Biologically inspired classification of microvessel histopathology via sparse coding

Recently the sparse coding approaches have been successfully applied to solve the image classification problems. However, the classification of microvessel regions from histopathology still rely on the hand designed low-level features. In this paper, we propose a novel method to classify region of microvessel by applying sparse coding on biological signals. The Single- and Double-Opponent signals from human visual cortex are simulated to capture microvessel properties. The SIFT (Scale Invariant Feature Transform)) descriptors of these signals are encoded via sparse coding and classified by SVM (Support Vector Machine) with the linear spatial pyramid matching kernel. We have carried out extensive experiments on the classification of microvessel histopathology and the proposed method achieves satisfactory classification rates.