中国实用外科杂志

中国实用外科杂志

• 论蓍 • 上一篇    下一篇

基于深度神经网络Mask R-CNN胆囊癌辅助识别系统临床应用价值研究

尹梓名1，孙大运1，2，翁    昊2，3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3，任    泰2，3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3，杨自逸2，3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3，李永盛3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3，4[Author]) AND 1[Journal]) AND year[Order])" target="_blank">4，王广义5，王传磊5，曹    宏6，刘颖斌3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3，4[Author]) AND 1[Journal]) AND year[Order])" target="_blank">4，7，束翌俊2，3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3[Author]) AND 1[Journal]) AND year[Order])" target="_blank">3   

  1. 1上海理工大学医疗器械与食品学院，上海 200093；2上海交通大学医学院附属新华医院普外科，上海 200092；3上海市胆道疾病重点实验室 上海交通大学医学院胆道疾病研究中心，上海 200092；4上海交通大学医学院附属仁济医院胆胰外科，上海 200127；5吉林大学第一医院肝胆胰外一科，吉林长春130021；6吉林大学中日联谊医院普外科，吉林长春 130033；7国家癌基因重点实验室，上海 200127
  • 出版日期:2021-03-01 发布日期:2021-03-18

  • Online:2021-03-01 Published:2021-03-18

摘要：  目的    探讨基于深度神经网络的目标检测技术在腹部双源CT胆囊癌辅助识别系统的临床应用价值。方法    选取2017年1月至2019年12月上海交通大学医学院附属新华医院普外科、吉林大学第一医院肝胆胰外一科和吉林大学中日联谊医院普外科收治的88例病理学检查诊断明确的胆囊癌，28例慢性胆囊炎胆囊结石病人和29例正常胆囊（影像学检查胆囊正常）病人，均行腹部双源CT检查。随机选取101例作为训练组，29例作为验证组，15例作为测试组。首先，利用已标注的10 409张腹部双源CT图像对Mask R-CNN模型进行学习，从而建立自动胆囊癌辅助识别系统。然后对验证组的2974张CT图像通过专业的医师对其进行判断识别，与Mask R-CNN得出的结果进行对比分析。通过不同交并比阈值（IoU）下的平均检测精度（AP）和平均召回率（AR）来对性能进行评估。结果    计算机通过学习组不断迭代训练，Mask R-CNN的损失函数值收敛，诊断误差不断降低。在IoU为0.5时，Mask R-CNN的边界框和掩膜的AP分别为0.929和0.929，IoU为0.75时的边界框和掩膜AP分别为0.901和0.890，IoU为0.5：0.95时的边界框和掩膜AP分别为0.723和0.707，平均召回率分别为0.794和0.774，模型的性能良好。结论    基于深度神经网络的Mask R-CNN胆囊癌辅助识别系统具有较高的准确率和性能，可辅助进行临床诊断。

关键词: 胆囊癌, 深度学习, 目标检测, 人工智能

Abstract: Application value of gallbladder cancer recognition based on deep neural network Mask R-CNN        YIN Zi-ming*,SUN Da-yun,WENG Hao,et al.  *School of Medical Instrument and Food Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
Corresponding author：LIU Ying-bin，E-mail：laoniulyb@163.com；SHU Yi-jun，E-mail：shuyijun19881125@163.com
Abstract    Objective    To investigate the clinical significance of gallbladder cancer recognition based on deep neural network Mask R-CNN. Methods    The research selected 88 patients with gallbladder cancer，28 patients with chronic cholecystitis with gallbladder stones and 29 patients with normal gallbladder tissues who underwent abdominal dual-source CT examination and diagnosed by pathology from January 2017 to December 2019 in Department of General Surgery, Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine,First Department of Hepatobiliary and Pancreatic Surgery of Jilin University and Department of General Surgery,China-Japan Friendship Hospital of Jilin University. One hundred and one cases as the training group,29 cases as the verification group and 15 cases as the test group were selected randomly. Firstly,using deep learning technique,researchers trained the 10409 abdominal dual-source CT images data of learning group with convolution neural network to simulate the judgment process of radiologists,and established an artificial intelligence automatic recognition system for gallbladder cancer. Then,2974 images of the validation group were clinically validated. The performance was evaluated by the average detection accuracy (AP) and average recall (AR) under different intersection ratio thresholds (IoU). Results    After continuous iteration training of the learning group data,the loss function value of Mask R-CNN decreased continuously, and the diagnostic error decreased continuously. When the IoU was 0.5,the AP of the bounding box and mask of Mask R-CNN were 0.929 and 0.929,respectively. When the IoU was 0.75,the AP of the bounding box and mask were 0.901 and 0.890,respectively.When the IoU was 0.5:0.95, the AP of the bounding box and mask were 0.723 and 0.707,and the AR was 0.794 and 0.774；respectively.It suggested the performance of the model was good. Conclusion    The Mask R-CNN for gallbladder cancer automatic recognition system based on deep neural network has high accuracy and high efficiency, and has the clinical significance of auxiliary diagnosis.

Key words: gallbladder cancer, deep learning, object detection, artificial intelligence