人工智能在成釉细胞瘤病理诊断中的初步应用  

Preliminary application of artificial intelligence in the pathological diagnosis of ameloblastoma

作　　者：乔馨玮 李茂 申泽良 张琳涵 郑志建 汤亚玲 QIAO Xin-wei;LI Mao;SHEN Ze-liang;ZHANG Lin-han;ZHENG Zhi-jian;TANG Ya-ling(State Key Laboratory of Oral Diseases&National Clinical Research Center for Oral Diseases,Department of Pathology,West China Stomatology Hospital,Sichuan University,Chengdu 610041,Sichuan Province,China)

机构地区：[1]口腔疾病防治全国重点实验室,国家口腔疾病临床医学研究中心,四川大学华西口腔医院病理科,四川成都610041

出　　处：《中国口腔颌面外科杂志》2025年第2期122-128,共7页China Journal of Oral and Maxillofacial Surgery

基　　金：四川大学华西口腔医院临床研究项目(LCYJ-MS-202308)。

摘　　要：目的:研究人工智能用于成釉细胞瘤病理诊断的效果,初步探索人工智能在口腔病理学领域中的应用。方法:以90例成釉细胞瘤的病理图像作为研究对象,构建U-net型结构的神经网络,将90幅成釉细胞瘤的H-E图像分为训练集72幅图、验证集9幅图和测试集9幅图,分别用于训练模型和测试模型,最后利用平均交并比(mean intersection over union,MIoU)和受试者工作特征(receiver operating characteristic,ROC)曲线评价U-net网络模型在成釉细胞瘤上皮识别中的能力。结果:U-net模型分割阴性区域的mIoU为0.818,分割阳性区域的mIoU为0.846,ROC曲线下面积为0.92。结论:U-net网络模型对成釉细胞瘤阳性区域和阴性区域具有良好的分割结果,同时能够鉴别阴性切片与阳性切片,能够初步用于成釉细胞瘤的病理诊断,有望进一步大样本验证后在临床逐步推广。PURPOSE:To investigate the effect of artificial intelligence in the pathological diagnosis of ameloblastoma,and to preliminarily explore the value of artificial intelligence in the field of oral pathology.METHODS:The pathological images of 90 cases of ameloblastoma were used as the research objects,and the U-net-like structure neural network was constructed.The 90 H-E images of ameloblastoma were divided into a training set(72 images),a validation set(9 images)and a test set(9 images)for training and testing the model respectively.The mIoU and ROC curve were used to evaluate the ability of the U-net network model in the identification of ameloblastoma epithelium.RESULTS:The mIoU of negative area segmented by U-net model was 0.818 and the positive area was 0.846.The area under the ROC curve was 0.92.CONCLUSIONS:The U-net network model has a good segmentation for the positive and negative regions of ameloblastoma,and can distinguish between negative and positive sections.It can be preliminarily applied to the pathological diagnosis of ameloblastoma,and is expected to be gradually popularized in clinical practice after further validation with large samples.

关 键 词：成釉细胞瘤 病理诊断 人工智能 口腔病理学 U-net网络 mIoU ROC曲线 

分 类 号：R739.8[医药卫生—肿瘤]