SVReg stands for Surface-Volume Registration. SVReg is an algorithm and a software tool for co-registration of human-brain MR images. It uses anatomical information from both the surface and volume of the brain images for accurate automated co-registration which allows consistent surface and volume mapping to a labeled atlas. Brain labeling is a necessary preprocessing step for several studies looking at group comparisons and regional brain analysis.
Deformation map from atlas to subject
By default SVReg will output:
- Labeled brain surfaces and volumes into segmentations of ~90 ROIs
- Computation of GM, WM, CSF, total volume, average cortical thickness, and surface area of each ROI
- Vertex-wise cortical thickness mapped to atlas surfaces
- Delineation of up to 76 sulci on the mid-cortical surface
SVReg utilizes a series of multi-step registration and refinement processes based on several consideration of useful morphological features and known variation of human brain anatomy. It has the ability to analyze both normal and abnormal brains of subjects starting at least 4 years old, even working well with lesion brains.
SVReg has been customized for registration of individual subjects to the atlas templates provided within BrainSuite’s installation package. Conversely, the provided atlas templates were created to cater to SVReg’s registration model for optimal results. Within the GUI interface, users may choose from two atlases that have been preprocessed and segmented by expert neuroanatomists.
- The Colin27 atlas is a stereotaxic average of 27 scans of an individual which is widely used by the neuroscience community.
- BCI-DNI_brain atlas is an in-house high-resolution atlas that was created with our collaborators at the Brain and Creativity Institute’s Dornsife Neuroscience Imaging Center at the University of Southern California.
Users may choose to run SVReg through the GUI or through command-line for batch processing and customization.
Users have the option to choose from a number of computational flags or designate manual inputs for highly controlled registration including manual masks, sulcal inputs, and atlases. SVReg can additionally be run in a one step process or be run piecemeal by executing individual modules.
Additional tools are provided for users who would like to further analyze SVReg.
Our data processing package includes stand-alone functions to read and write SVReg outputs and main scripts to perform the following functions:
- ROI-wise group comparison
- Vertex-wise group comparison
- Surface smoothing
- Regenerating statistics after manual corrections
- Creating custom atlas
- Transforming data to atlas space
- Edit ROIs on surfaces and transfer edits to volume labels
Additionally we provide tools for:
- Manual bias field correction
- Population or group modeling of cortical or sulcal morphology
- Anatomical ROI filtered tractography analysis