NAFLD Progression in Metabolic Syndrome: A Raman Spectroscopy and Machine Learning Approach in an Animal Model
Abstract
Nonalcoholic fatty liver disease (NAFLD) is emerging as the leading cause of chronic liver disease in many regions, particularly in association with the rising prevalence of metabolic syndrome (MetS), affecting more than 30% worldwide. It is a leading cause of liver-related diseases. Metabolic disorder is associated with the development and progression of NAFLD, demonstrating its relationship with MetS. However, the early detection and quantification of hepatic steatosis remains a challenge for current standard methods, such as liver biopsy and imaging. This work aimed to explore the potential of Raman spectroscopy as an alternative tool for NAFLD and MetS assessment. Spectral changes related to MetS and NAFLD were detected in liver samples at different points of time, using an animal model of MetS (8,18,28, and 52 weeks). The results show spectral differences associated with vitamin A, carotenoids, and lipids. A principal component analysis and a support vector machines (PCA-SVM) model for binary classification achieved an accuracy of 89.5%, while a multiclass classifier reached 91.6% accuracy on 52-week Raman spectra. These results underscore the potential of this approach for detecting liver disease and tracking temporal changes in NALFD during the development of MetS.