基于深度学习的龋齿辅助诊断系统  

A dental caries assistant diagnosis system based on deep learning

作　　者：项晋凡 缪佳欣 杨祎伟 余欣[1] 刘晓宇[1] XIANG Jinfan;MIAO Jiaxin;YANG Yiwei;YU Xin;LIU Xiaoyu(School of Mechanical Engineering,Sichuan University,Chengdu 610065,China)

机构地区：[1]四川大学机械工程学院,成都610065

出　　处：《智能计算机与应用》2025年第3期64-71,共8页Intelligent Computer and Applications

基　　金：2023年四川大学国家级大学生创新创业训练计划(202310610108)。

摘　　要：龋病作为一种常见病,长期以来一直影响着大众的身体健康。随着患龋人群规模不断扩大,牙医资源愈发紧缺。为解决目前缺少人工智能手段辅助医生进行龋齿诊断,且人工诊断效率难以保证等问题,设计了一种基于图像处理和深度学习的口腔小牙片龋齿辅助诊断方案。该方案基于U-Net语义分割模型和YOLOv5目标检测模型,对口腔小牙片中的龋齿病变进行识别,训练出的模型可有效识别龋坏区域,最优的mAP@0.5值达到95.3%,准确度达到95.9%。同时结合图形化界面开发工具PyQt5完成适医化的龋齿检测与辅助诊断系统的开发,给医生提供易操作的辅助软件,从而有效提高医疗效率。As a common disease,caries has affected the health of the public for a long time.As the size of the population suffering from caries continues to expand,dental resources are becoming more and more scarce.In order to solve the problems of lack of artificial intelligence to assist doctors in dental caries diagnosis and difficulty in ensuring the efficiency of manual diagnosis,an auxiliary diagnosis scheme for dental caries in oral small teeth based on image processing and deep learning is designed.Based on the U-Net semantic segmentation model and the YOLOv5 object detection model,this scheme identifies caries lesions in oral small dental films,and the trained model can effectively identify the caries area,and the optimal mAP@0.5 value reaches 95.3%and the accuracy reaches 95.9%.At the same time,combined with the graphical interface development tool PyQt5,the development of a medically suitable caries detection and auxiliary diagnosis system is completed,providing doctors with easy-to-operate auxiliary software,thereby effectively improving medical efficiency.

关 键 词：人工智能 深度学习 语义分割 目标检测 X射线图像 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术]