Medical image/ video analysis for therapeutic ingestible microcapsule

• Main Author: Tan, Ping Chun.
• Other Authors: Chan Kap Luk
• Format: Final Year Project
• Language: English
• Published: 2010
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/40201
• Institution: Nanyang Technological University

This report presents a study of multi-level local feature classification for bleeding detection in Wireless Capsule Endoscopy (WCE) images using MATLAB. The image feature that is used in classification is color. There are 3 levels of classification: low-level, intermediate-level and high-level classification. In low-level classification, each cell of N by N pixels is characterized by adaptive color histogram which is used as feature representation for WCE images. A Neural Network (NN) cell-classifier is trained to classify cells in an image as bleeding or non-bleeding patches. In the intermediate-level classification, a block which covers 3 by 3 cells is formed. The intermediate-level representation of the block is generated from the low-level classifications of the cells, which captures the spatial local correlations of the cell classifications. Again, a NN block classifier is trained to classify the blocks as bleeding or non-bleeding ones. In high-level classification, the low-level and intermediate-level classifications are used by decision making rule to make a final decision. Experiments on clinical WCE videos have shown that the method of classification is not only accurate in both detection and differentiating bleeding or non-bleeding in WCE images.