Fig. 1

This schematic illustrates an innovative workflow for diagnosing AMI using Raman spectroscopy integrated with machine learning. Initially, AMI was modeled in mice via coronary artery ligation, followed by collection of plasma samples 8 h post-surgery for Raman spectral and MS analysis. The study identified eight characteristic Raman peaks—corresponding to amino acids, lipids, and nucleic acids—that effectively distinguished the AMI group from the sham group. Five machine learning algorithms were applied to process the spectral data, with the optimal model achieving an accuracy of 79.6% by RF and LDA. Metabolomic validation confirmed that downregulated lipid metabolism correlated consistently with changes in the Raman spectral features.