SEMI-AUTOMATED CORTICAL TUBER IDENTIFICATION IN TUBEROUS SCLEROSIS: HIGH DEFINITION IMAGING AT 3 TESLA WITH SURFACE COILS
Abstract number :
1.210
Submission category :
Year :
2003
Submission ID :
2226
Source :
www.aesnet.org
Presentation date :
12/6/2003 12:00:00 AM
Published date :
Dec 1, 2003, 06:00 AM
Authors :
Colin P. Doherty, P. Ellen Grant, Bradford C. Dickerson, Evalina Busa, Elizabeth A. Thiele Pediatric Epilepsy Service, Department of Neurology, Massachusetts General Hospital and Harvard Medical School, Boston, MA; MGH/MIT/HMS Athinoula A. Martinos Center
Cortical tuber counting in tuberous sclerosis complex (TSC) has been suggested as a biomarker of disease burden. Most techniques employ simple manual counting methods using high resolution T2-weighted fast spin echo (FSE) or fluid attenuated inversion recovery (FLAIR) sequences on conventional 1.5 Tesla imaging machines. We chose to employ higher field strengths and whole head phased array surface coils for better image resolution and applied cortical reconstruction techniques in post-processing for semi-automated identification of cortical tubers.
8 adults with a diagnosis of TSC were imaged using a 3 Tesla Siemen[apos]s [apos]Trio[apos] whole body magnetic resonance scanner together with locally designed phased array surface coils. Detailed T1-weighted FSE scans and two 3D sagittal MPRAGE scans were collected along with coronal and axial FLAIR images. The two 3D acquisitions were averaged in the first post-processing step to further reduce signal to noise ambiguities. Novel image processing software (FREESURFER) was then employed for brain inflation and cortical reconstruction of the MPRAGE scans which resulted in cortical thickness measures. Cortical tubers were automatically identified using a statistical thickness map and this was compared to manual counting on the FLAIR images using the same rater in a blinded technique
Detailed T1-weighted images at 3 Tesla using the phased array coils visualized cortical tubers at a resolution unmatched by even pathological samples. Using a threshold significance level of 0.5 for cortex [gt] 0.4 cm in thickness there was a high correlation (r = 0.92) between the number of tubers identified using the semi-automated and manual techniques. Results further correlated with detailed clinical, neurophysiologic, molecular and neuropsychologic data.
We suggest that such enhanced techniques for tuber identification will refine the current specificity of cortical tuber count for disease status, surveillance and prognosis. The ability to so clearly distinguish dysplastic from normal tissue argues for wider use of 3T imaging and phased array coils for the study of the natural history of TSC.
[Supported by: Tuberous Sclerosis Alliance Innovative Science Research Award to Elizabeth A. Thiele MD PhD ]