3-D volume of interest based image classification

Abstract

This paper proposes a number of techniques for 3-D image classification according to the nature of a particular Volume of Interest (VOI) that appears across a given image set. Three VOI Based Image Classification (VOIBIC) approaches are considered: (i) Statistical metric based, (ii) Point series based and (iii) Tree based. For evaluation purpose, two 3-D MRI brain scan datasets, Epilepsy and Musicians, were used; the aim being to distinguish between: (i) epilepsy patients versus healthy people and (ii) musicians versus non-musicians. The paper also considers augmenting the VOI data with meta data. According to the reported experimental results the Point series based approach, augmented with meta data, is the most effective.