Advancing Coronary Risk Assessment Through Combined Radiomic, Mechanical, and Hemodynamic Analysis.
Authors
Affiliations (7)
Affiliations (7)
- Department of Electronics, Information and Bioengineering, Politecnico di Milano, Via Ponzio 34/5, 20133, Milan, Italy. [email protected].
- Laboratory of Biological Structure Mechanics (LaBS), Department of Chemistry, Materials and Chemical Engineering "Giulio Natta", Politecnico di Milano, Milan, Italy.
- Department of Electronics, Information and Bioengineering, Politecnico di Milano, Via Ponzio 34/5, 20133, Milan, Italy.
- Division of Cardiology, Department of Medical Sciences, "Città della Salute e della Scienza di Torino" Hospital, University of Turin, Turin, Italy.
- Radiology Unit, Department of Surgical Sciences, "Città della Salute e della Scienza di Torino" Hospital, University of Turin, Turin, Italy.
- Department of Industrial and Information Engineering, University of Pavia, Pavia, Italy.
- Cardiotech Lab, Centro Cardiologico Monzino IRCCS, Milan, Italy.
Abstract
Detecting vulnerable coronary plaques through coronary computed tomography angiography (CCTA) is a crucial, yet challenging task. To date, most of the proposed vulnerability markers have been studied in isolation. This study introduces the first integrated analysis combining radiomic, mechanical, and hemodynamic factors to explore their synergistic contribution to plaque vulnerability. The study analyzed 161 plaques in 46 coronary arteries from 39 patients, with 7 arteries (28 plaques) from 7 individuals, labeled as vulnerable from intravascular imaging. First, CCTA radiomic features were extracted. Second, mechanical markers were computed through finite element simulations conducted with varying material characteristics, accounting for the arterial wall mechanical properties' uncertainties. Third, hemodynamic markers were derived from transient computational fluid dynamics simulations. Finally, a machine learning pipeline was developed to classify coronary arteries and patients based on radiomic, mechanical, and hemodynamic features, both individually and in combination. Radiomics achieved the highest sensitivity (1.00), with all vulnerable patients identified, but lower specificity (0.69). Differently, mechanics and hemodynamics achieved higher specificities (0.94 and 0.97, respectively) but lower sensitivities (both 0.86). By integrating at least two out of the three models, the predictive performance improved, up to sensitivity = 1.00 and specificity = 0.97, with only one misclassified case. Although based on only 39 patients, the results highlight the power of a multi-level integrative approach for coronary plaque assessment. The study revealed that (i) hemodynamics outperformed mechanics and radiomics; (ii) while radiomics maximized sensitivity, mechanics and hemodynamics prioritized specificity, and (iii) integrating at least two variable types added value.