Correlation of quantitative MRI and neuropathology in epilepsy surgical resection specimens--T2 correlates with neuronal tissue in gray matter

Abstract

Newer MRI methods can detect cerebral abnormalities not identified on routine imaging in patients with focal epilepsy. Correlation of MRI with histopathology is necessary to understand the basis of MRI abnormalities and subsequently predict histopathology from in vivo MRI. The aim of this study was to determine if particular quantitative MR parameters were associated with particular histological features. Nine patients with temporal lobe epilepsy were imaged at 1.5 T using standard presurgical volumetric and quantifiable sequences: magnetization transfer and FFT2. The resected temporal lobe was registered with the volumetric MRI data according to our previously described method to permit correlation of the modalities. Stereologically measured neuronal densities and field fraction of GFAP, MAP2, synaptophysin and NeuN immunohistochemistry were obtained. Analyses were performed in the middle temporal gyrus and compared with quantitative MRI data from the equivalent regions. There was a significant Spearman Rho negative correlation between NeuN field fraction and the T2 value in gray matter (correlation coefficient -0.72, p=0.028). There were no significant correlations between any neuropathological and MR measures in white matter. These preliminary findings suggest that T2 in gray matter is sensitive to the proportion of neuronal tissue. Novel quantitative MRI measures acquired with higher field strength magnets, and so with superior signal to noise ratios, may generate data that correlate with histopathological measures. This will enable better identification and delineation of the structural causes of refractory focal epilepsy, and will be of particular benefit in patients in whom current optimal MRI does not identify a relevant abnormality.

Fig. 1

Example of images used for field fraction of NeuN immunohistochemistry. (a) Gyral crown of middle temporal gyrus at × 1.6 magnification (NeuN immunohistochemistry counterstained with H&E, 20 μm slice thickness). All cortical layers were included in the analysis. (b) Higher magnification (× 40) from the same specimen, used for field fraction analyses. Note that neuronal cell bodies, not only nuclei, are immunopositive, but there is only limited involvement of axons and dendrites. (c) Same field as in b with pixels immunopositive for NeuN highlighted in green. The RGB parameters representing immunopositive pixels (brown) are set in the first field analyzed.

Fig. 2

Normalization protocol. (a) Each subject's T1-weighted volume MRI scan was normalized to the T1 image template in SPM99, using the default normalization protocol. (b) The normalized T1-weighted image was hence in the same space as ROI1 and both could be reverse normalized to subject space using the parameters derived from the original normalization (a). (c) ROI1 could then be transferred to the equivalent histopathological section using the previously determined MRI:pathology correlation.

Fig. 3

Example of placement of ROI1. (a) The original T1-weighted volume MRI in subject space with a line demarcating the resection. (b) Zoomed image of right temporal lobe on T1-weighted volume MRI seen in a. (c) Zoomed T1-weighted volume MRI after normalization into SPM space, with region of interest ROI1 shown overlaid in white onto the T1-weighted image. (d) Zoomed image of the T1-weighted volume MRI and ROI1 after the back normalization to subject space. The same normalization parameters were used for both the T1-weighted image and ROI1. (e) The equivalent histopathological slice which correlates with the MRI slice (GFAP, 7 μm slice-thickness), showing ROI1, outlined with red marker.

Fig. 4

Registration protocol for ROI2. (a) The gyral crown of the middle temporal gyrus was identified on each patient's histopathological specimen (ROI2). (b) ROI2 was then determined on the corresponding T1-weighted volume scan using Analyze AVW 5.0. (c) The T1 volume scan was then registered to the PD images (and hence also the T2 images that were acquired simultaneously) using the SPM99 default registration options (registration 1). (d) The summed segmented T1 volume MRI (used to achieve image quality similar to MTR images) was registered to the MTR images using default options in SPM99 (registration 2). (e and f) The parameters from the two registrations were used to register ROI2 from T1 volume space to each patient's individual T2 and MTR space.

Fig. 5

Example of placement of ROI2. (a) A histopathological slice used to estimate quantitative histopathology in the gyral crown of the middle temporal gyrus, indicated in red (NeuN, 20 μm slice thickness). (b) The corresponding volume MRI slice with a line demarcating the resection. (c) Zoomed image of the right temporal lobe on T1-weighted volume MRI seen in b. (d) The same image as in c with the region of interest ROI2 (i.e. the cortical crown of the middle temporal gyrus defined after segmentation of the MRI volume data set) overlaid. (e) Zoomed image of the right temporal lobe on MTR images overlaid by ROI2 after appropriate registration.