Abstracts

Conventional and novel MRI techniques can detect cerebral abnormalities in patients with refractory focal epilepsies. Correlation of preoperative MRI data and neuropathological analysis of the neocortex is not straightforward. Per-operative neuronavigation and placement of markers on tissue is of limited use in temporal lobe resections. MRI scanning of the resected specimen for registration with in-vivo MRI is complicated by anisotropic deformation of tissue after resection. We have developed a method to facilitate precise registration of preoperative MRI with the resected specimen and enable accurate correlation of MRI findings with histopathology. Ten temporal lobe resections undertaken for refractory temporal lobe epilepsy were studied. [italic]En bloc[/italic] neocortical resections were performed followed by amygdalo-hippocampectomy. The middle temporal gyrus was marked with ink in the operating room, and the orientation of the specimen noted. The specimen was fixed in formalin for a week and then cut coronally using a specially manufactured cradle with parallel blades at 5 mm intervals to ensure evenly thick slices in the same orientation. The posterior face of each tissue block was photographed. Volumetric T1-weighted preoperative MRI were reformatted and sliced coronally in the same orientation as the fixed lobe. Consecutive MRI slices (0.94 mm) were compared to photographs of the 5 mm thick tissue blocks by two observers (SHE and SLF) separately and then together for a consensus. In eight cases hippocampal sclerosis seen on MRI was confirmed. There were 4-6 slices of temporal neocortex per case. In eight cases, one or more tissue block could be confidently matched with MRI slices, from which correlation of the remaining slices could be estimated. In two cases finding corresponding slices were more difficult, but there were one or two probable matches, from which the remaining correlations could be estimated. Matching was usually easiest 1.5-2 cm posterior from the temporal pole, where distinct anatomical features could be distinguished. Simultaneous review of postoperative MRI scans was useful, ensuring that all matched MRI slices were included in the resection. In all cases, consensus was reached by the two observers and the proposed MRI-pathology matches were plausible. Careful labelling, postoperative handling and the new method of orienting and slicing resected specimens ensured histopathological tissue blocks of uniform thickness and slicing angle. In 80% of cases confident and precise matching of MRI and blocks was possible, enabling MRI-pathological correlations. This technique can be applied to a range of MRI datasets, enabling exploration of the pathological basis of abnormalities on conventional and novel MRI. (Supported by The Wellcome Trust, UK)