Neural network based approaches for the classification of colonoscopic images

Abstract
A new method of colon status classification based on a set of quantitative parameters extracted from colonoscopic images is proposed. This can assist endoscopists for the early detection of abnormalities in the colon. Images captured by colonoscopic procedure are subjected to subsequent processing and analysis for the extraction of quantitative parameters, which form the input vectors to the three different neural networks selected for classification of colon. The three networks, viz. a two-layer perceptron trained with delta rule, a multilayer perceptron with backpropagation learning and a self-organising network, are used and the results obtained by the proposed methods are satisfactory. A comparative study of the three methods is also performed and it is observed that the self-organising network is more appropriate for the classification of colon status.

This publication has 1 reference indexed in Scilit: