Role of Computational Biology Models in Dermatoglyphics And Forensic Anatomy: A Review

Abstract

Automated fingerprint systems are very accurate with clear prints, but they do not work well with damaged or unclear evidence. This review assessed how well computational biology models perform in fingerprint analysis compared to manual methods. We searched five databases from 2018 to 2024, following PRISMA guidelines. We included 33 studies using machine learning for fingerprints, with data from 169 forensic examiners and 744 fingerprint pairs from real cases. We measured accuracy, false-negative rates, and group differences using meta-analysis in R. Computational models achieved 99.6% accuracy on clear prints, while manual checks achieved 92%. Both methods had a 7.5% false negative rate. 85% of examiners made at least one mistake, and algorithms worked 15–25% worse for less-represented groups. Deep learning achieved better fingerprint recognition but struggled with poor-quality prints. Computational biology models are best with clear evidence, but real forensic work still needs both algorithms and human experts.