Analysis on the morphological variation of brainstem in Alzheimer MR images using inverse Perona Malik diffusion filter and level set method

Alzheimer’s disease is a neurological disorder which undergoes a prodromal stage before the occurrence of cognitive decline, during which the changes in non-cognitive functions plays a significant role in the deterioration of cognitive abilities. Non-cognitive functional defects are due to dysfunction of the brain stem. In this work, the brainstem is segmented using the level set without re-initialization method along with inverse PM as edge stopping criterion. The segmented results are validated against the ground truth using statistical and overlap measures. Structural variations of the segmented brain stem are studied using geometric features. Results show that the ability of inverse Perona-Malik diffusion in detecting the weak edges improves the segmentation results of the variational level set method when compared with Gaussian and Perona-Malik based edge detection techniques. The segmented images are validated using regional statistics, overlap measures, and geometric features. It is also observed that the segmented brainstem using inverse PM has a high degree of similarity and sensitivity than other comparative methods. Extracted geometric features from segmented brainstem are able to differentiate the demented subjects from normal. Disease progression is better inflicted by the atrophy of brain stem. Hence the analysis seemed to be clinically significant.