AI-assisted forensic odontology: Advances in automated human identification and digital dental evidence

Artificial intelligence (AI) is highly recommended used in odontology to make dental-image interpretation more automated and aid in human identification. This critical review discusses the recent evidence of the ante-mortem and post-mortem radiographic matching, tooth detection and numbering, dental-age estimation, biological-sex estimation, bite-mark analysis, cheiloscopy, palatal-rugae comparison, recognition of restoration and implant, virtual autopsy, automated charting and multimodal dental-record integration. The dental-age models can yield low estimation errors when compared with sex-estimation models, which still require consideration of population structure and anatomical predictors chosen. New 3-D rugae-matching and lip-print systems are technically feasible but are not widely and independently validated. The limitations inherent in the biology and evidentiary use of bite marks cannot be solved by AI. Common threats include homogeneous datasets, overfitting, patient level data leakage, class imbalance and insufficient multicentre testing across the various application areas. In addition, maps of attention and confidence scores are not in and of themselves interpretable, valid or legally reliable. The age and sex estimation, restoration classification and automated charting are technically validated but not operationally established; and tooth numbering, candidate retrieval, treatment-pattern recognition and post-mortem imaging supports are closest to implementation. The next steps involve future development of prospective open set evaluations, data separation at the identity level, geographic and cross-device validation of digital records and auditable chains of custody. In light of this, AI should be used as a clear decision-making tool, under the supervision of trained forensic dentists who responsible for the interpretation and evidential aspects of the case.