Trustworthy Data-driven Chronological Age Estimation from Panoramic Dental Images

TL;DR

A system for dental age estimation from panoramic images that combines an opaque and a transparent method within a natural language generation (NLG) module that produces clinician-friendly textual explanations about the age estimations is proposed.

Abstract

Integrating deep learning into healthcare enables personalized care but raises trust issues due to model opacity. To improve transparency, we propose a system for dental age estimation from panoramic images that combines an opaque and a transparent method within a natural language generation (NLG) module. This module produces clinician-friendly textual explanations about the age estimations, designed with dental experts through a rule-based approach. Following the best practices in the field, the quality of the generated explanations was manually validated by dental experts using a questionnaire. The results showed a strong performance, since the experts rated 4.77+/-0.12 (out of 5) on average across the five dimensions considered. We also performed a trustworthy self-assessment procedure following the ALTAI checklist, in which it scored 4.40+/-0.27 (out of 5) across seven dimensions of the AI Trustworthiness Assessment List.

Figures (11)

• Figure 1: Two-stage CNN-based method for dental age estimation blanco2023XAS. The first convolutional network detects teeth in the input OPG. This information is used along with detected image features as input for the second network, which estimates the patient’s sex and age
• Figure 2: Measurements used to compute $CSM(4i)=A_i/L_i$, $i \in \{1,...,7\}$ by cameriere2006age. $A_i$ correspond to the apex openness, while $L_i$ correspond to each tooth height. The image shows the mandibular right quadrant of a patient's dentition, with tooth 45 (second premolar) highlighted
• Figure 3: The Text Description Generation system in the core of AgeX. The system uses patient information to generate narratives. This information is used to determine what to say and how to structure the content. Then, the system designs how to convey the information and finally generates the correct text, resulting in personalized dental reports
• Figure 4: Diagram illustrating the iterative and incremental development process of AgeX, with continuous feedback and validation at each stage
• Figure 5: Example of a heat-map created from the OPG of a patient. In the image, the patient's permanent mandibular teeth are highlighted. The most predictive teeth are marked in red, while the least predictive ones are in green. The uncolored lower teeth are primary (baby) teeth