A combined approach of non-subsampled contourlet transform and convolutional neural network to detect gastrointestinal polyp

The abnormal growth of tissues that disarray the typical organization of cells is popularly known as polyps. The polyp on the gastrointestinal is a primary sign of gastrointestinal cancer. False diagnosis is extremely high using traditional diagnosis procedures that make the polyp diagnosis is a cru...

Full description

Saved in:
Bibliographic Details
Published in:Multimedia tools and applications 2022-03, Vol.81 (7), p.9949-9968
Main Authors: Hasan, Md. Mahmodul, Hossain, Muhammad Minoar, Mia, Shisir, Ahammad, Mohd. Sultan, Rahman, Mohammad Motiur
Format: Article
Language:English
Subjects:
Algorithms
Artificial neural networks
Colonoscopy
Computer Communication Networks
Computer Science
Data Structures and Information Theory
Diagnosis
Feature extraction
Image enhancement
Medical imaging
Multimedia Information Systems
Multiple criterion
Neural networks
Polyps
Real time
Special Purpose and Application-Based Systems
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The abnormal growth of tissues that disarray the typical organization of cells is popularly known as polyps. The polyp on the gastrointestinal is a primary sign of gastrointestinal cancer. False diagnosis is extremely high using traditional diagnosis procedures that make the polyp diagnosis is a crucial task in real-time colonoscopy. We have developed a polyp detection methodology using a combination of hand-crafted and automated feature extraction techniques. In this study, we have experimented with different convolutional neural network (CNN) architectures and hand-crafted feature extractors to select the best combination. The combined approach of the fine-tuned Xception model with non-subsampled contourlet transform (NSCT) performed significantly well. Besides, we have applied the multi-criteria frame selection technique for selecting the best images from colonoscopy videos. Afterward, the feature extractors have worked on enhanced patch images of selected frames. This study has also experimented with dimensionality reduction techniques to remove irrelevant features from the combined feature vector. We designed an algorithm to localize the polyp regions using the outcomes of patch images. The method did significantly well on several available public datasets. This work might be helpful for the endoscopist during real-time endoscopy to detect polyps.
ISSN:1380-7501
1573-7721
DOI:10.1007/s11042-022-12250-2