Clinically applicable histopathological diagnosis system for gastric cancer detection using deep learning

Zhigang Song, et al.

Like Comment

Received Date: 27th January 20

Zhigang Song, Shuangmei Zou, Weixun Zhou, Yong Huang, Liwei Shao, Jing Yuan, Xiangnan Gou, Wei Jin, Zhanbo Wang, Xin Chen, Xiaohui Ding, Jinhong Liu, Chunkai Yu, Calvin Ku, Cancheng Liu, Zhuo Sun, Gang Xu, Yuefeng Wang, Xiaoqing Zhang, Dandan Wang, Shuhao Wang, Wei Xu, Richard C. Davis, Huaiyin Shi

Gastric cancer is among the malignant tumors with the highest incidence and mortality rates. Early detection and accurate histopathological diagnosis of gastric cancer are essential factors that can help increase the chances of successful treatment. While the worldwide shortage of pathologists has imposed burdens on the current histopathology service, it also offers a unique opportunity for the use of artificial intelligence assistance systems to alleviate the workload and increase diagnostic accuracy. To the best of our knowledge, there has not been a clinically applicable histopathological assistance system with high accuracy, and can generalize to whole slide images created with diverse digital scanner models from different hospitals. Here, we report the clinically applicable artificial intelligence assistance system developed at the Chinese PLA General Hospital, China, using a deep convolutional neural network trained with 2,123 pixel-level annotated whole slide images. The model achieved a sensitivity near 100% and an average specificity of 80.6% on a real world test dataset, which included 3,212 whole slide images digitalized with three scanner models. We showed that the system would aid pathologists in improving diagnostic accuracy and preventing misdiagnosis. Moreover, we demonstrated that our system could perform robustly with 1,582 whole slide images from two other medical centers. Our study proves the feasibility and benefits of using histopathological artificial intelligence assistance systems in routine practice scenarios.

Read in full at bioRxiv.

This is an abstract of a preprint hosted on an independent third party site. It has not been peer reviewed but is currently under consideration at Nature Communications

Go to the profile of Nature Communications

Nature Communications

Nature Research, Springer Nature