Abstracts

Can automated voxel-wise analysis of MRI help the detection of hippocampal sclerosis?

Abstract number : B.03;
Submission category : 5. Human Imaging
Year : 2007
Submission ID : 8124
Source : www.aesnet.org
Presentation date : 11/30/2007 12:00:00 AM
Published date : Nov 29, 2007, 06:00 AM

Authors :
L. Bonilha1, M. Eckert4, Z. Rumboldt2, C. Rorden3, D. R. Roberts2, J. J. Halford1

Rationale: In some patients with medial temporal lobe epilepsy (MTLE), hippocampal atrophy (HA) is not obvious in routine MRI. In this setting, manual morphometry of the hippocampus can aid the detection of HA by yielding a quantifiable measure of hippocampal volume, which is then compared to the hippocampal volume from a healthy population. Unfortunately, manual tracing is a technique that is difficult to implement in clinical practice because it is time-consuming and rater-dependent. High-resolution MRI segmentation algorithms such as voxel-based morphometry (VBM) may make it possible to measure HA using an automated process. In this study, we aimed to investigate if VBM can consistently identify HA in patients with MTLE with visually defined HA.Methods: Twenty consecutive patients with clinically defined MTLE with HA visually identified in routine clinical MRI were selected for this study. Thirty-four age-matched healthy individuals were also enrolled in the study. All subjects underwent high-resolution MRI in a Philips 3T scanner equipped with a multi-element head coil yielding T1-weighted images with 1mm isotropic voxels. Images underwent spatial normalization, segmentation of gray matter, and spatial smoothing using the software SPM5. Smoothed gray matter maps from controls were used to generate voxel-wise mean and standard deviation maps. Individual voxel-wise Z-score maps relative to the controls were generated for all patients. Each patient's hippocampal Z-score maps were quantified and compared to the clinical visual assessment for HA. We also assessed hippocampal Z-score maps from ten randomly selected control subjects as a way to investigate false positives. Results: In all patients with HA (left side, n=13, right side n=7), Z-score maps of the atrophied hippocampus showed regions with Z smaller than -2 standard deviations (SD), ranging from -2 to -4 SD. Additionally, Z-score maps of the contralateral hippocampus ranged from -0.5 to -3 SD, suggesting contralateral reduction of gray matter. In contrast, none of the control subjects showed hippocampal z-scores below -2 SD.Conclusions: Our results suggest that the use of VBM and standardized maps of gray matter can potentially aid in the detection of HA, as atrophied hippocampi tend to fall underneath a critical threshold of SD. This might prove to be beneficial clinical tool to help detect mild or bilateral HA. Further study of larger control groups, with smaller confidence intervals, may improve the sensitivity and specificity of this automated technique and allow its routine clinical use.
Neuroimaging