Heatstroke death identification using ATR-FTIR spectroscopy combined with a novel multi-organ machine learning approach

With global warming, the number of deaths due to heatstroke has drastically increased. Nevertheless, there are still difficulties with the forensic assessment of heatstroke deaths, including the absence of particular organ pathological abnormalities and obvious traces of artificial subjective assessment. Thus, determining the cause of death for heatstroke has become a challenging task in forensic practice. In this study, hematoxylin-eosin (HE) staining, attenuated total reflection-Fourier transform infrared spectroscopy (ATR-FTIR), and machine learning algorithms were utilized to screen the target organs of heatstroke and generate a multi-organ combination identification model of the cause of death. The hypothalamus (HY), hippocampus (HI), lung, and spleen are thought to be the target organs among the ten organs in relation to heatstroke death. Subsequently, the single-organ and multi-organ combined models were established, and it was found that the multi-organ combined approach yielded the most precise model, with a cross-validation accuracy of 1 and a test-set accuracy of 0.95. Additionally, the primary absorption peaks in the spectrum that differentiate heatstroke from other common causes of death are found in Amide I, Amide II, δ CH