Introduction Semah et al., 1998 McIntosh et al., 2001; Luders and Schuele, 2006 Blume et al., 2004; Jeha et al., 2006 Spencer et al., 2005 Palmini et al., 2004 Lee et al., 2005; Widdess-Walsh et al., 2006 Thus, there is a pressing need for further developments in MRI, such as higher field strengths to improve signal-to-noise ratio, and newer sequences that may better pick out culpable neuropathological lesions in the neocortex. Cascino et al., 1991; Van Paesschen et al., 1995 Thom et al., 2005 Eriksson et al., 2005 Methods Subjects Eriksson et al., 2005 The age range of subjects was 31–46 years (median 36 years) and 4 were male. Seven had a right anterior temporal lobe resection, and two had surgery on the left. The surgical procedure was selected after our standard presurgical work-up, including EEG-video telemetry, conventional MRI (T1- and T2-weighted, proton density and FLAIR), psychological and psychiatric assessments and multi-disciplinary case conference. Salmenpera et al., 2007 MR acquisition All scans were acquired on the same 1.5 T GE Signa MR scanner (GE Medical Systems, Milwaukee, WI). The T1-weighted volume sequence from the conventional scan protocol for all subjects was required for the MR:pathology correlation. The sequence parameters were: TE/TR/TI/NEX, 4.2/15/450/1; flip angle 20°; acquisition matrix 256 × 192; field of view 24 × 18 cm; 3/4 phase FOV, 124 contiguous 1.5 mm slices, giving a voxel size of 0.94 × 1.25 × 1.5 mm. Subjects had two additional scans to generate quantitative MR data: T2 mapping 1 2 2 1 S 1 S 2 S 1 S 2 Rugg-Gunn et al., 2005 Magnetization transfer imaging Rugg-Gunn et al., 2003 MRI quantitation Salmenpera et al., 2007 t p p Salmenpera et al., 2007 Quantitative MR values were obtained by overlaying a region of interest on the quantitative MR data. The mean intensity value of all voxels within the region of interest was calculated. MR:pathology visual correlation Eriksson et al., 2005 All subjects had post-operative MRI scans to indicate the extent of the resection. Histopathology Eriksson et al., 2006 Fig. 1 Histopathology and MRI correlation MRI-directed Table 1 Eriksson et al., 2006 Fig. 2 Fig. 2 Fig. 2 Fig. 2 Fig. 3 We also defined a region in the pathological data and determined its MR characteristics. To achieve this, the MRI volume data were segmented in subject space using intensity thresholding and manual drawing in Analyze AVW 5.0 (BIR, Mayo Clinic, Minnesota). The entire resection volume and the middle temporal gyrus within the resection volume were extracted by manual drawing guided by visual inspection of the previously determined MR:pathology correlation. Intensity thresholding and visual inspection were used to separate gray and white matter. A final manual segmentation step was used to separate the gray matter of the gyral crown (ROI2), from the gray matter in the rest of the middle temporal gyrus. This step used similar criteria to characterize the limits of the gyral crown as were used by the neuropathologist in the quantitative neuropathology. The data from the initial MR:pathology correlation were used to determine which MR slices within the middle temporal gyrus in each subject corresponded to ROI2. Each subject's MR volume data were registered to the advanced MR image data in the native subject space using SPM99 and the default registration option. Because of the different qualities of the T2 maps and MTR image data, different registration was performed and different templates used that more closely resembled the two data sets. Fig. 4 Fig. 5 Quantitative MR and histopathology measures were compared using SPSS (version 11.0.0). Results Table 1 Table 2 Table 3 Table 4 p Discussion Eriksson et al., 2005 We wished to compare the same area in all patients because it is not known how much histopathological measures and MR parameters differ as a function of location within cerebral tissue. Although some of the histopathological features of interest were readily quantified with a semi-automated protocol, stereological measures require significant expert time input. In consequence, we restricted this study to two regions of interest. The first region investigated corresponded to a region of abnormal T2 signal on SPM analysis in one patient and was chosen since it was of sufficient size to provide robust histopathological data and was available in all resections. Our second region was identified in each subject's histopathological sections, the gray matter of the crown of the middle temporal gyrus overlying ROI1, giving us data in adjacent gray matter. Eriksson et al., 2005 Rugg-Gunn et al., 2005 Lexa et al., 1994; Brochet and Dousset, 1999 Rugg-Gunn et al., 2003 Van Paesschen et al., 1997; Rugg-Gunn et al., 2005 Gray matter neuronal counts did not correlate with the field fraction estimate of NeuN. The gray matter neuronal counts are a count for each neuron observed in the region of interest expressed as a proportion of the total area assessed. The field fraction estimate of NeuN is an estimate of the area occupied by tissue immunopositive for NeuN as a proportion of the total area assessed. This tissue is predominantly neuronal cell bodies with some axons and dendrites. Thus 5 large neurons will occupy a greater area of immunopositivity than 5 small neurons. The neuronal count would be the same in each case but the field fraction would be different. This may explain the lack of correlation between the field fraction and the neuronal count. The relation of these two measures with the T2 relaxation value suggests that the T2 relaxation is dependent on more than just the number of neurons, also on their size or complexity, which may reflect their function. Further staining protocols and analyses of neuronal size are necessary to elucidate further this difference. The lack of any other correlations between the MR and the histopathology measures could be due to a number of factors. A real correlation may exist and not have been detected, possibly because of the limited number of subjects in the study. A change of MR scanner curtailed data collection for this series and we are now recruiting more subjects imaged using a new 3 T scanner. We could also have missed a true correlation because the registration between the MR region and the pathological region was inadequate as discussed previously. Alternatively, there may not be any other true correlations between the MRI sequences and pathological measures studied. The pathological measures characterize real physical features — for example, the actual number of neurons and the proportion of tissue affected by gliosis. However, the underlying physical significance of the MR data is not yet well characterized. We may have chosen pathological measures which do not relate to the features characterized by the MR signals. Myelin content, for example, might influence both T2 and MTR. We did not think it likely that myelin studies would be of relevance in epilepsy in the context of this paper, but additional histopathological analyses, such as LFB for myelin or myelin basic protein, may lead to further useful data in future studies. Tables 3 and 4 Our protocol for matching MR and pathology has demonstrated how we can investigate both the histopathology directed by an MR finding and the MR data directed by a histopathological finding. We identified a biologically plausible correlation between an increased T2 relaxation time and a reduction in neuronal labeling in the cortex of the middle temporal gyrus. If newer, quantitative MRI sequences are to be adopted more widely in routine clinical practice, further research exploring their underlying biological and pathological substrates will be necessary and may ultimately lead to a reduction in the proportion of patients with drug-resistant focal epilepsy whose optimal MRI is said to be “normal”.