A new strategy for skeletal muscle wound age estimation using machine learning and ATR-FTIR spectroscopy: Eliminating early postmortem interference.

Accurate wound age estimation is of great significance in forensic practice. However, postmortem changes often obscure or even obliterate the biological information of skeletal muscle injuries, making it extremely challenging to accurately estimate their age. In this study, we combined ATR-FTIR spectroscopy with multiple machine learning algorithms to establish three regression models optimized by genetic algorithms (GA-Ridge, GA-Lasso, and GA-PLS) to estimate wound age in skeletal muscle, targeting the impact of early post-mortem changes. The results indicate that these models exhibit strong resistance to postmortem changes and demonstrate excellent predictive performance, with the best CV-R of 0.78, test R of 0.77, CV-MAE of 4.84 h, and test MAE of 5.01 h. In addition, eight spectral feature bands were found that are highly correlated with wound age. These features were located in absorption bands corresponding to amide II, CO stretching, C-O-C stretching, and PO₄ stretching vibrations, suggesting that changes in proteins, phospholipids, and nucleic acids may represent key biochemical events in the temporal evolution of muscle injury. In conclusion, this study proposes a new method for estimating skeletal muscle wound age based on ATR-FTIR and machine learning, taking into account the interference of early postmortem changes. This research offers a novel technical approach for forensic wound age estimation.