Relative localizing value of amygdalo-hippocampal MR biometry in temporal lobe epilepsy

The aims of the study were (i) to examine the localizing value of three MRI quantitative modalities (qMRI) currently used for the analysis of the hippocampus and amygdala in the context of pre-surgical screening and (ii) to propose a step-by-step protocol based on the sensitivity and performance of the different MR techniques.

Ninety-two adults with chronic mesial temporal lobe epilepsy (TLE) of which 28 underwent amygdalo-hippocampal resection, and 34 age-matched controls were included in the study. High-resolution qMRI was performed at 1.5 T, including a tilted T1-weighted 3D-dataset for volumetry and four-echoes T2 relaxometry (both for hippocampus and amygdala quantifications) and multi-voxel spectroscopy [NAA/(Cho+Cre)] (exclusively in the hippocampus). Individual qMRI data were compared with electroencephalography regarding the localization of the epileptogenic area, with the neuropathological data and with postoperative outcome. MRI pathology was defined based on 99% confidence ellipses. Ten controls were used to assess the quantitative MRI intra- and inter-observer variability for all variables.

Volumetric measurements revealed unilateral damage in 77% of the patients, T2-relaxometry in 64% and spectroscopy in 53%. Additional measurements of the amygdalae (T2-relaxometry) allowed us to localize pathology that coexists with that of the hippocampus in 34%, and isolated unilateral amygdala damage in 8% of patients. Volumetry and T2-relaxometry (not spectroscopy) were associated with postoperative outcome, but accurate predictive models were computed based on hippocampal measures only. At least at 1-year follow-up, volumetry predicts outcome correctly in 100% of the cases, whilst T2-relaxometry classified 96.4% (27/28) of these patients. All operated patients had hippocampal sclerosis.

Hippocampal structural damage is equivocally depicted by spectroscopy. For diagnostic and pre-operative evaluation, hippocampal volumetry and T2-relaxometry provide maximal accuracy. Amygdala quantifications are irrelevant in the pre-operative evaluation but may be useful for diagnostic purposes. Of the three qMRI modalities tested, T2-relaxometry provided the best balance between diagnosis accuracy and time-efficiency to lateralize a sclerotic lesion on the majority of the patients. Cases that remain undecided after T2-relaxometry may benefit from additional measurements based on hippocampal volumetry.