Accelerated DSI with Compressed Sensing using Adaptive Dictionaries by Berkin Bilgic, Kawin Setsompop, Julien Cohen-Adad, Van Wedeen, Lawrence Wald, Elfar Adalsteinsson. The abstract reads:
Diffusion spectrum imaging offers detailed information on complex distributions of intravoxel fiber orientations at the expense of extremely long imaging times (~1 h). Recent work by Menzel et al. demonstrated successful recovery of diffusion probability density functions from sub-Nyquist sampled q-space by imposing sparsity constraints on the probability density functions under wavelet and total variation transforms. As the performance of compressed sensing reconstruction depends strongly on the level of sparsity in the selected transform space, a dictionary specifically tailored for diffusion probability density functions can yield higher fidelity results. To our knowledge, this work is the first application of adaptive dictionaries in diffusion spectrum imaging, whereby we reduce the scan time of whole brain diffusion spectrum imaging acquisition from 50 to 17 min while retaining high image quality. In vivo experiments were conducted with the 3T Connectome MRI. The root-mean-square error of the reconstructed ‘‘missing’’ diffusion images were calculated by comparing them to a gold standard dataset (obtained from acquiring 10 averages of diffusion images in these missing directions). The root-meansquare error from the proposed reconstruction method is up to two times lower than that of Menzel et al.’s method and is actually comparable to that of the fully-sampled 50 minute scan. Comparison of tractography solutions in 18 major whitematter pathways also indicated good agreement between the fully-sampled and 3-fold accelerated reconstructions. Further, we demonstrate that a dictionary trained using probability density functions from a single slice of a particular subject generalizes well to other slices from the same subject, as well as to slices from other subjects
This is a pretty interesting study as there seems to be a possible speed-up of about ten times for a similar reconstruction error. It also seems that training on patient data seems to generalize to other patients.
A related presentation is here and an implementation is here.
Liked this entry ? subscribe to Nuit Blanche's feed, there's more where that came from. You can also subscribe to Nuit Blanche by Email, explore the Big Picture in Compressive Sensing or the Matrix Factorization Jungle and join the conversations on compressive sensing, advanced matrix factorization and calibration issues on Linkedin.
No comments:
Post a Comment