The splenic evanescent trauma identification and the injured time estimation in decomposing cadavers based on ATR-FTIR and chemometrics

Trauma is one of the most prevalent disorders in the biomedical field. Abdominal trauma is widely encountered in accidents and some criminal situations, and it has a significant influence on the investigation or conviction of violent crimes. With about 33% of cases involving blunt abdominal trauma, the spleen is the organ that suffers the most and is associated with a significant risk of morbidity and mortality, particularly in patients who have accumulated multiple or compound injuries. “Evanescent trauma” describes an injury that becomes invisible as time passes, decomposition occurs, and other factors involved. In this study, we have found methods to identify splenic evanescent trauma and estimate the injured time in postmortem cadavers by FTIR combination with chemometrics. The PLSDA model to identify splenic evanescent trauma in decomposing cadavers' specificity of externally validation was 0.917, sensitivity was 0.958, and accuracy was 0.950. The PLSDA model to identify DSR and SR in decomposing cadavers' specificity was 0.833, sensitivity was 0.889, and accuracy was 0.875. To eliminate the impact cause by the cadaver's autolysis and putrefaction, and the redundancy of full-spectral data modeling, we performed variables selection with the methods of CARS, GA, SPA. The performance of CARS-PLS was superior to GA-PLS. The SPA-CARS-PLS regression model for the injured time estimation in decomposing cadavers at PMI 2d and 10d had the better performance and better interpretability of spectral. The SPA-CARS-PLS regression model to estimate the injured time at PMI 2d was constructed with 6 latent variables, the R2 CV was 0.911 and the RMSECV was 1.477 d. For external validation, the R2Pred value was 0.884. The root mean square error of prediction(RMSEP) was 2.404 d. The SPA-CARS-PLS regression model to estimate the injured time at PMI 10d was built with 6 latent variables, the R2CV was 0.858 and the RMSECV was 1.868 d. For external validation, the prediction model's R2Pred value was 0.930, the RMSEP was 1.487 d. FTIR combine with chemometrics is useful in identifying the splenic evanescent trauma and estimating the injured time in postmortem decomposed cadavers.