THE ARTIFICIAL INTELLIGENCE IN DENTAL FORENSICS: AGE ESTIMATION ACCURACY COMPARED WITH MANUAL METHODS

Background: Age estimation has an important role in forensic sciences, with dental radiographs serving as a reliable
source. Conventional methods which are widely used such as Demirjian’s technique, may vary depending on examiner
subjectivity. Artificial intelligence (AI) offers potential for speed and automation. This study compared the accuracy of
dental age estimation using a modified Demirjian’s method and ChatGPT-based AI estimation.
Methods: 50 digital orthopantomograms (25 males and 25 females, aged 10–20 years) were collected from digital OPG archives and analysed. Chronological age was calculated from the date of birth. Manual dental age estimation was estimated using the Indian modification of Demirjian’s method. ChatGPT was prompted to assign tooth development stages, calculate maturity scores, and estimate age based on the OPGs uploaded. Spearman’s correlation, Mann– Whitney U test, mean absolute error (MAE), and root mean square error (RMSE) were applied for statistical analysis.
Results: All age datasets showed deviation from normality (Shapiro–Wilk, p < 0.05). Manual dental age showed strong correlation with chronological age (ρ = 0.871, p < 0.001), at the same time AI-based estimation showed weaker
correlation (ρ = 0.394, p = 0.0047). A moderate correlation was observed between manual and AI estimates (ρ = 0.418, p = 0.0025). No significant gender-based differences were found. Accuracy was higher for the manual method (MAE = 1.22 years, RMSE = 1.55 years) compared to AI (MAE = 2.92 years, RMSE = 3.66 years).
Conclusion: In this study it was found out that manual age estimation using the modified Demirjian method remains
more accurate than ChatGPT based age estimation, confirming its role as the gold standard in forensic cases. While AI
demonstrated speed and consistency, its current limitations are the absence of datasets. In the future AI can be trained on wide radiographic datasets so that it can be an adjunct to. conventional methods , but manual estimation method remain unavoidable at present.