Abstracts

Rationale: Diffusion tensor imaging (DTI) is highly senstive to alterations in the architecture of tissue at the microstructural level and may add to hippocampal volume (HCV) loss in predicting verbal memory performance and/or seizure lateralization in mesial temporal lobe epilepsy (MTLE). Methods: DTI and volumetric MRI were performed in 39 patients with video-EEG confirmed unilateral MTLE (20 left; 19 right). Apparent diffusion coefficient (ADC) of the hippocampus (HC), parahippocampal cingulum (PHC) and fornix (FORX), and HCV were calculated for all subjects using DTI and MRI, respectively. Verbal memory was evaluated with Logical Memory-2 of the Wechsler Memory Scale-Third Edition. Discriminant function analyses (DFA) were performed to examine whether a linear combination of DTI and volumetric measurements enhanced the prediction of seizure lateralization, and hierarchical regression analyses were performed to determine their aggregate contributions to memory, after taking HV into account. Results: For verbal memory, left HCV alone explained 22% of the variance; However, left HC-ADC explained an additional 21% of the variance, resulting in 43% total explained variance. PHC ADC correlated with verbal memory performance, but did not contribute additional variance after accounting for HCV and H-ADC. FORX ADC was not related to memory performance. With respect to seizure lateralization, HCV asymmetry alone correctly classified 84% of the original sample (84% with cross-validation; CV). Using a linear combination of HCV, HC-ADC, FORX-ADC, and PHC-ADC asymmetry values, classification accuracy increased to 92% (89% with CV). Although this linear classifier did not provide a statistically significant improvement in classification, the addition of DTI measures correctly classified 3 of the 6 patients who were missclassfied on the basis of HCV asymmetry alone and who did not show evidence of hippocampal sclerosis on MRI. Conclusions: Our results suggest that combining microstructural and macrostructural measures of HC pathology enhances the prediction of verbal memory performances in MTLE. However, the utility of DTI measures for improving seizure lateralization warrants further investigation in a large sample of patients without obvious HCV loss. HC-ADC has been shown to predict memory decline in patients with mild cognitive impairment before the onset of HCV loss, and it may be particulary sensitive for explaining memory loss in patients new-onset MTLE who likely have changes in HC tissue properties prior to visible HCV loss.