Machine Learning for Skeletal Trauma Detection in Biological Anthropology

Archeo-anthropology and forensic anthropology increasingly rely on digital imaging, three-dimensional documentation, and quantitative analysis of skeletal remains. However, the interpretation of skeletal trauma remains largely expert-driven, time-consuming, and vulnerable to inter-observer variability, especially when remains are fragmentary, burned, weathered, or affected by taphonomic processes. The major issue remains the lack of standardization, both for the terminology and the workflow for the differential diagnosis. This paper argues that machine learning can support trauma analysis by detecting, localizing, classifying, and explaining traumatic lesions from photographs, radiographs, computed tomography scans, and three-dimensional bone models. After reviewing progress in clinical fracture detection, bone surface modification analysis, and forensic skeletal trauma resources, the paper identifies open research problems in computational anthropology.