Accurately recovering the shape features against rough and noisy segmentations is critical to achieving good anatomical correspondence between individual brain shape models. Our framework provides various tools for individual shape modeling, group-wise template construction, and shape deformity computation. And it has been used for large datasets of the human brain.
Demonstrating this procedure will be Dr.Jaeil Kim, a former grad student from my laboratory who developed the software for the brain shape modeling. For manual editing of the hippocampal segmentation, open the T1 weight magnetic resonance image and the automatic hippocampal segmentation results in the graphic user interface software. Click the icon in the display window to select the coronal view and scroll through the volume until the uncus is located.
Including the uncus in the hippocampal mask where it is present, use the add and subtract functions to edit the mask of the hippocampal body after the uncus has receded. Continue editing the hippocampal mask until the hippocampal tail is found. As the pulvinar nucleus of the thalamus recedes superior to the hippocampus, the fornix will emerge.
Finish editing the last coronal slice of the hippocampus in which the entire length of the fornix is visible, but not yet continuous with the splenium of the corpus callosum. Then, save the masks for both the left and right hippocampi in NifTI format. To construct a group template model load the shape modeling plug-in and click Open Directory to open a directory containing the binary masks of the study population of interest.
Input the desired number of vertices and click Template Construction for the group template construction. Then check the mean shape mesh. For individual shape reconstruction, load the T1-weighted magnetic resonance image of interest and its corresponding segmentation mask and select the working directory to save the files.
Select a template model for the individual shape modeling and check and modify the modeling parameters in the shape modeling plug-in as necessary. So our modeling framework is almost automated, however, some steps require user confirmation. For example, if the value for the hippocampal regions is not one users must change the intensity parameter.
Then, check the result in the 3D view of the toolkit workbench. To perform a shape and deformity measurement select the shape model of the subject of interest in the data manager of the software and click Select Template to select the template of interest to obtain the measurement. Here, a representative deformation of the hippocampal template model for individual shape reconstruction can be observed.
The method induces a large to small scale deformation of the template model to minimize the distortion of its point distribution while restoring the individual shape characteristics. In this figure, reconstructed shape models from two subjects with their segmentation masks are shown. In these images aligned individual shape models, their average model, and the shape difference vectors with an individual shape model can be observed.
These data represent average shape deformity maps projected onto the average model. For one group with a small brain tissue volume, and one group with a large brain tissue volume. The shape deformity maps of the two groups present opposite patterns of hippocampal shape difference at corresponding regions.
Check the segmentation mask and the individual or either the shared model together. If the model is not fitted to the image boundary, adjust the modeling parameters to achieve better results. Statistical analysis using the shape deformity can be performed to investigate the group-wise shape We have also provided MATLAB code for the analysis at our project page.
This robust method has been applied to a number of clinical studies, not only involving critical structure modeling such as Alzheimer's disease or aging study, but also foot-bone disorders which require the analysis of compound bones.
