New Horizons of Science, Technology and Culture Vol. 12

Forensic pathology rests on the microscopic examination of postmortem tissue to establish how, and roughly when, a person died, yet histopathological interpretation is slow, dependent on scarce expertise, and prone to disagreement between examiners. Over the past decade, artificial intelligence, and deep learning in particular, has reshaped diagnostic histopathology in clinical medicine, and these same techniques are now being adapted to the very different problem of postmortem tissue, with its autolysis, putrefaction, and lack of any living clinical reference point. This critical review brings together the literature on artificial intelligence in forensic pathology, focusing specifically on histopathological interpretation. It traces the field from early convolutional neural networks through to the foundation models now appearing in computational pathology, and it examines specific forensic applications: postmortem interval estimation through pathomics, wound age and vitality assessment, cardiac and other organ-based determination of cause of death, diatom-based diagnosis of drowning, and age estimation from skeletal and dental maturation. It also weighs the methodological and ethical obstacles that stand between this research and routine practice — small and uneven datasets, limited external validation, opaque model reasoning, demographic bias, and open questions about evidentiary admissibility and professional accountability. The picture that emerges is one in which artificial intelligence functions, and seems likely to keep functioning for some time, as a decision-support tool that sharpens rather than replaces the forensic pathologist's own judgement. Real progress will depend less on further algorithmic refinement in isolation than on multicentre data sharing, prospective and externally validated studies, transparent model design, and regulatory frameworks built specifically for the medicolegal setting in which forensic pathology operates.