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 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.

[1]  P. Fockens,et al.  Detection and classification of the mucosal and vascular patterns (mucosal morphology) in Barrett's esophagus by using narrow band imaging. , 2006, Gastrointestinal endoscopy.

[2]  L. Gossner,et al.  Long-term efficacy and safety of endoscopic resection for patients with mucosal adenocarcinoma of the esophagus. , 2014, Gastroenterology.

[3]  Henry Horng-Shing Lu,et al.  Accurate Classification of Diminutive Colorectal Polyps Using Computer-Aided Analysis. , 2017, Gastroenterology.

[4]  D. Whiteman,et al.  Cost-effectiveness of endoscopic surveillance of non-dysplastic Barrett's esophagus. , 2014, Gastrointestinal endoscopy.

[5]  A. Rastogi,et al.  Observer agreement in the assessment of narrowband imaging system surface patterns in Barrett’s esophagus: a multicenter study , 2011, Endoscopy.

[6]  C. Hassan,et al.  Endoscopic management of Barrett’s esophagus: European Society of Gastrointestinal Endoscopy (ESGE) Position Statement , 2017, Endoscopy.

[7]  Kazufumi Kaneda,et al.  Computer-Aided Colorectal Tumor Classification in NBI Endoscopy Using CNN Features , 2016, ArXiv.

[8]  A. Meining,et al.  Deep-Learning System Detects Neoplasia in Patients With Barrett's Esophagus With Higher Accuracy Than Endoscopists in a Multi-Step Training and Validation Study with Benchmarking. , 2019, Gastroenterology.

[9]  Irving Waxman,et al.  Development and Validation of a Classification System to Identify High-Grade Dysplasia and Esophageal Adenocarcinoma in Barrett's Esophagus Using Narrow-Band Imaging. , 2016, Gastroenterology.

[10]  Christoph Palm,et al.  Computer-aided diagnosis using deep learning in the evaluation of early oesophageal adenocarcinoma , 2018, Gut.

[11]  Carmen C. Y. Poon,et al.  Automatic Detection and Classification of Colorectal Polyps by Transferring Low-Level CNN Features From Nonmedical Domain , 2017, IEEE Journal of Biomedical and Health Informatics.

[12]  A. Meining,et al.  The Argos project: The development of a computer-aided detection system to improve detection of Barrett's neoplasia on white light endoscopy , 2019, United European gastroenterology journal.

[13]  Nicolas Chapados,et al.  Real-time differentiation of adenomatous and hyperplastic diminutive colorectal polyps during analysis of unaltered videos of standard colonoscopy using a deep learning model , 2017, Gut.