A computer-assisted algorithm for narrow-band imaging-based tissue characterization in Barrett’s esophagus - 02/09/20

Doi : 10.1016/j.gie.2020.05.050

Maarten R. Struyvenberg, MD ^1,^{^{∗
Drs Struyvenberg and de Groof contributed equally to this article and share first authorship.}}, Albert J. de Groof, MD ^1,^{^{∗
Drs Struyvenberg and de Groof contributed equally to this article and share first authorship.}}, Joost van der Putten, MSc ², Fons van der Sommen, PhD ², Francisco Baldaque-Silva, MD, PhD ⁴, Masami Omae, MD ⁴, Roos Pouw, MD, PhD ¹, Raf Bisschops, MD, PhD ⁵, Michael Vieth, MD, PhD ⁶, Erik J. Schoon, MD, PhD ³, Wouter L. Curvers, MD, PhD ³, Peter H. de With, PhD ², Jacques J. Bergman, MD, PhD ^1,^⁎
¹ Department of Gastroenterology and Hepatology, Amsterdam UMC, University of Amsterdam, Amsterdam, the Netherlands
² Department of Electrical Engineering, VCA Group, Eindhoven University of Technology, Eindhoven, the Netherlands
³ Department of Gastroenterology and Hepatology, Catharina Hospital Eindhoven, Eindhoven, the Netherlands
⁴ Department of Gastroenterology and Hepatology, Karolinska University Hospital, Stockholm, Sweden
⁵ Department of Gastroenterology and Hepatology, University Hospitals Leuven, Leuven, Belgium
⁶ Institute of Pathology, Bayreuth Clinic, Bayreuth, Germany

^∗Reprint requests: J.J. Bergman, MD, PhD, Professor of Gastrointestinal Endoscopy, Director of Endoscopy, Academic Medical Center Amsterdam, Meibergdreef 9, 1105 AZ Amsterdam, the Netherlands.Professor of Gastrointestinal EndoscopyDirector of EndoscopyAcademic Medical Center AmsterdamMeibergdreef 9AZ Amsterdam1105the Netherlands

Sous presse. Épreuves corrigées par l'auteur. Disponible en ligne depuis le Wednesday 02 September 2020

Abstract

Background and Aims

The endoscopic evaluation of narrow-band imaging (NBI) zoom imagery in Barrett’s esophagus (BE) is associated with suboptimal diagnostic accuracy and poor interobserver agreement. Computer-aided diagnosis (CAD) systems may assist endoscopists in the characterization of Barrett’s mucosa. Our aim was to demonstrate the feasibility of a deep-learning CAD system for tissue characterization of NBI zoom imagery in BE.

Methods

The CAD system was first trained using 494,364 endoscopic images of general endoscopic imagery. Next, 690 neoplastic BE and 557 nondysplastic BE (NDBE) white-light endoscopy overview images were used for refinement training. Subsequently, a third dataset of 112 neoplastic and 71 NDBE NBI zoom images with histologic correlation was used for training and internal validation. Finally, the CAD system was further trained and validated with a fourth, histologically confirmed dataset of 59 neoplastic and 98 NDBE NBI zoom videos. Performance was evaluated using fourfold cross-validation. The primary outcome was the diagnostic performance of the CAD system for classification of neoplasia in NBI zoom videos.

Results

The CAD system demonstrated accuracy, sensitivity, and specificity for detection of BE neoplasia using NBI zoom images of 84%, 88%, and 78%, respectively. In total, 30,021 individual video frames were analyzed by the CAD system. Accuracy, sensitivity, and specificity of the video-based CAD system were 83% (95% confidence interval [CI], 78%-89%), 85% (95% CI, 76%-94%), and 83% (95% CI, 76%-90%), respectively. The mean assessment speed was 38 frames per second.

Conclusion

We have demonstrated promising diagnostic accuracy of predicting the presence/absence of Barrett’s neoplasia on histologically confirmed unaltered NBI zoom videos with fast corresponding assessment time.

Le texte complet de cet article est disponible en PDF.

Abbreviations : AUC, BE, CADe, CADx, CI, EAC, HGD, IQR, NBI, NDBE, ROC, WLE

Plan

Introduction

Methods

Setting

CAD system architecture

Datasets used for CAD (pre)training and testing

Automated video analysis

Comparison of pretraining strategies

Outcome measures

Statistical analysis

Results

Primary outcome measurements

Secondary outcome measures

Discussion

DISCLOSURE: Dr Bergmann has received research support from NinePoint Medical and speaker fees from Fujifilm. Dr Baldaque-Silva has received research support from Boston Scientific. Dr Bisschops has received research support, consulting fees, and speaker fees from Fujifilm. Dr Vieth has received honoraria for lecturing from Falk, Shire, and Olympus. All other authors disclosed no financial relationships relevant to this publication.