Research Article
Identification of Polyp from Colonoscopy Images by Deep Belief Network based Polyp Detector Integration Model
@ARTICLE{10.4108/eetpht.9.3964, author={A. B. Dash and S. Dash and S. Padhy and R. K. Das and B. Mishra and B. K. Paikaray}, title={Identification of Polyp from Colonoscopy Images by Deep Belief Network based Polyp Detector Integration Model}, journal={EAI Endorsed Transactions on Pervasive Health and Technology}, volume={9}, number={1}, publisher={EAI}, journal_a={PHAT}, year={2023}, month={9}, keywords={DBN, Deep Residual Network, Resnet-50, Polyp Detector, Two Stage Model, Polyp Network}, doi={10.4108/eetpht.9.3964} }
- A. B. Dash
S. Dash
S. Padhy
R. K. Das
B. Mishra
B. K. Paikaray
Year: 2023
Identification of Polyp from Colonoscopy Images by Deep Belief Network based Polyp Detector Integration Model
PHAT
EAI
DOI: 10.4108/eetpht.9.3964
Abstract
Cancer is a disease involving unusual cell growth likely to spread to other parts of the body. According to WHO 2020 report, colorectal malignancy is the globally accepted second leading cause of cancer related deaths. Colorectal malignancy arises when malignant cells often called polyp, grow inside the tissues of the colon or rectum of the large intestine. Colonoscopy, CT scan, Histopathological analysis are some manual approaches of malignancy detection that are time consuming and lead to diagnostic errors. Supervised CNN data model requires a large number of labeled training samples to learn parameters from images. In this study we propose an expert system that can detect the colorectal malignancy and identify the exact polyp area from complex images. In this approach an unsupervised Deep Belief Network (DBN) is applied for effective feature extraction and classification of images. The classified image output of DBN is utilized by Polyp Detector. Residual network and feature extractor components of Polyp Detector helps polyp inspector in pixel wise learning. Two stage polyp network (PLPNet) is a R-CNN architecture with two stage advantage. The first stage is the extension of R-CNN to detect the polyp lesion area through a location box also called Polyp Inspector. Second Stage performs polyp segmentation. Polyp Inspector transfers the learned semantics to the polyp segmentation stage. It helps to enhance the ability to detect polyp with improved accuracy and guide the learning process. Skip schemes enrich the feature scale. Publicly available CVC-Clinical DB and CVC Colon DB datasets are used for experiment purposes to achieve a better prediction capability for clinical practices.
Copyright © A. Dash et al., licensed to EAI. This is an open access article distributed under the terms of the CC BY-NC-SA 4.0, which permits copying, redistributing, remixing, transformation, and building upon the material in any medium so long as the original work is properly cited.