Can AI-derived digital diagnostic tools for toothwear aid clinicians to detect toothwear independent of clinical experience?

• Grant number: 25/15843-0
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• Start date: January 21, 2026
• End date: July 20, 2026
• Field of knowledge: Health Sciences - Dentistry - Dental Clinics
• Principal Investigator: Taís Scaramucci Forlin
• Grantee: Claudia Allegrini Kairalla
• Supervisor: Saoirse O'Toole
• Host Institution: Faculdade de Odontologia (FO). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Institution abroad: King's College London, England
• Associated to the scholarship: 24/01944-7 - Validation of two methods of detection and clinical measurement of erosive tooth wear: A prospective Coort study, BP.DR

Abstract

This project aims to evaluate the performance of a pre-trained artificial intelligence (AI) tool in detecting early erosive tooth wear (ETW) using retrospective 3D intraoral scan data. Twelve pairs of scans will be selected from an ethically approved dataset with annotated clinical BEWE scores at surface level. Three examiners with different levels of clinical experience (an expert, a postgraduate student, and an undergraduate student) will independently annotate signs of tooth wear using the 3Shape annotation tool. The overlap between these annotations and those generated by the AI will be calculated using DICE similarity scores. After the initial evaluation, examiners will have access to the AI-generated segmentations and will re-analyze the same scans. Changes in their annotations will be compared to assess the influence of AI. Finally, a consensus ground truth will be established by a panel including an additional expert clinician on the intraoral scans of 60 participants and these will be released as an open source intraoral scan dataset. Statistical analyses will include inter-examiner agreement (Kappa), DICE scores, and subgroup evaluations by tooth type and location. This study seeks to determine the potential of AI as a support tool for early ETW diagnosis, improving accuracy and consistency among clinicians.