Modified residual attention network for abnormalities segmentation and detection in WCE images
Charfi, Said,Ansari, Mohamed El,Ellahyani, Ayoub,Eljaafari, Ilyas
DOI: https://doi.org/10.1007/s00500-023-09576-w
IF: 3.732
2024-01-15
Soft Computing
Abstract:Wireless capsule endoscopy (WCE) is a technology that uses a pill-sized camera to visualize images of the digestive tract. It presents several advantages, since it is far less invasive, does not require sedation and has less potential complications compared to standard endoscopy. Hence, it might be exploited as alternative to the standard procedure. WCE is used to diagnosis a variety of gastro-intestinal diseases such as polyps, ulcers, Crohn's disease and hemorrhages. Nevertheless, WCE videos can contain thousands of images per patient that must be screened by medical specialists, besides, the capsule free mobility and technological limits cause production of a low quality images. In this paper, a Nouvel method based on Dense-UNet deep learning segmentation model is presented. This approach aims at red lesion, ulcer and polyp detection from WCE images. Then, we propose a modified residual attention network for images classification. The proposed methods training and validation accuracies are 97.57% and 92.70%, with training and validation intersection over union and dice coefficients of 75.31%, 71.29%, 83.50% and 80.66%, respectively, in the red lesion dataset. In the polyp dataset, the method achieved a training and validation accuracies of 98.26% and 92.33%, with training and validation intersection over union and dice coefficients of 92.09%, 95.62%, 80.13% and 87.14%, respectively. Besides, the proposed architecture reached an average accuracy of 99.30%, sensitivity, specificity and F1 score of 98.61%, 100% and 99.30%, respectively. These results demonstrate that the proposed approach is satisfactory in polyp and red lesion segmentation and ulcer detection.
computer science, artificial intelligence, interdisciplinary applications