Authors | K. Pogorelov, O. Ostroukhova, A. Petlund, P. Halvorsen, T. de Lange, H. Espeland, T. Kupka, C. Griwodz and M. Riegler |
Title | Deep Learning and Handcrafted Feature Based Approaches for Automatic Detection of Angiectasia |
Afilliation | Communication Systems |
Project(s) | Efficient EONS: Execution of Large Workloads on Elastic Heterogeneous Resources |
Status | Published |
Publication Type | Proceedings, refereed |
Year of Publication | 2018 |
Conference Name | 2018 IEEE Conference on Biomedical and Health Informatics (BHI) |
Pagination | 365-368 |
Publisher | IEEE |
Keywords | Angiectasia, computer aided diagnosis, deep learning, Machine learning, video capsular endoscopy |
Abstract | Angiectasia, formerly called angiodysplasia, is one of the most frequent vascular lesions and often the cause of gastrointestinal bleedings. Medical specialists assessing videos or images of examinations reach a detection performance of 16% for the detection of bleeding to 69% for the detection of angiectasia. This shows that automatic detection to support medical experts can be useful. In this paper, we present several machine learning-based approaches for angiectasia detection in wireless video capsule endoscopy frames. In summary, the most promising results for pixel-wise localization and framewise detection are obtained by the proposed deep learning method using generative adversarial networks (GANs). Using this approach, we achieve a sensitivity of 88% and specificity of 99.9% for pixel-wise localization, and a sensitivity of 98% and a specificity of 100% for frame-wise detection. Thus, the results demonstrate the capability of using deep learning for automatic angiectasia detection in real clinical settings.
|
DOI | 10.1109/BHI.2018.8333444 |
Citation Key | 25794 |