Serial fusion of random subspace ensemble for subcellular phenotype images classification

Abstract

Subcellular localisation is a key functional characteristic of proteins. In this paper, we apply Haralick texture analysis and Curvelet Transform for feature description and propose a cascade Random Subspace (RS) ensemble with rejection options for subcellular phenotype classification. Serial fusions of RS classifier ensembles much improve classification reliability. The rejection option is implemented by relating the consensus degree from majority voting to a confidence measure and abstaining to classify ambiguous samples if the consensus degree is lower than a threshold. Using the public 2D HeLa cell images, classification accuracy 93% is obtained with rejection rate 2.7% from the proposed system.