Age Estimation in Forensic Dentistry Using Artificial Intelligence: An Umbrella Review
TL;DR
Artificial intelligence-driven dental age estimation models—specifically deep CNN frameworks—exhibit high precision and clinical reproducibility for forensic applications, however, current conclusions are limited by the low methodological quality of the underlying systematic reviews.
Abstract
Artificial intelligence (AI) frameworks, particularly deep learning architectures, have emerged as powerful tools to optimize the objectivity and efficiency of dental age estimation. This umbrella review systematically synthesized and critically appraised the current evidence from existing systematic reviews regarding the diagnostic accuracy and clinical performance of AI-driven dental age estimation in forensic odontology. A comprehensive search protocol was executed across PubMed, PubMed Central, Google Scholar, and IEEE Xplore to identify systematic reviews published up to December 2024. The population of interest focused on systematic reviews evaluating age estimation accuracy in living or deceased individuals. Methodological quality and risk of bias of the included systematic reviews were independently evaluated using the AMSTAR 2 tool, and literature redundancy was assessed via the corrected covered area (CCA) index. Out of 474 initially screened records, five systematic reviews tracking diverse machine learning networks were included. Convolutional neural networks (CNNs) and deep learning architectures consistently demonstrated superior diagnostic precision, yielding mean absolute errors frequently below 1.0 year and overall accuracies ranging from 94.7% to 99.98% compared to traditional manual approaches (e.g., Demirjian and Nolla methods). However, the AMSTAR 2 assessment revealed critical methodological vulnerabilities, grading all five included systematic reviews as having low quality due to systemic deficiencies in protocol registration, duplicate data extraction protocols, and risk of bias tracking. The calculated CCA demonstrated minimal primary study overlap across the synthesized literature. AI-driven dental age estimation models—specifically deep CNN frameworks—exhibit high precision and clinical reproducibility for forensic applications. However, current conclusions are limited by the low methodological quality of the underlying systematic reviews. Future research must prioritize standardized external validation across diverse ethnic populations and implement explainable AI models to ensure diagnostic transparency before integration into routine forensic casework.