic resonance imaging (CMRI). Bi-ventricular function and IVIM parameters of 16 myocardial segments for left ventricle were measured and compared between different groups. Subsequently, machine learning models were utilized to predict CR and/or MF based on numerous clinical data and these IVIM parameters. The fast apparent diffusion coefficient (ADCfast) and perfusion fraction (f) of 16 myocardial segments and each blood supply area in athletes were significantly lower than in controls (p < 0.05). There were statistical differences for IVIM parameters among different exercise types (p<0.05). The myocardial slow apparent diffusion coefficient (ADCslow) values of athletes with CR and/or MF were lower than the remaining athletes (p < 0.05).The IVIM value were related with CMRI first-pass perfusion parameters and the presence of myocardial delayed enhancement (p < 0.05). Gradient boosting machine (GBM) was the best model to predict CR and/or MF based on these hypo-perfusion parameters, with an accuracy of 78.7%, area under the receiver operating characteristic curve of 0.865, sensitivity of 82.8%, and specificity of 77.9%. The ADCslow value of the 5th segment was the most important predictor for CR and/or MF.

本研究旨在评估运动员的心肌微循环，采用体素内不相关运动（IVIM）扩散加权成像技术，并探讨其与心脏重塑（CR）及心肌纤维化（MF）之间的关系。对174名运动员和54名健康非运动性受试者进行前瞻性招募，进行心脏磁共振成像（CMRI）。测量并比较了左心室16个心肌节段的二心室功能和IVIM参数。随后，利用机器学习模型根据众多临床数据和这些IVIM参数预测CR和/或MF。运动员的16个心肌节段和每个血供区域的快速表观扩散系数（ADCfast）和灌注分数（f）显著低于对照组（p < 0.05）。不同运动类型间的IVIM参数存在统计学差异（p < 0.05）。具有CR和/或MF的运动员的心肌缓慢表观扩散系数（ADCslow）值低于其他运动员（p < 0.05）。IVIM值与CMRI首次通过灌注参数及心肌延迟增强的存在相关（p < 0.05）。基于这些低灌注参数，梯度提升机（GBM）是预测CR和/或MF的最佳模型，其准确率为78.7%，受试者工作特征曲线下面积为0.865，灵敏度为82.8%，特异度为77.9%。第5节段的ADCslow值是预测CR和/或MF最重要的指标。