Deep-Learning for the Diagnosis of Esophageal Cancers and Precursor Lesions in Endoscopic Images: A Model Establishment and Nationwide Multicenter Performance Verification Study
Background: Suspicion of lesions and prediction of the histology of esophageal cancers or premalignant lesions in endoscopic images are not yet accurate. The local feature selection and optimization functions of the model enabled an accurate analysis of images in deep learning. Objectives: To establ...
Main Authors: | , , , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2022-06-01
|
Series: | Journal of Personalized Medicine |
Subjects: | |
Online Access: | https://www.mdpi.com/2075-4426/12/7/1052 |
_version_ | 1797405905190387712 |
---|---|
author | Eun Jeong Gong Chang Seok Bang Kyoungwon Jung Su Jin Kim Jong Wook Kim Seung In Seo Uhmyung Lee You Bin Maeng Ye Ji Lee Jae Ick Lee Gwang Ho Baik Jae Jun Lee |
author_facet | Eun Jeong Gong Chang Seok Bang Kyoungwon Jung Su Jin Kim Jong Wook Kim Seung In Seo Uhmyung Lee You Bin Maeng Ye Ji Lee Jae Ick Lee Gwang Ho Baik Jae Jun Lee |
author_sort | Eun Jeong Gong |
collection | DOAJ |
description | Background: Suspicion of lesions and prediction of the histology of esophageal cancers or premalignant lesions in endoscopic images are not yet accurate. The local feature selection and optimization functions of the model enabled an accurate analysis of images in deep learning. Objectives: To establish a deep-learning model to diagnose esophageal cancers, precursor lesions, and non-neoplasms using endoscopic images. Additionally, a nationwide prospective multicenter performance verification was conducted to confirm the possibility of real-clinic application. Methods: A total of 5162 white-light endoscopic images were used for the training and internal test of the model classifying esophageal cancers, dysplasias, and non-neoplasms. A no-code deep-learning tool was used for the establishment of the deep-learning model. Prospective multicenter external tests using 836 novel images from five hospitals were conducted. The primary performance metric was the external-test accuracy. An attention map was generated and analyzed to gain the explainability. Results: The established model reached 95.6% (95% confidence interval: 94.2–97.0%) internal-test accuracy (precision: 78.0%, recall: 93.9%, F1 score: 85.2%). Regarding the external tests, the accuracy ranged from 90.0% to 95.8% (overall accuracy: 93.9%). There was no statistical difference in the number of correctly identified the region of interest for the external tests between the expert endoscopist and the established model using attention map analysis (<i>P</i> = 0.11). In terms of the dysplasia subgroup, the number of correctly identified regions of interest was higher in the deep-learning model than in the endoscopist group, although statistically insignificant (<i>P</i> = 0.48). Conclusions: We established a deep-learning model that accurately classifies esophageal cancers, precursor lesions, and non-neoplasms. This model confirmed the potential for generalizability through multicenter external tests and explainability through the attention map analysis. |
first_indexed | 2024-03-09T03:17:58Z |
format | Article |
id | doaj.art-7133ae2ddd224b1eae6ca8029b1eb974 |
institution | Directory Open Access Journal |
issn | 2075-4426 |
language | English |
last_indexed | 2024-03-09T03:17:58Z |
publishDate | 2022-06-01 |
publisher | MDPI AG |
record_format | Article |
series | Journal of Personalized Medicine |
spelling | doaj.art-7133ae2ddd224b1eae6ca8029b1eb9742023-12-03T15:16:17ZengMDPI AGJournal of Personalized Medicine2075-44262022-06-01127105210.3390/jpm12071052Deep-Learning for the Diagnosis of Esophageal Cancers and Precursor Lesions in Endoscopic Images: A Model Establishment and Nationwide Multicenter Performance Verification StudyEun Jeong Gong0Chang Seok Bang1Kyoungwon Jung2Su Jin Kim3Jong Wook Kim4Seung In Seo5Uhmyung Lee6You Bin Maeng7Ye Ji Lee8Jae Ick Lee9Gwang Ho Baik10Jae Jun Lee11Department of Internal Medicine, Gangneung Asan Hospital, University of Ulsan College of Medicine, Gangneung 25440, KoreaDepartment of Internal Medicine, Hallym University College of Medicine, Chuncheon 24253, KoreaDepartment of Internal Medicine, Kosin University College of Medicine, Busan 49267, KoreaDepartment of Internal Medicine, Pusan National University School of Medicine and Biomedical Research Institute, Pusan National University Yangsan Hospital, Yangsan 50615, KoreaDepartment of Internal Medicine, Inje University Ilsan Paik Hospital, Goyang 10380, KoreaDepartment of Internal Medicine, Hallym University College of Medicine, Chuncheon 24253, KoreaDepartment of Medicine, Hallym University College of Medicine, Chuncheon 24253, KoreaDepartment of Medicine, Hallym University College of Medicine, Chuncheon 24253, KoreaDepartment of Biomedical Science, Hallym University, Chuncheon 24252, KoreaDepartment of Life Science, Hallym University, Chuncheon 24252, KoreaDepartment of Internal Medicine, Hallym University College of Medicine, Chuncheon 24253, KoreaInstitute of New Frontier Research, Hallym University College of Medicine, Chuncheon 24253, KoreaBackground: Suspicion of lesions and prediction of the histology of esophageal cancers or premalignant lesions in endoscopic images are not yet accurate. The local feature selection and optimization functions of the model enabled an accurate analysis of images in deep learning. Objectives: To establish a deep-learning model to diagnose esophageal cancers, precursor lesions, and non-neoplasms using endoscopic images. Additionally, a nationwide prospective multicenter performance verification was conducted to confirm the possibility of real-clinic application. Methods: A total of 5162 white-light endoscopic images were used for the training and internal test of the model classifying esophageal cancers, dysplasias, and non-neoplasms. A no-code deep-learning tool was used for the establishment of the deep-learning model. Prospective multicenter external tests using 836 novel images from five hospitals were conducted. The primary performance metric was the external-test accuracy. An attention map was generated and analyzed to gain the explainability. Results: The established model reached 95.6% (95% confidence interval: 94.2–97.0%) internal-test accuracy (precision: 78.0%, recall: 93.9%, F1 score: 85.2%). Regarding the external tests, the accuracy ranged from 90.0% to 95.8% (overall accuracy: 93.9%). There was no statistical difference in the number of correctly identified the region of interest for the external tests between the expert endoscopist and the established model using attention map analysis (<i>P</i> = 0.11). In terms of the dysplasia subgroup, the number of correctly identified regions of interest was higher in the deep-learning model than in the endoscopist group, although statistically insignificant (<i>P</i> = 0.48). Conclusions: We established a deep-learning model that accurately classifies esophageal cancers, precursor lesions, and non-neoplasms. This model confirmed the potential for generalizability through multicenter external tests and explainability through the attention map analysis.https://www.mdpi.com/2075-4426/12/7/1052convolutional neural networkdeep learningendoscopyesophageal cancers |
spellingShingle | Eun Jeong Gong Chang Seok Bang Kyoungwon Jung Su Jin Kim Jong Wook Kim Seung In Seo Uhmyung Lee You Bin Maeng Ye Ji Lee Jae Ick Lee Gwang Ho Baik Jae Jun Lee Deep-Learning for the Diagnosis of Esophageal Cancers and Precursor Lesions in Endoscopic Images: A Model Establishment and Nationwide Multicenter Performance Verification Study Journal of Personalized Medicine convolutional neural network deep learning endoscopy esophageal cancers |
title | Deep-Learning for the Diagnosis of Esophageal Cancers and Precursor Lesions in Endoscopic Images: A Model Establishment and Nationwide Multicenter Performance Verification Study |
title_full | Deep-Learning for the Diagnosis of Esophageal Cancers and Precursor Lesions in Endoscopic Images: A Model Establishment and Nationwide Multicenter Performance Verification Study |
title_fullStr | Deep-Learning for the Diagnosis of Esophageal Cancers and Precursor Lesions in Endoscopic Images: A Model Establishment and Nationwide Multicenter Performance Verification Study |
title_full_unstemmed | Deep-Learning for the Diagnosis of Esophageal Cancers and Precursor Lesions in Endoscopic Images: A Model Establishment and Nationwide Multicenter Performance Verification Study |
title_short | Deep-Learning for the Diagnosis of Esophageal Cancers and Precursor Lesions in Endoscopic Images: A Model Establishment and Nationwide Multicenter Performance Verification Study |
title_sort | deep learning for the diagnosis of esophageal cancers and precursor lesions in endoscopic images a model establishment and nationwide multicenter performance verification study |
topic | convolutional neural network deep learning endoscopy esophageal cancers |
url | https://www.mdpi.com/2075-4426/12/7/1052 |
work_keys_str_mv | AT eunjeonggong deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT changseokbang deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT kyoungwonjung deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT sujinkim deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT jongwookkim deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT seunginseo deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT uhmyunglee deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT youbinmaeng deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT yejilee deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT jaeicklee deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT gwanghobaik deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy AT jaejunlee deeplearningforthediagnosisofesophagealcancersandprecursorlesionsinendoscopicimagesamodelestablishmentandnationwidemulticenterperformanceverificationstudy |