Abstracts

Rationale: Although for many years extensive animal data have suggested an important role for a network involving the thalamus in hippocampal epilepsy, this issue has only recently been examined in humans. Our recent work has demonstrated that in patients with temporal lobe epilepsy a macroscopic network of energetic impairment and neuronal function exists involving the hippocampus, bilateral thalami and bilateral basal ganglia. This network is reflected by a hierarchy of correlated decrements in N-acetyl aspartate, a surrogate measure of neuronal mitochondrial function. The goal of this work was to determine if measurements of NAA in the network provides predictive power for temporal lobectomy.Methods: 1H MR spectroscopic images of the hippocampus and thalamus were acquired from 25 patients prior to temporal lobe resection using a 4T whole body MR system. Spectroscopic images were acquired with a nominal resolution of 0.64cc using a quadrature TEM volume head coil. Voxels from both hippocampi (4 each), thalami (6 each) and basal ganglia (3 each) were reconstructed using an automated image guided analysis routine. To minimize the number of variables (8) anterior and posterior loci for each hippocampus and thalamus were averaged. Patient outcomes (more than 1 year post surgery) were divided into two groups, those attaining an ILAE classification of I or II and those continuing to experience seizures (Class III and higher). To classify the data a neural network using a single layer of perceptron neurons was used with 8 inputs (anterior and posterior hippocampus and thalamus) for each subject.Results: Displayed in Figure 1 are the correlation coefficients between the ipsilateral and contralateral hippocampi and the other components of the network in patients with temporal lobe epilepsy. Of the 25 patients, 21 achieved a class I or II outcome, while 4 continued to experience seizures (Class III and above). The network correctly classified (23/25) patients or 92% of the group. The two misclassified patients included one patient who was predicted to have a class I or II outcome who experienced breakthrough seizures and one patient who was predicted to have a Class III outcome who achieved a seizure free outcome.Conclusions: Temporal lobe epilepsy is characterized by a hierarchy of reductions in NAA in the hippocampi, thalami and basal ganglia. Using the characteristics of this network as an input to a neural network allowed 92% of the outcomes to be correctly predicted.