EndoCV 2021 3rd International Workshop and Chal- lenge on Computer Vision in En- doscopy

-

https://biomedicalimaging.org/2021/challenges-2

Preface for EndoCV2021 Challenge

Endoscopy is a widely used clinical procedure for early detection and prevention of cancer and in ammations in hollow organs such as oesophagus, stomach, colon, rectum and bladder. EndoCV is a crowd sourcing initiative to bring together both the computational scientists and clinical colleagues together to solve imminent challenges in endoscopy image analysis. EndoCV tackles existing challenges that is critical for success in clinical application of computer-aided systems. As part of EndoCV road-map, we have already achieved some important milestones that includes: { Identifying artefacts in endoscopy imaging for quality quanti cation (see our EAD2019 [ 1 ] and EAD2020 [ 2 ] editions) { Detection and segmentation of multi-class disease instances in the entire GI tract (see our EDD2020 [ 2 ] edition)

Albeit there are important steps taken to solve computer-based detection and segmentation problems for polyps, a known cancer precursor, however, due to limited publicly available datasets and lack of heterogeneous population samples robustness and accuracy of methods cannot be guaranteed for varied clinical settings. This is a current bottleneck in robust and accurate method development. This year we extended our collaboration further to Faculty of Medicine, University of Alexandria, Egypt (Prof. Osama E. Salem); Medical Department, Sahlgrenska University Hospital-Molndal, Sweden (Prof. Thomas de Lange); and Oslo University Hospital Ulleval, Oslo, Norway (Kim V. Anonsen). Together with our senior gastroenterologists from previous editions of this challenge and new collaborators from these three centers we have put an e ort to establish a diverse multi-population dataset from 6 di erent centers. We believe that this initiative will further assist in the development of endoscopy image analysis research for improved patient care. EndoCV2021 possess important generalisabilty questions and the participants were challenged to address issues of data shifts due to population variability, acquisition variability and modality under the theme: Addressing generalisability in polyp detection and segmentation.

Training data was released in two phases with all together 5050 frames released at the nal phases II consisting of data from 5 centers and sequence data (mixed center). The data from 6th center was hidden and was part of generalisability test data. Related dataset paper is already available [ 3 ]. All algorithms were evaluated online with the same evaluation metrics for detection, localisation and semantic segmentation. To steer the detection (with localisation) and segmentation tasks research in the right direction we used classically used state-of-the-art metrics1 in computer vision. We established an online leaderboard for round I and round II based on which challenge winners were decided. For the rst time, we introduced a cloudbased inference on NVIDIA V100 GPU to assess methods potential for clinical translation. A third round is the organiser's evaluation round which will be based on rigorous generalisability tests whose results will be presented in a prospective joint-journal paper.

We would like to thank all the participants, organising committee members, and IEEE ISBI 2021 challenge committee for their tremendous support. We would also like to thank all the keynotes, reviewers and sponsors.

Sharib Ali, Ph.D. (Lead organiser)

1 https://github.com/sharibox/EndoCV2021-polyp_det_seg_gen

ii

Preface for EndoCV2021 Workshop Proceeding This volume contains the proceedings of the third edition of the international workshop and challenge on computer vision in endoscopy(EndoCV). Due to the COVID-19 outbreak, the workshop was virtually held as a webinar on the 13th Arpil 2021 (initially planned to be held in Nice, France). For the third time this challenge was colocated with the 18th IEEE International Symposium on Biomedical Imaging (ISBI2021).

This year we received 15 full paper submissions. All the papers were reviewed through CMT by at least 3 reviewers and 1 metareviewer. Ten high quality papers were accepted for publication.

Sharib Ali,

Noha Ghatwary,

Debesh Jha, & Pal Halvorsen (Vol. Editors) Copyright ©2021 for the individual papers by the papers' authors. Copyright ©2021 for the volume as a collection by its editors. This volume and its papers are published under the Creative Commons License Attribution 4.0 International (CC BY 4.0).

EndoCV2021 Challenge Organization

Organising committee

Sharib Ali (lead) Debesh Jha Noha Ghatwary

Program committee

Christian Daul Michael A. Riegler Pal Halvorsen Jens Rittscher

Clinical collaborators

Dominique Lamarque James East Osama Ebada Salem Stefano Realdon Renato Cannizzaro Thomas de Lange IBME, Big Data Institute, University of Oxford, Ofxord, UK SimulaMet, Oslo, Norway University of Lincoln, UK University of Lorraine, CNRS, CRAN, UMR 7039, Nancy, France SimulaMet, Norway SimulaMet, Norway and Oslo Metropolitan University, Norway Department of Engineering Science, Big Data Institute, University of Oxford, UK Consultation Gastroenterology, Ho^pital Ambroise Pare, Paris, France Translational Gastroenterology Unit, John Radcli e Hospital, Oxford, UK Faculty of Medicine, University of Alexandria, Egypt Istituto Oncologico Veneto, IOV-IRCCS, Padova, Italy Centro di Riferimento Oncologico di Aviano (CRO) IRCCS, Italy Medical Department, Sahlgrenska University Hospital-Molndal, Sweden

IT support for GPU cloud inference

Adam Hu man Big Data Institute, University of Oxford, Oxford, UK

Event support

Charlotte Rush

Sponsors

BRC Coordinator for Cardiovascular and Imaging Themes, Division of Cardiovascular Medicine, University of Oxford, Oxford, UK

Workshop Organization

Workshop (co)-chair(s)

Sharib Ali Pal Halvorsen

Keynote Speakers

James East Peter Moutney Lena M. Hein

Reviewers

Reinke, Annika Papiez, Bartlomiej Khanal, Bishesh Daul, Christian Jha, Debesh IBME, BDI, Department of Engineering Science, University of Oxford, Ofxord, UK SimulaMet, Oslo, Norway Translational Gastroenterology Unit, John Radcli e Hospital, University of Oxford, Oxford, UK Odin Vision and University College London, London, UK German Cancer Research Center (DKFZ), University of Heidelberg, Heidelberg, Germany Sarikaya, Duygu Zhou, Felix Dmitrieva, Mariia Riegler, Michael Ghatwary, Noha Celik, Numan Gupta, Soumya

References

1. Sharib Ali , Felix Zhou, Barbara Braden, Adam Bailey, Suhui Yang, Guanju Cheng, Pengyi Zhang, Xiaoqiong Li, Maxime Kayser , Roger D. Soberanis-Mukul , Shadi Albarqouni, Xiaokang Wang, Chunqing Wang , Seiryo Watanabe, Ilkay Oksuz, Qingtian Ning, Shufan Yang, Mohammad Azam Khan, Xiaohong W. Gao, Stefano Realdon, Maxim Loshchenov, Julia A. Schnabel , James E. East, Georges Wagnieres, Victor B. Loschenov , Enrico Grisan, Christian Daul, Walter Blondel, and Jens Rittscher . An objective comparison of detection and segmentation algorithms for artefacts in clinical endoscopy . Scienti c Reports , 10 ( 1 ): 2748 , 2020 .

2. Sharib Ali , Mariia Dmitrieva, Noha Ghatwary, Sophia Bano, Gorkem Polat, Alptekin Temizel, Adrian Krenzer, Amar Hekalo, Yun Bo Guo, Bogdan Matuszewski, Mourad Gridach, Irina Voiculescu, Vishnusai Yoganand, Arnav Chavan, Aryan Raj, Nhan T. Nguyen, Dat Q. Tran , Le Duy Huynh, Nicolas Boutry, Shahadate Rezvy, Haijian Chen, Yoon Ho Choi, Anand Subramanian, Velmurugan Balasubramanian, Xiaohong W. Gao, Hongyu Hu, Yusheng Liao, Danail Stoyanov, Christian Daul, Stefano Realdon, Renato Cannizzaro, Dominique Lamarque, Terry Tran-Nguyen, Adam Bailey, Barbara Braden, James E. East , and Jens Rittscher . Deep learning for detection and segmentation of artefact and disease instances in gastrointestinal endoscopy . Medical Image Analysis , 70 : 102002 , 2021 .

3. Sharib Ali , Debesh Jha, Noha Ghatwary, Stefano Realdon, Renato Cannizzaro, Osama E. Salem, Dominique Lamarque, Christian Daul, Kim V. Anonsen , Michael A. Riegler , Pal Halvorsen, Jens Rittscher, Thomas de Lange, and James E. East . Polypgen: A multi-center polyp detection and segmentation dataset for generalisability assessment . arXiv, 2021 .