Segmentation of the subthalamic nucleus in MR images using information fusion - A preliminary study for a computed-aided surgery of Parkinson's disease

Abstract

Subthalamic nucleus (ST) stimulation is proved to have beneficial effects on the symptoms of Parkinson.s disease. We propose a fully automated segmentation method of this structure on MR images based on an information fusion technique. Information is provided by both images and expert knowledge, and consists in morphological, topological and tissue constitution data. All this ambiguous, complementary and redundant information is managed using a three steps fusion scheme based on fuzzy logic. Information is first modeled into a common theoretical frame managing imprecision and uncertainty. Models are then fused and a segmentation is finally performed that reduce the imprecision and increase the certainty in the location of the structure. Computed locations are compared with those obtained during a stereotactic surgical procedure. Results on ten patients are very promising, and suggest that this method may be applied during the surgical procedure as an help for the location of the subthalamic nucleus.