# Plasma-based Raman spectroscopy for early detection of acute myocardial infarction in murine models

## TL;DR

This study explores using Raman spectroscopy of plasma to detect heart attacks in mice, showing promising accuracy with machine learning.

## Contribution

The novel use of Raman spectroscopy combined with machine learning for early detection of acute myocardial infarction in murine models.

## Key findings

- Eight Raman peaks were identified that differentiate AMI from sham groups with 79.6% accuracy.
- Machine learning algorithms RF and LDA achieved highest accuracy, specificity, and sensitivity.
- Raman spectroscopy results aligned with metabolomic analysis, showing lipid downregulation in AMI.

## Abstract

Acute myocardial infarction (AMI) is a leading cause of cardiovascular mortality. Current diagnostic dilemma suffers from limited sensitivity, insufficiently timely and effective specificity. To address this dilemma, Raman spectroscopy, a rapid and non-invasive technique with significant potential for plasma metabolic profiling, although AMI involves rapid metabolic alterations, diagnostic approaches based on plasma metabolites remain underexplored. In murine AMI model induced by coronary artery ligation, we acquired Raman spectra from ultrafiltered plasma samples 8 h post-surgery. We identified eight distinct Raman peaks, assigned to amino acids, lipids, and nucleic acids, which collectively differentiated AMI group (n = 27) from the Sham group (n = 27). To ensure optimal accuracy, we employed five different algorithms including SVM, LR, RF, LDA and PLS-DA to analyze the Raman spectra, both RF and LDA achieved the highest accuracy of 79.6%, specificity of 85.2%, and sensitivity of 74.1%. Furthermore, metabolomic analysis revealed significant down-regulation of most lipid classes, aligning with the downregulation observed in the Raman peaks at 2885 cm⁻¹ and 2940 cm⁻¹. These results demonstrate a high concordance between plasma metabolic profiling via Raman spectroscopy and MS analysis. The integration of Raman spectroscopy with machine learning has remarkable potential for the early and accurate diagnosis of AMI, offering a promising approach for clinical translation.

The online version contains supplementary material available at 10.1038/s41598-025-30292-y.