ausblenden:
Schlagwörter:
-
Zusammenfassung:
Defining and measuring cortical thickness: histology and MRI
Abstract No:
2580
Authors:
Miriam Wähnert1, Marcel Weiss1, Stefan Geyer1, Robert Turner1
Institutions:
1Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany
Introduction:
Measurement of cortical thickness has gained increasing recent interest as a potential replacement for the artificial parameter of gray matter density, as used in VBM. MRI data has been used for estimation of cortical thickness in vivo, after gray matter (GM) segmentation [Fischl 2000, Hutton 2008]. However, MRI contrast mostly derives from the presence of myelin, so that images show myeloarchitecture. The problem is that the GM-WM boundary is strictly defined only by cytoarchitecture, since white matter (WM) does not contain neuronal cell bodies. Furthermore, myelinated axons are often found to extend far into cortex. T1-weighted MR images therefore show several regions where the GM-WM boundary is poorly defined. Investigation of brain sections stained alternately for myelin and neuronal cell bodies may provide detailed criteria for determining the GM-WM boundary from MR images with suitable contrast, and thus enable much more accurate in vivo estimation of cortical thickness.
Methods:
We sectioned the visual cortex of a human postmortem brain (fixed in 4% formalin) with a freezing microtome at 30 µm and stained the sections alternately for myelin (Gallyas stain) and cell bodies (Merker silver impregnation). The sections included primary visual cortex, showing the stria of Gennari (fig. 4) and the V1/V2 boundary. Micrographs of the sections were obtained with a Zeiss Axio Imager microscope (resolution 51.6 µm/pixel). The images from adjacent section pairs were masked and co-registered. Linear registration was performed by image translation and rotation. This approach proved to be inadequate, since due to their fragility, even consecutive sections can undergo significant mutual distortion during sectioning, mounting, and staining. Successful unwarping was achieved using the software package AIR 5 (Automated Image Registration, RP Woods). This provides good registration between cell body- and myelin-stained sections except where a section has actually been damaged.
We used the Laplace equation approach to determine 427 individual cortical profiles [Jones 2000, Annese 2004, Hutton 2008]. As preliminary boundaries, we used the pial boundary and the GM-WM boundary obtained by thresholding the cell-body stained section (fig. 1). The relaxation method was applied in 2D to calculate contours. To obtain a profile (fig. 1, red lines), we took the gradient of the solution of the Laplace equation using a weighted average at a starting point (fig. 1, yellow line, starting points are 52 µm apart). Then we advanced 1/5 of a pixel length in the gradient direction. At this new point, we took the gradient again. This was repeated until we reached the preliminary GM-WM boundary (fig. 1, blue line). Then the profile was extended by 30% at both ends. The profile was constructed at 100 equidistant points along a field line to ensure normalization.
Results:
Cortical profiles (fig. 1, red lines) derived using the Laplace equation are more realistic than the straight lines used in earlier studies [Schleicher, 2005]. These do not intersect, and they strike the cortical boundaries perpendicularly [Annese, 2004]. They are often consistent in direction with the larger intracortical blood vessels (fig. 1 & 3, circle), which implies that in several regions the plane of section was parallel to the vertical organization of the cortical columns [Schleicher, 2005]. In the myelin-stained section at 5 µm/pixel resolution (not shown), in these regions, the myelin fibres can be seen along their whole length across the cortex.
Then we used a sliding average. Each average profile results from 25 neighboring individual profiles on the cell stained section. The final pial boundary is identified as the maximum in the left half of each average profile (fig. 2, green dashed line). The final GM-WM boundary (fig. 2, red dashed line) is identified as the point just before the plateau representing the consistently lower cell density of white matter. This is where the intensity reached 96 % of its value in white matter.
The revised pial and GM-WM boundaries thus determined are plotted for each individual profile onto the cell stained section (fig. 3, green and red crosses). The lengths of three different profiles in between those final boundaries, i. e. their cortical thicknesses, are 3.50 mm, 1.74 mm and 2.32 mm (fig. 3, dark blue lines from left to right). The changes of cortical thickness may be due to a variation of the cutting angle.
To compare the location of the GM-WM boundary determined from the cell stained section to the myelin stained section [Eickhoff, 2005], we plotted the boundary at exactly the same location onto the myelin stained section (fig. 4). This was possible because of the registration. The cell GM-WM boundary corresponds well to what one would expect to be the GM-WM boundary on the myelin stained section.
Conclusions:
In order to compare myelin and cell stained sections they must be precisely registered.
The GM-WM boundary determined by our 96%-criterion from the average profiles of the cell stained section corresponds very well with the GM-WM boundary seen in the myelin stained section. Further studies will show, whether this also holds for other cortical areas. We plan to extend this to a more robust and quantitative definition of this boundary in high resolution (200 µm/pixel) MR images of postmortem and living human brains. This should enable anatomically much more valid in vivo mapping of cortical thickness.
