Gastrointestinal Disorder Detection using Endoscopy Images with RNN compared over SVM

Abstract

This research investigates the application of Recurrent Neural Networks (RNNs) for automated detection of gastrointestinal (GI) disorders from endoscopic images, comparing its performance against Support Vector Machines (SVMs). The objective is to develop a robust and accurate computer-aided diagnosis (CAD) system. The study uses a large dataset of endoscopic images, pre-processed and augmented to improve model training. Results demonstrate that RNNs, leveraging the temporal nature of endoscopic image sequences, achieve superior accuracy and sensitivity compared to SVMs in identifying various GI disorders. This system holds significant potential for improving the efficiency and accuracy of GI disease diagnosis.

Introduction

Early and accurate diagnosis of gastrointestinal disorders is crucial for effective treatment and improved patient outcomes. Traditional methods rely heavily on the expertise of gastroenterologists, which can be subjective and time-consuming. Endoscopy, while providing detailed visual information, requires significant expertise for interpretation. Computer-aided diagnosis (CAD) systems offer a promising solution by automating the analysis of endoscopic images. However, existing CAD systems often struggle with the variability in image quality, subtle visual cues, and the need to consider the temporal context of endoscopic procedures. This research addresses these challenges by employing RNNs, specifically designed to handle sequential data, and comparing them to the widely-used SVM classifier for improved accuracy and efficiency.

Objectives

• To develop an RNN-based CAD system for detecting GI disorders from endoscopic images.
• To compare the performance of the RNN-based system with an SVM-based system.
• To evaluate the accuracy, sensitivity, and specificity of both systems.