Proton magnetic resonance spectroscopy of mesial temporal sclerosis: Analysis of voxel shape and position to improve diagnostic accuracy.

Objective: To determine if change in voxel shape and position improves the lateralization of magnetic resonance (MR) spectroscopy to the side of the MR imaging and/or electroencephalography (EEG) abnormality in mesial temporal sclerosis.
Methods: In 10 patients with unilateral mesial temporal sclerosis and 5 controls, MR spectroscopy was performed. Long echo time single-voxel spectroscopy was obtained in the right and left hippocampus. First, the standard 2-cm x 2-cm x 2-cm cubic voxels were placed bilaterally. Then, 1-cm x 2-cm x 4-cm rectangular voxels were used. With this rectangular voxel shape, more of the hippocampus and less of the adjacent medial temporal lobe were included in the interrogation voxel. N-acetylaspartate-to-creatine (NAA/Cr) ratios and choline-to-creatine (Cho/Cr) ratios were obtained. The difference between the affected and unaffected sides in the patients was calculated and compared with the controls. Additionally, the mean difference and standard deviation were determined. Predictive values and Student t tests were also performed.
Results: In all 10 patients, using the NAA/Cr ratios, the rectangular voxel correctly identified the abnormal side. The cubic voxel only identified the abnormal side in 6 of the 10 patients, however. Additionally, the mean difference between the affected and unaffected sides was 0.30 for the rectangular voxel but only 0.003 for the cubic voxel; that is, there was no overall significant difference when using the cubic voxel (P = 0.007). No significant correlation between the affected and unaffected sides was found using the Cho/Cr ratios. No significant difference between the right and left sides was found in the controls.
Conclusion: Use of a 1-cm x 2-cm x 4-cm rectangular voxel places more of the hippocampus in the region of interrogation with less partial volume artifact from the adjacent brain parenchyma, which occurs with use of the standard 2-cm x 2-cm x 2-cm cubic voxel, resulting in improved correct lateralization of MR spectroscopy to the side of the MR imaging and/or EEG abnormality.