摘要
目的验证基于快速区域卷积神经网络萎缩性胃炎-胃癌胃镜图像自动识别系统,并探讨其临床应用价值。方法回顾性收集80例萎缩性胃炎-胃癌患者的640张胃镜图像序列,将所有图像序列作为训练组输入卷积神经网络系统,建立图像自动识别模型,通过读取图像进行验证,记录识别准确率,绘制精确回归曲线。结果模型训练损失在12轮训练后基本趋于0,准确率在第29轮达到最大值98%,在42轮以后也基本稳定在最大值。结论卷积神经网络系统能够帮助消化科医师对萎缩性胃炎-胃癌进行诊断,具有一定的临床应用价值。
Objective To verify the automatic recognition system of atrophic gastritis-gastric cancer gastroscopy pictures based on fast regional convolutional neural network,and to explore its clinical application value.Methods A retrospective collection of 640 gastroscopy picture sequences of 80 patients with atrophic gastritis and gastric cancer was conducted.All picture sequences were input into the convolutional neural network system as a training group.An automatic picture recognition model was established,and the recognition accuracy was recorded by reading the pictures.An accurate regression curve was drawn.Results The model training loss basically tended to 0 after 12 rounds of training,and the accuracy reached the maximum 98%in the 29 thround,and basically stabilized at the maximum after 42 rounds.Conclusion The convolutional neural network system can help gastroenterologists diagnose atrophic gastritisgastric cancer,and has certain clinical application value.
作者
徐伟超
李博林
许亚培
才艳茹
杜朋丽
赵丹阳
王旭东
杨怀金
夏克文
杨倩
Xu Weichao;Li Bolin;Xu Yapei;Cai Yanru;Du Pengli;Zhao Danyang;Wang Xudong;Yang Huaijin;Xia Kewen;Yang Qian(Department of Spleen and Stomach Diseases,Hebei Provincial Hospital of Traditional Chinese Medicine,Shijiazhuang 050011,China;Hebei Provincial Key Laboratory of gastroenteropathy with combination of Traditional Chinese and Western Medicine,Shijiazhuang 050011,China;The 13th Institute of China Electronics Technology Group Corporation,Beijing 100081,China;School of Electronic Information Engineering,Hebei University of Technology,Tianjin 300401,China;Key Laboratory of Big Data Computing of Hebei Province,Tianjin 300401,China)
出处
《世界科学技术-中医药现代化》
CSCD
北大核心
2021年第9期3274-3280,共7页
Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology
基金
河北省科学技术厅省级科技计划资助项目(21377724D、21377740D):香连化浊颗粒阻断慢性萎缩性胃炎“炎-癌转化”的临床研究,负责人:杨倩
河北省自然科学基金委员会青年项目(H2020423017):LSM-POMDP联用在名老中医诊疗胃癌前病变经验诠释中的“算-优”研究,负责人:徐伟超
河北省中医药管理局科研计划项目(2022070):基于模型驱动和机器学习的胃癌中医诊疗数据优化研究,负责人:徐伟超
关键词
萎缩性胃炎
胃癌
卷积神经网络
临床应用
Atrophic gastritis
Gastric cancer
Convolutional neural network
Clinical application
