Automated segmentation of spinal diffusion tensor MR imaging

Abstract

A novel automated segmentation technique is presented for the delineation of white matter and gray matter regions in diffusion tensor magnetic resonance imaging of the spine. The technique involves an automated method for the extraction of the spinal cord regions from the diffusion tensor imaging data and relies on the fuzzy means clustering approach, which is inherently robust. Experimental results obtained for the segmentation of in vitro spinal cord sections of varying ages from 48 to 80 years demonstrate the viability of the automated segmentation technique. Statistical comparison with manually delineated white matter regions indicates the potential of the automated technique for the investigation and analysis of white matter abnormalities in diffusion tensor magnetic resonance imaging of the spine.