This wiki page describes a procedure for cortical surface reconstruction from ExVivo brain scans.

The procedure is tailored for a particular ExVivo scan protocol used at MGH, which acquires a series of MEF scans for each hemisphere of a postmortem brain sample. The method has only been tested on a few data sets, and further development and improvement is still on the way.

The method contains two major stages. The first stage aims to get a good WM segmentation (with necessary manual editing involved). The second stage performs a multi-spetrcal surface deformation to find the gray/white and pial surfaces.

The first stage consists of the following steps:

Step 1. Preparing the data. The MEF protocol used for the ExVivo brain analysis provides a large number of image volumes. For example, typically 6 different flip-angles are used. Each flip-angle corresponds to multiple volumes acquired at different echo times. In addition, multiple acquisitions are typically performed for each flip angle. In the first processing step, we create a single average volume for each flip angle, by averaging across the multiple acquisitions and the multiple echo volumes. "mri_average" is sufficient for this purpose. Note that, in this step, we keep the original scan resolution and size by specifying "-noconform" in the "mri_average" command line. We will convert relevant volumes to the COR format later before surface topology correction and surface deformation. This is based on the consideration that the extra interpolation (as performed when converting an input volume to the COR format) may interfere with the intensity clustering based tissue segmentation. Also, the COR format forces the intensity values to be of 8-bit BYTE type, which drops information from the original float-typed image.