Automatic analysis of normal brain dissymmetry of males and females in MR images

Abstract

In this paper, we present a statistical analysis of the normal dissymmetry in brain MR images. Our method consists in: computing a dense 3D field characterizing the dissymmetry of a subject, analysing it with an operator (relative volume variation), computing 3D statistical maps from populations of subjects (learning series) and extracting interesting regions. The results are then tested on an independent validation series. We use a database of 36 normal young subjects, 18 males and 18 females, strongly right-handed. We show that the brain hemispheres can be reliably discriminated thanks to a consistent size difference in the white matter of the temporal and frontal lobes, with a very high level of confidence (36/36, α-value=0. 000001). We also exhibit a difference of brain dissymmetry with respect to sex.