AI Oral Segmentation — Enhancing Cone-Beam Computed Tomography Scan Analysis through Deep Learning

Grade:

12

Teacher:

Mari Echols

This project presents a real-time YOLOv8 pipeline that automates per-tooth instance segmentation of three-dimensional dental CBCT volumes. We trained two models on raw images and preprocessed images, finding that preprocessing raised mean accuracy by 6%. If adopted, this system could save dentists time, reduce mistakes, and one day make the dental field quicker, cheaper, and more accessible for all.

Project presentation

Research paper