Multi-class Semantic Segmentation of Tooth Pathologies and Anatomical Structures on Bitewing and Periapical Radiographs [Machine Vision Applications, Hamamatsu, Japan, 2023]
J.A. Sarmiento, L. Chen, P.C. Naval, Jr.,Â
Solution
We developed an AI model based on the U-Net Deep Learning architecture that has the ability to identify particular dental pathologies, as well as various anatomical features that are important for accurate diagnosis.This model was trained on 344 images and can recognize two types of dental problems and eight features of teeth. It has shown good performance in identifying these issues—scoring 0.794 overall for accuracy and 0.787 in sensitivity. For specific problems like tooth decay and root infections, the scores were lower but still promising.
Problem Statement
Traditional dental exams may not effectively identify all dental issues, leading to the need for radiography (X-rays) for better detection. X-ray interpretation is often time-consuming, subjective, and requires specialized knowledge, which can result in errors that may delay treatment and worsen dental health problems. Given these challenges, there is a pressing need for an automated solution that enhances the accuracy and efficiency of diagnosing dental issues.