Summary form only given. Neurosurgery is a very complex procedure and the surgeon has to integrate multi-modal data to produce an optimum surgical plan. Many times the region of interest is surrounded by vital structures (such as motor cortex, temporal cortex, vision and audio sensors, etc.) and has irregular configurations. Slight damage to such eloquent brain structures can severely impair the patient. CASMIL, an image guided neurosurgery toolkit is being developed at Wayne State University to produce optimum plans resulting in minimally invasive surgeries. This system has many innovative features needed by neurosurgeons that are not available in other academic and commercial systems. CASMIL is an integration of various vital modules, such as rigid and non-rigid co-registration (image-image, image-atlas, and image-patient), 3D segmentation, brain shift predictor (BSP), knowledge based query tools, intelligent planning, and augmented reality. The above modules, though necessary for improved surgical outcomes, require tremendous amount of computing typically requiring a large cluster. In this paper, the author identifies the opportunities of HPC in computer assisted surgery (CAS) and the challenges specific to CAS