A Novel Concept of the Management of Coronary Artery Disease Patients Based on Machine Learning Risk Stratification and Computational Biomechanics: Preliminary Results of SMARTool Project

Abstract

Coronary artery disease (CAD) is one of the most common causes of death in western societies. SMARTool project proposes a new concept for the risk stratification, diagnosis, prediction and treatment of CAD. Retrospective and prospective data (clinical, biohumoral, computed tomography coronary angiography (CTCA) imaging, omics, lipidomics, inflammatory and exposome) have been collected from ~250 patients. The proposed patient risk stratification, relying on machine learning analysis of non-imaging data, discriminates low and medium-to-high risk patients. The CAD diagnosis module is based on the 3D reconstruction and automatic blood flow dynamics of the coronary arteries, and the non-invasive estimation of smartFFR, an index correlated with invasively measured fractional flow reserve (FFR). CAD prediction is based on complex computational models of plaque growth considering the blood rheology, the lipoproteins transport and the major mechanisms of plaque growth, such as the inflammation and the foam cells formation. Finally, the treatment module is based on the simulation of virtual stent deployment. Preliminary analysis of 101 patients yielded an overall accuracy of 85.2% with the sensitivity of Class II reaching 98%. The reconstruction methodology is validated against intravascular ultrasound data and the correlation of the geometry derived metrics such as the degree of stenosis, minimal lumen area, minimal lumen diameter, plaque burden are 0.79, 0.85, 0.81 and 0.75, respectively. SmartFFR has been validated compared to invasively measured FFR with a correlation coefficient of 0.90. Plaque growth modelling demonstrates that the inclusion of variables such as the macrophages and foam cells concentrations can increase to 75% the prediction accuracy of regions prone to plaque formation.