Abstracts

RATIONALE: MRI characteristics of focal cortical dysplasia (FCD) include cortical thickening, blurring of the gray matter (GM)-white matter (WM) junction, and hyperintense signal within the lesion. In many patients, FCD is characterized only by minor structural changes that often go unrecognized by standard radiological analysis. Computer-based texture analysis provides quantitative information about spatial gray level variations in MR images. To increase the sensitivity of MRI in the detection of subtle lesions of FCD, we developed a voxel-based image processing method modeled on in vivo features of FCD. METHODS: Patients: We selected 11 patients who had focal cortical resection for treatment of intractable extra-temporal partial epilepsy. All had histologically proven FCD. In 4 patients, FCD was seen on preoperative MRI, and in 7 the MRI was reported as normal. Image processing: Preoperative T1-weighted three-dimensional MRI was normalized to 128 gray levels, and segmented into GM and WM using a histogram-based method. Texture features were calculated over a neighborhood for each voxel in the MRI. Three texture maps were generated: 1. Gray matter thickness map (number of consecutive GM voxels) to model cortical thickening; 2. Gradient map (rate of change in gray level intensities) to model blurring of the GM-WM junction; 3. Relative intensity map (distance between intensity of a given voxel and the intensity at the boundary between GW and WM) to model hyperintense signal within the lesion. Finally, a ratio map (GM thickness x relative intensity/gradient) was generated. RESULTS: Lesions were seen on ratio maps in 11/11 patients, compared to 4/11 on the preoperative MRI. CONCLUSIONS: MRI texture analysis improves visual detection of FCD, even in cases where no lesion is obvious on MRI. This image processing technique could increase the number of patients with so-called non-lesional partial epilepsy who could benefit from surgery.