Automatic diagnosis of prostate cancer using multispectral based linear binary pattern bagged codebooks

Abstract

Cancer is classified by the World Health Organisation as a worldwide problem and its effect is increasing. A timely diagnosis is crucial and an early detection can be vital leading to an effective diagnosis. This paper proposes an automated classification system of prostate cancer using multispectral imagery for an early detection. It revolves around a block based texture analysis that uses multiscale multispectral local binary pattern texture features combined with a bagging ensemble method and codebooks. Extensive experiments have been carried out using a real dataset and the result obtained show an accuracy of 96.0%. The findings were also analysed and compared against a few existing and similar techniques and the results suggest that the proposed approach is attractive.