A large number of sophisticated techniques have been proposed over the last few decades for automatic analysis of brain MR images to help clinicians better diagnose and understand anatomical changes due to neurological disorders. While significant improvements in performance have been achieved, almost all techniques suffer from a common limitation of high computational complexity due to the large number of voxels present in a typical MR volume. Computational complexity is a major hurdle in the clinical application of these sophisticated image analysis techniques. Brain MR volumes consist of approximately piecewise constant tissue regions with high redundancy among voxel intensities, which can be grouped into perceptually meaningful entities (superpixels) to reduce the complexity. In this study, we investigate the utility of superpixels (2D) and supervoxels (3D) in reducing computational complexity of brain MR analysis tasks. We investigate the extent of spatial and intensity distortions introduced in superpixel representation of MR images and evaluate its effect on brain tissue segmentation as an example task. We observe that superpixels are highly promising for significantly reducing the computational complexity of the lower-level image analysis tasks that are often essential components of MR analysis pipelines.