Automatic deep learning-based colorectal adenoma detection system and its similarities with pathologists
Objectives The microscopic evaluation of slides has been gradually moving towards all digital in recent years, leading to the possibility for computer-aided diagnosis. It is worthwhile to know the similarities between deep learning models and pathologists before we put them into practical scenarios....
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMJ Publishing Group
2020-09-01
|
| Series: | BMJ Open |
| Online Access: | https://bmjopen.bmj.com/content/10/9/e036423.full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841554153436872704 |
|---|---|
| author | Xin Chen Wei Xu Yong Huang Wei Jin Gang Xu Jing Yuan Zhigang Song Chunkai Yu Shuangmei Zou Wenmiao Wang Xiaohui Ding Jinhong Liu Liwei Shao Xiangnan Gou Zhanbo Wang Huang Chen Cancheng Liu Zhuo Sun Calvin Ku Yongqiang Zhang Xianghui Dong Shuhao Wang Ning Lv Huaiyin Shi |
| author_facet | Xin Chen Wei Xu Yong Huang Wei Jin Gang Xu Jing Yuan Zhigang Song Chunkai Yu Shuangmei Zou Wenmiao Wang Xiaohui Ding Jinhong Liu Liwei Shao Xiangnan Gou Zhanbo Wang Huang Chen Cancheng Liu Zhuo Sun Calvin Ku Yongqiang Zhang Xianghui Dong Shuhao Wang Ning Lv Huaiyin Shi |
| author_sort | Xin Chen |
| collection | DOAJ |
| description | Objectives The microscopic evaluation of slides has been gradually moving towards all digital in recent years, leading to the possibility for computer-aided diagnosis. It is worthwhile to know the similarities between deep learning models and pathologists before we put them into practical scenarios. The simple criteria of colorectal adenoma diagnosis make it to be a perfect testbed for this study.Design The deep learning model was trained by 177 accurately labelled training slides (156 with adenoma). The detailed labelling was performed on a self-developed annotation system based on iPad. We built the model based on DeepLab v2 with ResNet-34. The model performance was tested on 194 test slides and compared with five pathologists. Furthermore, the generalisation ability of the learning model was tested by extra 168 slides (111 with adenoma) collected from two other hospitals.Results The deep learning model achieved an area under the curve of 0.92 and obtained a slide-level accuracy of over 90% on slides from two other hospitals. The performance was on par with the performance of experienced pathologists, exceeding the average pathologist. By investigating the feature maps and cases misdiagnosed by the model, we found the concordance of thinking process in diagnosis between the deep learning model and pathologists.Conclusions The deep learning model for colorectal adenoma diagnosis is quite similar to pathologists. It is on-par with pathologists’ performance, makes similar mistakes and learns rational reasoning logics. Meanwhile, it obtains high accuracy on slides collected from different hospitals with significant staining configuration variations. |
| format | Article |
| id | doaj-art-ff9006e64fcf44978c6e18372264a33e |
| institution | Kabale University |
| issn | 2044-6055 |
| language | English |
| publishDate | 2020-09-01 |
| publisher | BMJ Publishing Group |
| record_format | Article |
| series | BMJ Open |
| spelling | doaj-art-ff9006e64fcf44978c6e18372264a33e2025-01-08T19:50:10ZengBMJ Publishing GroupBMJ Open2044-60552020-09-0110910.1136/bmjopen-2019-036423Automatic deep learning-based colorectal adenoma detection system and its similarities with pathologistsXin Chen0Wei Xu1Yong Huang2Wei Jin3Gang Xu4Jing Yuan5Zhigang Song6Chunkai Yu7Shuangmei Zou8Wenmiao Wang9Xiaohui Ding10Jinhong Liu11Liwei Shao12Xiangnan Gou13Zhanbo Wang14Huang Chen15Cancheng Liu16Zhuo Sun17Calvin Ku18Yongqiang Zhang19Xianghui Dong20Shuhao Wang21Ning Lv22Huaiyin Shi23Department of Pathology, Chinese PLA General Hospital, Beijing, ChinaNational Clinical Research Center for Cardiovascular Diseases, NHC Key Laboratory of Clinical Research for Cardiovascular Medications, State Key Laboratory of Cardiovascular Disease, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, ChinaDepartment of Immunization Programme Planning, Guangzhou Center for Disease Control and Prevention, Guangzhou, China8 Heart Failure Center, Luwan Branch, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, ChinaRenal Department, Leicester General Hospital, Leicester, UKDepartment of Pathology, Chinese PLA General Hospital, Beijing, ChinaDepartment of Pathology, Chinese PLA General Hospital, Beijing, ChinaDepartment of Pathology, Capital Medical University Affiliated Beijing Shijitan Hospital, Beijing, China3 Department of Pathology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, ChinaDepartment of Pathology, National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, ChinaDepartment of Pathology, Chinese PLA General Hospital, Beijing, ChinaDepartment of Pathology, Chinese PLA General Hospital, Beijing, ChinaDepartment of Pathology, Chinese PLA General Hospital, Beijing, ChinaDepartment of Pathology, Chinese PLA General Hospital, Beijing, ChinaDepartment of Pathology, Chinese PLA General Hospital, Beijing, ChinaDepartment of Pathology, China-Japan Friendship Hospital, Beijing, ChinaThorough Images, Beijing, ChinaThorough Images, Beijing, ChinaThorough Images, Beijing, ChinaDepartment of Pathology, Chinese PLA General Hospital, Beijing, ChinaDepartment of Pathology, Chinese PLA General Hospital, Beijing, ChinaThorough Lab, Thorough Future, Beijing, ChinaDepartment of Pathology, National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, ChinaDepartment of Pathology, Chinese PLA General Hospital, Beijing, ChinaObjectives The microscopic evaluation of slides has been gradually moving towards all digital in recent years, leading to the possibility for computer-aided diagnosis. It is worthwhile to know the similarities between deep learning models and pathologists before we put them into practical scenarios. The simple criteria of colorectal adenoma diagnosis make it to be a perfect testbed for this study.Design The deep learning model was trained by 177 accurately labelled training slides (156 with adenoma). The detailed labelling was performed on a self-developed annotation system based on iPad. We built the model based on DeepLab v2 with ResNet-34. The model performance was tested on 194 test slides and compared with five pathologists. Furthermore, the generalisation ability of the learning model was tested by extra 168 slides (111 with adenoma) collected from two other hospitals.Results The deep learning model achieved an area under the curve of 0.92 and obtained a slide-level accuracy of over 90% on slides from two other hospitals. The performance was on par with the performance of experienced pathologists, exceeding the average pathologist. By investigating the feature maps and cases misdiagnosed by the model, we found the concordance of thinking process in diagnosis between the deep learning model and pathologists.Conclusions The deep learning model for colorectal adenoma diagnosis is quite similar to pathologists. It is on-par with pathologists’ performance, makes similar mistakes and learns rational reasoning logics. Meanwhile, it obtains high accuracy on slides collected from different hospitals with significant staining configuration variations.https://bmjopen.bmj.com/content/10/9/e036423.full |
| spellingShingle | Xin Chen Wei Xu Yong Huang Wei Jin Gang Xu Jing Yuan Zhigang Song Chunkai Yu Shuangmei Zou Wenmiao Wang Xiaohui Ding Jinhong Liu Liwei Shao Xiangnan Gou Zhanbo Wang Huang Chen Cancheng Liu Zhuo Sun Calvin Ku Yongqiang Zhang Xianghui Dong Shuhao Wang Ning Lv Huaiyin Shi Automatic deep learning-based colorectal adenoma detection system and its similarities with pathologists BMJ Open |
| title | Automatic deep learning-based colorectal adenoma detection system and its similarities with pathologists |
| title_full | Automatic deep learning-based colorectal adenoma detection system and its similarities with pathologists |
| title_fullStr | Automatic deep learning-based colorectal adenoma detection system and its similarities with pathologists |
| title_full_unstemmed | Automatic deep learning-based colorectal adenoma detection system and its similarities with pathologists |
| title_short | Automatic deep learning-based colorectal adenoma detection system and its similarities with pathologists |
| title_sort | automatic deep learning based colorectal adenoma detection system and its similarities with pathologists |
| url | https://bmjopen.bmj.com/content/10/9/e036423.full |
| work_keys_str_mv | AT xinchen automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT weixu automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT yonghuang automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT weijin automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT gangxu automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT jingyuan automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT zhigangsong automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT chunkaiyu automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT shuangmeizou automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT wenmiaowang automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT xiaohuiding automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT jinhongliu automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT liweishao automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT xiangnangou automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT zhanbowang automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT huangchen automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT canchengliu automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT zhuosun automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT calvinku automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT yongqiangzhang automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT xianghuidong automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT shuhaowang automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT ninglv automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists AT huaiyinshi automaticdeeplearningbasedcolorectaladenomadetectionsystemanditssimilaritieswithpathologists |