摘要： 目的 基于牙根尖X线片建立计算机人工智能诊断模型，辅助口腔医生对牙隐裂进行有效诊断。方法 收集182例患者的牙根尖影像数据。其中，确诊牙隐裂77例，未确诊105例。通过蟒蛇编程从牙根尖X线片中提取影像特征，采用套索算法（LASSO）对特征进行筛选，并建立机器学习模型判断牙齿是否患有牙隐裂。结果 筛选出3个高维度纹理特征，机器学习模型在测试集上的受试者操作特性曲线下面积、特异度和灵敏度分别为0.805、0.727和0.815。结论 基于牙根尖X线片建立的机器学习模型能够对牙隐裂进行比较有效的诊断。

机器学习方法辅助诊断牙隐裂的研究

刘洁¹, 张洪晓², 付朗远², 蒋西然²

1. 中国医科大学口腔医学院·附属口腔医院修复二科, 辽宁省口腔疾病重点实验室, 沈阳 110002;
2. 中国医科大学智能医学学院生物医学工程教研室, 沈阳 110042

收稿日期:2022-09-30出版日期:2023-04-30发布日期:2023-04-15
通讯作者:刘洁E-mail:493165009@qq.com
作者简介:刘洁(1986-),女,主治医师,硕士.
基金资助:辽宁省自然科学基金（2021-MS-205）


关键词: 牙折, 机器学习, 计算机辅助诊断, 人工智能
Abstract: Objective A computer-aided artificial intelligence diagnosis model based on the X-ray periapical film was created to help dentists effectively diagnose cracked teeth. Methods A total of 182 periapical radiographs were reviewed. Seventy-seven cases were diagnosed with cracked teeth, while the remaining 105 cases were undiagnosed. Imaging features were extracted from the periapical radiographs using Python programming, and features were selected using the Lasso algorithm to develop a machine learning model that identifies a cracked tooth. Results Three high-dimensional texture features were selected. The area under the curve (AUC), specificity, and sensitivity of the machine learning model on the test set were 0.805, 0.727, and 0.815, respectively. Conclusion The machine learning model based on periapical radiographs is effective in diagnosing cracked teeth.
Key words: tooth fractures, machine learning, computer-aided diagnosis, artificial intelligence
PDF全文下载地址:

https://journal.cmu.edu.cn/CN/article/downloadArticleFile.do?attachType=PDF&id=3197