Abstracts

Rationale: Pharmacoresistant neocortical epilepsy is highly associated with focal cortical dysplasias, malformations of neuronal growth and morphology. Identification of these dysplasias on neuroimaging is crucial for the patients in order to undergo presurgical planning and resective surgery. Surgical outcome is significantly improved in patients that have identifiable lesions on imaging versus those who are "nonlesional" utilizing current imaging techniques (Bien et al Arch Neurol 2009). Methods: We apply a combination of voxel-based morphometry, surface-based morphometry and deformation-based morphometry on multimodal imaging (T₁-weighted MPRAGE sequence, FLAIR, FDG-PET) from ECoG-implanted patients to automatically identify regions with cortical malformations. Five patients images (T₁, T₂, PET) were mapped into a common template domain and then histogram matched to equalize intensities. Grey matter, white matter segmentation maps along with FLAIR and PET images were mapped into the same image domain. These multimodal intensities were then converted into z scores through voxel-by-voxel comparison with 24 normal healthy controls with the same image acquisition parameters. The z-score maps were then smoothed to reveal areas of high and low differences. Predictions from these z maps were compared to regions that were surgically resected and confirmed by histopathology. Results: Brain morphometry was applied to neuroimaging data from 5 patients. One patient was revealed to have a small left Frontal Type II cortical dysplasia that was consistent with the localization of surgical resection. The tool also identified another possible lesion that was not identified in the radiology reading (see figure). In another patient, the tool was able to identify a lesion in the left parietal occipital region that was consistent with either a closed lip schizencephaly or cortical dysplasia. The tool was less successful identifying small scattered lesions that were visible on FLAIR images but not on T₁ weighted images. Conclusions: We introduce a tool that utilizes a combination of methods of brain morphometry to identify lesions suspected to be cortical dysplasias. Preliminary results show promise in assisting conventional neuroradiology analysis to improve diagnostic yield of MRI studies in patients with neocortical epilepsy.