To whom it may concern, I recently saw a video regarding bundle warp and I had a few questions: 1) Are we able to use this on whole brain streamlines or only smaller bundles? 2) If we are only able to do smaller bundles, is there an MNI atlas you are currently using to warp to MNI space? If so are you just taken segment at a time and getting individual warp fields? 3)Are the deformations in a format that we can convert to be ANTS compatible for other applications (such as improving T1 warping to an atlas by incorporating streamline warp calculations)? Thanks, Ajay
Dear Ajay, Thank you for your email. 1) BundleWarp is specifically designed to register individual bundles rather than whole-brain tractograms. While it is technically possible to use BundleWarp to register two tractograms by treating them as sets of streamlines (static and moving), it will require significantly more memory and time. Additionally, registration may not be optimal due to the presence of false positives in whole-brain tractograms. A better way to register whole brain tractograms with BundleWarp involves clustering the tractograms first to establish correspondences between clusters of two tractograms, then registering the corresponding clusters individually using BundleWarp. However, we have yet to test this approach thoroughly. In short, we recommend using BundleWarp exclusively for bundles. 2) We use the HCP atlas of bundles in the MNI space, which can be downloaded from https://figshare.com/articles/dataset/Atlas_of_30_Human_Brain_Bundles_in_MNI... You can register bundles from your subject to the same type of bundles in the atlas. BundleWarp outputs a warp map for each bundle. 3) It saves the Gaussian kernel and transformation matrix containing transforms for every streamline in the moving bundle as NumPy files. Best, Bramsh On Fri, May 17, 2024 at 2:09 PM Ajay Kurani <dr.ajay.kurani@gmail.com> wrote:
Dear Ajay, Thank you for your email. 1) BundleWarp is specifically designed to register individual bundles rather than whole-brain tractograms. While it is technically possible to use BundleWarp to register two tractograms by treating them as sets of streamlines (static and moving), it will require significantly more memory and time. Additionally, registration may not be optimal due to the presence of false positives in whole-brain tractograms. A better way to register whole brain tractograms with BundleWarp involves clustering the tractograms first to establish correspondences between clusters of two tractograms, then registering the corresponding clusters individually using BundleWarp. However, we have yet to test this approach thoroughly. In short, we recommend using BundleWarp exclusively for bundles. 2) We use the HCP atlas of bundles in the MNI space, which can be downloaded from https://figshare.com/articles/dataset/Atlas_of_30_Human_Brain_Bundles_in_MNI... You can register bundles from your subject to the same type of bundles in the atlas. BundleWarp outputs a warp map for each bundle. 3) It saves the Gaussian kernel and transformation matrix containing transforms for every streamline in the moving bundle as NumPy files. Best, Bramsh On Fri, May 17, 2024 at 2:09 PM Ajay Kurani <dr.ajay.kurani@gmail.com> wrote:
participants (2)
-
Ajay Kurani -
Bramsh Chandio