Segmentation of the hippocampus for detection of Alzheimer's disease

Abstract

Since hippocampal volume measurement is often used in detection and progression of Alzheimer's disease (AD), segmentation of hippocampus is a significant clinical application. However, it is relatively hard task, due to low signal to noise ratio (SNR), low contrast, indistinct boundary and intensity inhomogeneities. This paper uses Wave Atom shrinkageas an efficient method for enhancing the noisy magnetic resonance images to improve segmentation accuracy followed by a region-scalable active contour model that uses intensity information in local regions. A data fitting energy functional is incorporated into a level set formulation, from which a curve evolution equation is derived for energy minimization. Experimental results of segmenting the hippocampus in T1-weighted MR images yield promising results in the presence of intensity inhomogeneities and low SNR images. © 2012 Springer-Verlag.