SURF-SVM Based Identification and Classification of Gastrointestinal Diseases in Wireless Capsule Endoscopy

TL;DR

This paper presents a SURF-SVM based method for automatic detection and classification of gastrointestinal diseases in wireless capsule endoscopy images, achieving high accuracy and improving multi-class classification over previous bi-class approaches.

Contribution

It introduces a novel SURF feature extraction combined with SVM classification for multi-class GIT disease detection in WCE images, enhancing previous bi-class methods.

Findings

94.58% accuracy in normal/abnormal classification

82.91% accuracy in multi-class disease classification

Improved over previous bi-class analysis methods

Abstract

Endoscopy provides a major contribution to the diagnosis of the Gastrointestinal Tract (GIT) diseases. With Colon Endoscopy having its certain limitations, Wireless Capsule Endoscopy is gradually taking over it in the terms of ease and efficiency. WCE is performed with a miniature optical endoscope which is swallowed by the patient and transmits colour images wirelessly during its journey through the GIT, inside the body of the patient. These images are used to implement an effective and computationally efficient approach which aims to detect the abnormal and normal tissues in the GIT automatically, and thus helps in reducing the manual work of the reviewers. The algorithm further aims to classify the diseased tissues into various GIT diseases that are commonly known to be affecting the tract. In this manuscript, the descriptor used for the detection of the interest points is Speeded Up Robust Features (SURF), which uses the colour information contained in the images which is converted to CIELAB space colours for better identification. The features extracted at the interest points are then used to train and test a Support Vector Machine (SVM), so that it automatically classifies the images into normal or abnormal and further detects the specific abnormalities. SVM, along with a few parameters, gives a very high accuracy of 94.58% while classifying normal and abnormal images and an accuracy of 82.91% while classifying into multi-class. The present work is an improvement on the previously reported analyses which were only limited to the bi-class classification using this approach.