Dr. Jurgen Germann, Dr. Flavia Venetucci Gouveia, Dr. Mallar Chakravarty, Dr. Gabriel A. Devenyi
A recent publication, led by Drs. Gabriel A. Devenyi and Mallar Chakravarty from the CoBrALab describes an openly available pipeline that facilitates obtaining optimized measurements of volumetric differences in MRI and other types of brain imaging. The publication “Longitudinal deformation-based morphometry pipeline to study neuroanatomical differences in structural MRI based on SyN unbiased templates” compares this new pipeline to standard methods, demonstrating its superior sensitivity.
What are the benefits of this pipeline?
The open two-level deformation-based morphometry (DBM) pipeline is 4.5 times more sensitive in detecting longitudinal within-subject volume changes than the standard methods that are derived from registration to a common template brain space (one-level DBM). It is particularly suitable for capturing these changes in vivo.
This method was validated using an open dataset, OASIS-2, of 150 individuals with and without dementia. They found that the two-level DBM was superior in capturing cortical volume changes associated with cognitive decline across study participants. Thus, it has superior sensitivity and accuracy while limiting false positives. Their key innovation is constructing unbiased templates from the data itself, minimizing bias and registration errors.
When should I use DBM?
DBM allows detecting volumetric differences cross-sectionally (e.g., between groups) as well as longitudinal brain changes. DBM is a registration-based technique on a common template brain space (one-level DBM) that can be used without any tissue classification or atlas priors. It allows for arbitrary contrasts and application to novel species or anatomy. It is therefore unbiased and can be widely applied, particularly where tissue classification is not possible, as compared to voxel-based morphometry (VBM) pipelines.
How does this project contribute to open science?
The outcome of this project, an open software pipeline, is free to use and available to all on GitHub. The pipeline is licensed under a modified academic public license that, in contexts other than commercialization, allow users to use the software, distribute copies, receive source code, modify it, and distribute the modified versions under the same license. The project relies on additional open science practices: the article is published open access while open data was used to validate the new tools and test the robustness of results.
Figure 1. Illustration of the processing underlying the two-level DBM pipeline proposed.
Jacobian determinants for each individual input MRI are calculated at the subject level to capture differences and changes with maximal sensitivity. The subject-level average brain is then used to generate an unbiased common space, and the Jacobians are transformed into that space for statistical analysis.