Voxel-based analysis of whole brain FLAIR at 3T detects focal cortical dysplasia
Abstract number :
2.083;
Submission category :
5. Human Imaging
Year :
2007
Submission ID :
7532
Source :
www.aesnet.org
Presentation date :
11/30/2007 12:00:00 AM
Published date :
Nov 29, 2007, 06:00 AM
Authors :
N. Focke1, 2, M. R. Symms1, R. Samson1, J. S. Duncan1
Rationale: Focal cortical dysplasia (FCD) is an important cause for pharmaco-resistant focal epilepsy. The most evident features on clinical MRI are an increased T2 signal in the involved cortex and a tail extending into the white matter. These features are most conspicuous on a CSF suppressed T2 weighted scan (FLAIR). The purpose of this study was to determine whether voxel-based analysis could reliably detect abnormalities of FLAIR intensity in cases of neuroradiologically demonstrated FCD. If so, this would be promising tool for the investigation of MRI negative patients. Methods: We studied 25 patients in whom FCD was reported by two experienced neuroradiologists on 3T (GE Excite) MRI and 25 normal controls. All had a high-resolution T1 3D-volume (hr3D) and a standard coronal T2-FLAIR scan (cFLAIR). Each hr3D scan was coregistered, segmented into tissue classes and normalised to 1x1x1mm resolution using the unified segmentation model of SPM5. The cFLAIR scans were spatially normalised using the same parameters and intensity normalised by extracting the anterior frontal white matter and the pons via an anatomical mask in combination with the individual white matter class and setting the average intensity to 1000. Afterwards the cFLAIR scans were brain extracted using bet (FSL) and convolved with a Gaussian kernel (8mm FMHW). An SPM5 GLM analysis was performed using a two-sample t-test to compare the signal increases in each individual cFLAIR scan against the controls (one-sided t contrast, p=0.0001 uncorrected and no extent threshold). The contrast images were then overlaid on the individual normalised hr3D scans for visualisation.Results: In 22/25 cases (88%) the FCD was correctly identified. In 2 of these the FCD was identified but was not the most significant finding in the whole brain. In 3/25 (12%) cases the lesion was not found at the given threshold level. In controls there was one false positive finding (cluster size < 20 voxel) when comparing each of the controls against the rest of the control group. The highest T scores were noted in the subcortex in the majority of cases but clusters did reach into cortex as well (see images). Clearly vascular lesions (mostly periventricular) were ignored since they are very frequent in the general population. At a threshold of p=0.0005 2 of the missed findings were detected as well but at the cost of more false positive findings in controls (6 findings with clusters > 5 voxel). Conclusions: FLAIR intensity mapping detected most cases of focal cortical dysplasia in our group of patients. It is of note that our data set included all cases for which hr3D and cFLAIR data was available at 3T and included very subtle lesions. It appears promising for other pathologies and / or MRI negative patients as well since it is a lot less susceptible to partial volume problems given its excellent inherent CSF suppression.
Neuroimaging