AGE ESTIMATION BY ASSESSMENT OF DENTAL PULP VOLUME: A BAYESIAN METHOD APPLIED TO DENTAL EVIDENCE

Background: The pulp-dentin complex is a dental structure that exhibits age-dependent modifications, which result primarily in reduction of pulp chamber volume due to the continual deposition of secondary dentin. The aim of this study was to investigate the performance of a Bayesian method in the evaluation of dental age-related evidence collected by means of a geometrical approximation procedure of the pulp chamber volume. Measurement of dental volumes were based on Cone Beam Computed Tomography images. Methods: Two operators randomly analyzed 286 CBCT (114 males and 172 females) and considered the upper left central incisor. This research was designed to simplify the volume measurement through a geometric approximation of different parts of the tooth. The ratio between the pulp volume and the hard tissue volume (PHr) was assumed as a variable according to the following formula: PHr=V(P)/V(H). The Bayesian method was applied by means of a probabilistic graphical model, namely a Bayesian network. Performance of that method was investigated in terms of accuracy and bias of the decisional outcomes. Influence of an informed elicitation of the prior belief on the chronological age was also studied by means of a sensitivity analysis. Results: The physical measurements revealed that the CBCT analysis consistently underestimated the real dental pulp volumes by 53%-70%; however, the error occurred for estimation of both pulp and hard tissue volumes, and it tended to be eliminated by the ratio. The Bayesian method for adult age estimation appears to be a logical and robust approach for age estimation. The Bayesian network optimizes the evaluative procedure, which is relevant for daily forensic practice. Conclusions support the integration of expert initial belief in order to improve estimation results. Sensitivity analysis of the initial belief on the chronological age shows that an informed elicitation seems to considerably increase the quality of the estimates. Even when the initial belief on the most probable age is elicited with a 10-years distance from the real age of a given individual, results tend to be more accurate than those obtained with uniform prior. These findings encourage to use background information in the inferential process