Skip to main content

Research Repository

Advanced Search

SegCrop: Segmentation-based dynamic cropping of endoscopic videos to address label leakage in surgical tool detection

Qayyum, Adnan; Bilal, Muhammad; Qadir, Junaid; Caputo, Massimo; Vohra, Hunaid; Akinosho, Taofeek; Berrou, Ilhem; Niyi-Odumosu, Faatihah; Loizou, Michael; Ajayi, Anuoluwapo; Abioye, Sofiat

Authors

Adnan Qayyum

Muhammad Bilal Muhammad.Bilal@uwe.ac.uk
Associate Professor - Big Data Application

Junaid Qadir

Massimo Caputo

Hunaid Vohra

Profile Image

Taofeek Akinosho Taofeek.Akinosho@uwe.ac.uk
Research Associate - Big Data Application Development

Profile Image

Ilhem Berrou Ilhem.Berrou@uwe.ac.uk
Senior Lecturer in Applied Pharmacology

Profile Image

Dr Michael Loizou Michael2.Loizou@uwe.ac.uk
Wallscourt Associate Professor in Health Technology and Life Sciences

Anuoluwapo Ajayi Anuoluwapo.Ajayi@uwe.ac.uk
Associate Professor - Big Data Application



Abstract

In recent times, surgical data science has emerged as an important research discipline in interventional healthcare. There are many potential applications for analysing endoscopic surgical videos using machine learning (ML) techniques such as surgical tool classification, action recognition, and tissue segmentation. However, the efficacy of ML algorithms to learn robust features drastically deteriorates when models are trained on noise-affected data [1]. Appropriate data preprocessing for endoscopic videos is thus crucial to ensure robust ML training. To this end, we demonstrate the presence of label leakage when surgical tool classification is performed naively and present SegCrop, a dynamic U-Net model with an integrated attention mechanism to dynamically crop the arbitrary field of view (FoV) in endoscopic surgical videos to remove spurious label-related information from the data. In addition, we leverage explainability techniques to demonstrate how the presence of spurious correlations influences the model's learning capability.

Citation

Qayyum, A., Bilal, M., Qadir, J., Caputo, M., Vohra, H., Akinosho, T., …Abioye, S. (2023, April). SegCrop: Segmentation-based dynamic cropping of endoscopic videos to address label leakage in surgical tool detection. Paper presented at IEEE International Symposium on Biomedical Imaging (ISBI), 2023, Colombia

Presentation Conference Type Conference Paper (unpublished)
Conference Name IEEE International Symposium on Biomedical Imaging (ISBI), 2023
Conference Location Colombia
Start Date Apr 18, 2023
End Date Apr 21, 2023
Deposit Date Mar 13, 2023
Publicly Available Date Mar 29, 2024
DOI https://doi.org/10.1109/isbi53787.2023.10230822
Public URL https://uwe-repository.worktribe.com/output/10461127
Related Public URLs https://2023.biomedicalimaging.org/en/