We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
Classification of Helicobacter Pylori infection based on deep convolutional neural network with visual attention and self-supervised learning for endoscopic images.
- Authors
Jian, Guo-Zhang; Lin, Guo-Shiang; Wang, Chuin-Mu; Yan, Sheng-Lei
- Abstract
Helicobacter pylori (HP) is a major risk factor for chronic gastritis as well as gastroduenal disease and gastric endoscopic examination is the standard procedure to diagnose HP infection. A computer-aided diagnosis (CAD) scheme based on convolutional neural network (CNN) and self-supervised learning was proposed for classification of HP infection in this paper. The proposed scheme is composed of an encoder and a prediction head. The encoder composed of backbone network, visual attention module (VAM), and feature fusion module (FFM) is trained by using self-supervised contrastive learning for feature extraction. After obtaining the trained encoder, the whole network is trained by using a small image dataset with annotations. To evaluate the performance of the proposed scheme, some BLI (Blue Laser Imaging) images are collected for testing. According to experimental results of 5-fold cross validation, the average F1-score rates of the proposed scheme are from 0.885 to 0.915 for diagnosing HP infection. Therefore, the experimental results demonstrate that the proposed scheme can effectively extract visual representative features from BLI images for HP infection classification. Furthermore, the proposed scheme is superior to two existing methods in terms of accuracy and F1-score.
- Publication
Multimedia Tools & Applications, 2023, Vol 82, Issue 24, p37731
- ISSN
1380-7501
- Publication type
Article
- DOI
10.1007/s11042-023-14764-9