In vivo raman spectroscopy of muscle is highly sensitive for detection of healthy muscle and highly specific for detection of disease

Raman spectroscopy of muscle provides a molecular fingerprint to identify the disease. Previous work has demonstrated effectiveness in differentiating between two groups of equal sizes (e.g., healthy vs disease) but imbalanced multiclass scenarios are more common in medicine. We performed in vivo Raman spectroscopy in a total of 151 mice across four different histopathologies (healthy, acute myopathy, chronic myopathy, neurogenic), with variable numbers in each (class “imbalance”). Using hierarchical modeling and synthetic data generation, we demonstrate high sensitivity (94%) for detection of healthy muscle and high specificity (≥97%) for disease. Further, we demonstrate the potential for unique biomarker development by demonstrating variations in the protein structure across different pathologies. The findings demonstrate the potential of Raman spectroscopy to provide accurate disease identification and unique molecular insights.