Using intravascular ultrasound image-based fluid-structure interaction models and machine learning methods to predict human coronary plaque vulnerability change

Plaque vulnerability prediction is of great importance in cardiovascular research. In vivo follow-up intravascular ultrasound (IVUS) coronary plaque data were acquired from nine patients to construct fluid-structure interaction models to obtain plaque biomechanical conditions. Morphological plaque vulnerability index (MPVI) was defined to measure plaque vulnerability. The generalized linear mixed regression model (GLMM), support vector machine (SVM) and random forest (RF) were introduced to predict MPVI change (ΔMPVI = MPVIfollow-up‒MPVIbaseline) using ten risk factors at baseline. The combination of mean wall thickness, lumen area, plaque area, critical plaque wall stress, and MPVI was the best predictor using RF with the highest prediction accuracy 91.47%, compared to 90.78% from SVM, and 85.56% from GLMM. Machine learning method (RF) improved the prediction accuracy by 5.91% over that from GLMM. MPVI was the best single risk factor using both GLMM (82.09%) and RF (78.53%) while plaque area was the best using SVM (81.29%).