Abstracts

Rationale: Identification of epileptogenic lesions plays an important role in presurgical evaluation of patients with epilepsy. Identification of focal cortical dysplasia (FCD) is a particular challenge, as subtle lesions can be overlooked. In these cases, quantitative analysis can aid in lesion identification, but often requires significant manual intervention or large training sets. We implemented a simple, automated atlas-free within subject similarity metric for visual lesion identification based on previous observations that subtle focal cortical dysplasias are often found to have increased cortical thickness. Methods: We obtained MRIs from 7 healthy volunteers (3 female, age 10-40) and 12 patients with focal epilepsy (8 females, ages 15-61), 10 with FCD (6/10 were MRI negative, positive on histology) and 2 had more diffuse malformations of cortical development on MRI. Data were acquired on NIH Clinical Center 3T Philips Achieva MRI Scanners, including 3D volumetric images (MPRAGE, T2, and FLAIR). Surface reconstruction and cortical thickness measurements were performed with FreeSurfer. Multiscale 3D local image features including Gaussian and gradient filters computed for three MR contrasts were computed for each gray-white surface node. Curvature and sulcal depth effects were regressed out of feature sets. A target feature vector was computed by averaging the feature vectors for surface nodes with cortical thickness 5mm or larger; cosine similarity to this target vector was computed for each cortical node; and the similarities were normalized across the surface. Lesions were identified by visualizing the normalized similarity score surface maps. Results: In healthy volunteers and patients, increased similarity scores were observed in the occipital pole, temporal pole, and insular regions, while increased dissimilarity scores were observed primarily in the precentral region. High similarity scores were also observed in abnormal cortical regions in 10/12 patients, 2 with larger malformations of cortical development and 6/10 with FCD.   2 additional FCD patients had lesional areas with high dissimilarity scores.  Use of thickness maps alone detected abnormalities in 5/12 patients.  In 2 cases with parietal dysplasias near to or within primary sensory cortex by histology, the lesions were not detected visually, or by using thickness or similarity score maps<./p>