Endoscopy-Driven Pretraining for Classification of Dysplasia in Barrett’s Esophagus with Endoscopic Narrow-Band Imaging Zoom Videos
Endoscopic diagnosis of early neoplasia in Barrett’s Esophagus is generally a two-step process of primary detection in overview, followed by detailed inspection of any visible abnormalities using Narrow Band Imaging (NBI). However, endoscopists struggle with evaluating NBI-zoom imagery of subtle abn...
Main Authors: | , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-05-01
|
Series: | Applied Sciences |
Subjects: | |
Online Access: | https://www.mdpi.com/2076-3417/10/10/3407 |
id |
doaj-d32130ee93344c02b3609fd633504f8e |
---|---|
record_format |
Article |
spelling |
doaj-d32130ee93344c02b3609fd633504f8e2020-11-25T02:20:04ZengMDPI AGApplied Sciences2076-34172020-05-01103407340710.3390/app10103407Endoscopy-Driven Pretraining for Classification of Dysplasia in Barrett’s Esophagus with Endoscopic Narrow-Band Imaging Zoom VideosJoost van der Putten0Maarten Struyvenberg1Jeroen de Groof2Wouter Curvers3Erik Schoon4Francisco Baldaque-Silva5Jacques Bergman6Fons van der Sommen7Peter H.N. de With8Department of Electrical Engineering, Video Coding and Architectures, Eindhoven University of Technology, 5612 AZ Eindhoven, Noord-Brabant, The NetherlandsDepartment of Gastroenterology and Hepatology, Amsterdam UMC, University of Amsterdam, 1105 AZ Amsterdam, Noord-Holland, The NetherlandsDepartment of Gastroenterology and Hepatology, Amsterdam UMC, University of Amsterdam, 1105 AZ Amsterdam, Noord-Holland, The NetherlandsDepartment of Gastroenterology and Hepatology, Catharina Hospital, 5623 EJ Eindhoven, Noord-Brabant, The NetherlandsDepartment of Gastroenterology and Hepatology, Catharina Hospital, 5623 EJ Eindhoven, Noord-Brabant, The NetherlandsDepartment of Gastroenterology and Hepatology, Karolinksa University Hospital, SE-171 76 Solna, Stockholm, SwedenDepartment of Gastroenterology and Hepatology, Amsterdam UMC, University of Amsterdam, 1105 AZ Amsterdam, Noord-Holland, The NetherlandsDepartment of Electrical Engineering, Video Coding and Architectures, Eindhoven University of Technology, 5612 AZ Eindhoven, Noord-Brabant, The NetherlandsDepartment of Electrical Engineering, Video Coding and Architectures, Eindhoven University of Technology, 5612 AZ Eindhoven, Noord-Brabant, The NetherlandsEndoscopic diagnosis of early neoplasia in Barrett’s Esophagus is generally a two-step process of primary detection in overview, followed by detailed inspection of any visible abnormalities using Narrow Band Imaging (NBI). However, endoscopists struggle with evaluating NBI-zoom imagery of subtle abnormalities. In this work, we propose the first results of a deep learning system for the characterization of NBI-zoom imagery of Barrett’s Esophagus with an accuracy, sensitivity, and specificity of 83.6%, 83.1%, and 84.0%, respectively. We also show that endoscopy-driven pretraining outperforms two models, one without pretraining as well as a model with ImageNet initialization. The final model outperforms absence of pretraining by approximately 10% and the performance is 2% higher in terms of accuracy compared to ImageNet pretraining. Furthermore, the practical deployment of our model is not hampered by ImageNet licensing, thereby paving the way for clinical application.https://www.mdpi.com/2076-3417/10/10/3407endoscopic zoom imageryBarrett’s esophagusdeep learningclassificationmachine learning |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Joost van der Putten Maarten Struyvenberg Jeroen de Groof Wouter Curvers Erik Schoon Francisco Baldaque-Silva Jacques Bergman Fons van der Sommen Peter H.N. de With |
spellingShingle |
Joost van der Putten Maarten Struyvenberg Jeroen de Groof Wouter Curvers Erik Schoon Francisco Baldaque-Silva Jacques Bergman Fons van der Sommen Peter H.N. de With Endoscopy-Driven Pretraining for Classification of Dysplasia in Barrett’s Esophagus with Endoscopic Narrow-Band Imaging Zoom Videos Applied Sciences endoscopic zoom imagery Barrett’s esophagus deep learning classification machine learning |
author_facet |
Joost van der Putten Maarten Struyvenberg Jeroen de Groof Wouter Curvers Erik Schoon Francisco Baldaque-Silva Jacques Bergman Fons van der Sommen Peter H.N. de With |
author_sort |
Joost van der Putten |
title |
Endoscopy-Driven Pretraining for Classification of Dysplasia in Barrett’s Esophagus with Endoscopic Narrow-Band Imaging Zoom Videos |
title_short |
Endoscopy-Driven Pretraining for Classification of Dysplasia in Barrett’s Esophagus with Endoscopic Narrow-Band Imaging Zoom Videos |
title_full |
Endoscopy-Driven Pretraining for Classification of Dysplasia in Barrett’s Esophagus with Endoscopic Narrow-Band Imaging Zoom Videos |
title_fullStr |
Endoscopy-Driven Pretraining for Classification of Dysplasia in Barrett’s Esophagus with Endoscopic Narrow-Band Imaging Zoom Videos |
title_full_unstemmed |
Endoscopy-Driven Pretraining for Classification of Dysplasia in Barrett’s Esophagus with Endoscopic Narrow-Band Imaging Zoom Videos |
title_sort |
endoscopy-driven pretraining for classification of dysplasia in barrett’s esophagus with endoscopic narrow-band imaging zoom videos |
publisher |
MDPI AG |
series |
Applied Sciences |
issn |
2076-3417 |
publishDate |
2020-05-01 |
description |
Endoscopic diagnosis of early neoplasia in Barrett’s Esophagus is generally a two-step process of primary detection in overview, followed by detailed inspection of any visible abnormalities using Narrow Band Imaging (NBI). However, endoscopists struggle with evaluating NBI-zoom imagery of subtle abnormalities. In this work, we propose the first results of a deep learning system for the characterization of NBI-zoom imagery of Barrett’s Esophagus with an accuracy, sensitivity, and specificity of 83.6%, 83.1%, and 84.0%, respectively. We also show that endoscopy-driven pretraining outperforms two models, one without pretraining as well as a model with ImageNet initialization. The final model outperforms absence of pretraining by approximately 10% and the performance is 2% higher in terms of accuracy compared to ImageNet pretraining. Furthermore, the practical deployment of our model is not hampered by ImageNet licensing, thereby paving the way for clinical application. |
topic |
endoscopic zoom imagery Barrett’s esophagus deep learning classification machine learning |
url |
https://www.mdpi.com/2076-3417/10/10/3407 |
work_keys_str_mv |
AT joostvanderputten endoscopydrivenpretrainingforclassificationofdysplasiainbarrettsesophaguswithendoscopicnarrowbandimagingzoomvideos AT maartenstruyvenberg endoscopydrivenpretrainingforclassificationofdysplasiainbarrettsesophaguswithendoscopicnarrowbandimagingzoomvideos AT jeroendegroof endoscopydrivenpretrainingforclassificationofdysplasiainbarrettsesophaguswithendoscopicnarrowbandimagingzoomvideos AT woutercurvers endoscopydrivenpretrainingforclassificationofdysplasiainbarrettsesophaguswithendoscopicnarrowbandimagingzoomvideos AT erikschoon endoscopydrivenpretrainingforclassificationofdysplasiainbarrettsesophaguswithendoscopicnarrowbandimagingzoomvideos AT franciscobaldaquesilva endoscopydrivenpretrainingforclassificationofdysplasiainbarrettsesophaguswithendoscopicnarrowbandimagingzoomvideos AT jacquesbergman endoscopydrivenpretrainingforclassificationofdysplasiainbarrettsesophaguswithendoscopicnarrowbandimagingzoomvideos AT fonsvandersommen endoscopydrivenpretrainingforclassificationofdysplasiainbarrettsesophaguswithendoscopicnarrowbandimagingzoomvideos AT peterhndewith endoscopydrivenpretrainingforclassificationofdysplasiainbarrettsesophaguswithendoscopicnarrowbandimagingzoomvideos |
_version_ |
1724873697773223936 |