Localization of Target Structures through Data Fusion Applied to Neurostimulation

Symptoms of Parkinson's disease can be relieved by deep brain stimulation (DBS), which is based on an accurate positioning of electrodes in specific structures of the basal ganglia, e.g. the subthalamic nucleus (STN) and the internal globus pallidus (GPi), and of the thalamus. We planned to develop an automatized method for the identification of these structures, deeply located in the brain, through data fusion. Information, provided by both anatomical MR images and expert knowledge, was managed in a possibilist frame by a fuzzy logic approach. A multi-scale graph-based virtual atlas, modeling the theoretical anatomical knowledge, was matched to each patient's data (MRI), in order to compute an estimation of the structures' location, thus assisting the physician in the definition of the target. The method has been tested on several images with promising results. In the long run, it might open opportunities to a fully automatized surgical planning