딥러닝 기반의 상하악 사랑니 매복 유형 분류

Abstract

This paper proposes an deep learning method to predict impaction types of mandibular and maxillary third molar. We developed a semantic segmentation model that finds third teeth in panoramic radiography and a classification model that classifies the impaction type of third molars. Our deep learning model has a high accuracy of 90.7% for third molar semantic segmentation and has a high accuracy of over 80% for classification of impaction type of mandibular and maxillary third molar. It greatly help diagnose the extraction difficulty of mandibular and maxillary third molar.