Innovative application of confocal Raman spectroscopy and Machine learning in cardiovascular diseases identification

Abstract

Myocardial hypertrophy and heart failure are leading causes of mortality in cardiovascular diseases, yet current diagnostic techniques lack the resolution to monitor molecular changes effectively. In this study, we employed confocal Raman spectroscopy combined with machine learning to evaluate myocardial tissue in a transverse aortic constriction (TAC) mouse model. Mice were divided into three groups: control (CON), myocardial hypertrophy (TAC 2 W), and heart failure (TAC 4 W). The model was validated using echocardiography, histopathology, and transmission electron microscopy. Raman spectroscopy revealed significant changes in chemical composition, including decreased peak intensities at 748, 1309, 1584, 1170, 1359, 1222, 1636, 1335, and 1393 cm⁻¹, increased intensity at 2847 cm⁻¹, and a rightward shift of the peak at 2927 cm⁻¹, which correlated with disease progression. Machine learning analysis demonstrated that the random forest model achieved 85 % accuracy in classifying normal, hypertrophic, and failing myocardial tissues. This study highlights the potential of confocal Raman spectroscopy combined with machine learning for the identification and real-time monitoring of cardiovascular diseases, offering a novel approach to understanding disease mechanisms and improving patient outcomes. Although the model achieved high classification accuracy, the small sample size may limit generalizability and requires further validation in larger cohorts.