Primatologist is the main entry point of the Primatologist toolbox. It is a pipeline, i.e. a sequence of basic processes. Primatologist performs atlas registration, tissue segmentation and surfaces and sulci extraction. Its building blocks (the individual processing steps) are stored in blocks.
To this day, only the segmentation of anatomical regions has been thorouglly validated. Structures volume can be computed from a posteriori probability maps generated by the Primatologist pipeline with the hierarchical analysis process.
To process data, it is advised to use BrainVISA's databasing system. The usual practice is to create one database per study. Raw MR images should be imported with the dedicated importation process.
EM Segmentation: fits a statistical model to the MRI by expectation-maximization. Intensities are assumed to originate from a gaussian mixture model on a MRF lattice with non-stationary priors. This step also includes densoising and bias field estimation .
The Center for In Vivo Microscopy, in Duke University, has published a high resolution, post mortematlas of the Rhesus Macaque (M. mulatta) brain. It contains anatomical (T2-weighted) and diffusion templates as well as a parcellation of the brain into 241 anatomic and functional structures.
We have added to this parcellation two missing regions : a pseudo-CSF region was obtained by a 1mm dilation of the brain, and corpus callosum was manually segmented in sagittal incidence by a single operator. We then used our Atlas Hierarchy Tools to extract a subset of nodes (Figure 2), corresponding in the end to only 17 labels (Figure 3).
Figure 2. Selected nodes from the complete hierarchy. Pure virtual classes (i.e. without a label) are written in italic.Figure 3. CIVM template and 17 selected labels.
Pseudo prior probabilities were obtained by extracting each label mask and smoothing it with a Gaussian kernel. Several of such probabilistic atlases are provided, with different smoothing radius. This task can also be performed during the segmentation process (see Basic Mixture).
This atlas has a very fine resolution (0.150 mm). We have thus also generated templates, labels and priors at coarser scales (0.3 and 0.6 mm). When using the optional resampling step (Align and Resmple), the MRI resampling and atlas resolutions should be the same.
To be usable with Primatologist, an atlas should also contain a hierarchy that associates labels with normalized names, another hierarchy that classifies structures between grey and white matters and CSF, and a Clique matrix that contains conditional neighbouring probabilities and which can be obtained from a volume of labels (see Compute Cliques from Labels).
Note:
For now, the naming norm is not described anywhere. To create your own Primatologist-compatible atlas, you should copy what was done for the CIVM atlas.
Any point located in the left hemisphere. If the MRI is in radiological orientation, this point should be located, on the screen, at the right of the interhemispheric plane.
Primatologist
