Extraction of epileptogenic foci from PET and SPECT images by fuzzy modeling and data fusion.

During the presurgical evaluation of medically intractable epilepsy, isotopic functional imagery provides an increasing amount of data concerning the potential location of the focus. The aim of this study is to facilitate the surgical decision by presenting an image fusion method able to extract epileptogenic foci from periictal single photon emission computed tomography (SPECT), interictal SPECT, fluoro-desoxy-glucose (FDG) position emission tomography (PET), and flumazenil PET. After spatial coregistration, the images are converted into fuzzy maps whose membership functions indicate the pathological degree of each voxel, according to each modality. These maps are then fused together thanks to a combination operator managing uncertainty (due to the sensitivity) and imprecision (due to poor resolution and partial volume effect) of the images. In the framework of possibilistic theory, this operator mimics the way the physicians evaluate and compare the various exams. The technique was successfully tested on simulated images with well-defined abnormalities, in terms of size and intensity. A preliminary clinical study was also performed and gave results in accordance with the "gold standard" investigation (deep electrodes or postsurgical outcome) in 11 patients out of 12.