Abstract: Objective To infer postmortem interval （PMI） based on spectral changes of the dorsal skin of rats within 15 days postmortem using Fourier transform infrared （FTIR） spectroscopy. Methods The rats were sacrificed by cervical dislocation after anesthesia, and then placed at 25 ℃ and relative humidity of 50%. The FTIR spectral data collected from the dorsal skin at PMI points were modeled with machine learning technique. Results There was no significant difference of absorption peak location among all the PMI groups but their peak intensities changed as a function of PMIs. The model for PMI estimation was constructed using partial least squares （PLS） regression, reaching a R2 of 0.92 and a root mean square error （RMSE） of 1.30 d. As shown in variable importance for projection （VIP）, four spectral bands including 1 760-1 700 cm-1, 1 660-1 640 cm-1, 1 580-1 540 cm-1 and 1 460-1 420 cm-1 were determined as important contributions to model prediction. Conclusion Application of the FTIR technique to detect postmortem spectral changes of the rat skin provides a novel proposal for PMI estimation.

Key words: forensic pathology, spectroscopy, Fourier transform infrared, machine learning, postmortem interval, skin, rats