The interplay of coronary serial stenoses and their deceptive impact on hemodynamic and surgical planning

Assessing serial stenoses in coronary artery disease (CAD) remains a difficult task, as hemodynamic interactions between lesions can mask individual severity. In this work, we present an automated and non-invasive decision tree-based framework that integrates computational fluid dynamics (CFD)-derived indices with CAD-RADS classification to guide revascularization strategies in cases of serial stenoses.

A dataset of 400 virtual coronary artery models was generated from 20 patients' anatomies, augmented with serial stenosis severities (CAD-RADS system) and boundary condition configurations. CFD simulations were performed to compute key hemodynamic parameters, including FFR, ΔFFR, ΔP, and PPG. Simulations also evaluated the impact of different angioplasty strategies, including proximal, distal, and combined revascularization. A decision tree was developed to classify cases based on lesion severity and hemodynamic interaction and was validated through a clinical survey answered by experienced cardiologists.

CFD simulations revealed notable hemodynamic interactions between lesions. In over 20% of cases, distal angioplasty revealed proximal FFR changes from a non-significant to a significant value (<0.8). Conversely, proximal angioplasty improved distal FFR beyond 0.8. The proposed decision tree outperformed FFR- and stenosis grade-based approaches, achieving an AUC of 0.98 and an accuracy of 0.97 when validated against CFD-based angioplasty simulations. Furthermore, it showed 81.2% agreement with clinicians' final decisions when informed by CFD and reduced iterative invasive procedures by 20%.

This study showed that the proposed CFD-based decision framework enhances accuracy in managing serial CAD by capturing lesion interactions, improving surgical planning precision, reducing patient risk, and supporting more consistent, efficient clinical decision-making.