Routine colon cancer detection using local image descriptors

In the present era of information technology, medical imaging and analysis is very active area of research, because of its use in various disease diagnosis. Colon cancer is among one of the diseases to be identified effectively and accurately. In this paper various state of the image descriptors are investigated for their suitability in computer assisted colon cancer detection. Five recent local descriptors namely, LCOD, RSHD, LWP, LDEP and LBDP have been considered for the same. Image retrieval framework has been used to test the efficacy of the descriptors and tested on routine colon cancer H&E stained images. Descriptors are evaluated on the basis of precision, recall, retrieval time and feature length. On the basis of evaluation parameter comparisons, proposition of using LCOD and RSHD descriptors have been presented in this paper.