Spectral Region Optimization and Machine Learning-Based Nonlinear Spectral Analysis for Raman Detection of Cardiac...

Area of Science:

• Biomedical Optics
• Cardiovascular Pathology
• Spectroscopic Analysis

Background:

• Cardiac fibrosis post-myocardial infarction is crucial for scar formation and ventricular arrhythmias.
• Current diagnostics lack precision due to limited spatial and molecular information.
• Collagen is a key molecular biomarker in fibrotic remodeling.

Purpose of the Study:

• To classify myocardial tissue types using line scan Raman microspectroscopy.
• To utilize collagen-associated Raman bands for precise fibrosis characterization.
• To develop a robust framework for potential real-time intraoperative diagnostics.

Main Methods:

• Line scan Raman microspectroscopy applied to sheep myocardial tissue.
• Evaluation of three spectral regions, with a focus on collagen peaks (1448 cm⁻¹ and 1652 cm⁻¹).
• Application of principal component analysis (PCA) and support vector machines (SVMs) for classification.

Main Results:

• Collagen-associated Raman bands in region C showed superior performance for tissue classification.
• PCA-SVM models, both linear and nonlinear, achieved high explained variance and class separation.
• Stage-dependent tissue characterization was enabled by collagen peak ratios, confirming nonlinear fibrotic remodeling.

Conclusions:

• Raman bands associated with collagen show significant diagnostic potential for myocardial fibrosis.
• The PCA-SVM framework is robust, even with limited sample sizes.
• This approach could form the basis for real-time intraoperative diagnostics of cardiac fibrosis.