Application-driven MRI

Matlab demo for the joint reconstruction-segmentation method from undersampled MRI data presented in MICCAI 2014;1:106-113 [pdf]. A 2D brain phantom is retrospectively undersampled and jointly reconstructed and segmented identifying four classes with a Gaussian mixture model. The accuracy of the segmentation relative to a fully sampled segmentation is compared with a separate method, where an intensity image is first fully reconstructed and then segmented.

Note
The demo requires to download and install the packages KSVD-Box v13 and OMP-Box 10, which can be obtained from Ron Rubinstein's website.

Application-driven MRI demo code


DLMRI

Matlab implementation for the reconstruction of undersampled dynamic magnitude MR data using the algorithm presented in TMI 2014;33(4):979-994 [pdf]. The demo will restrospectively undersample a 32-coil recombined (no SENSE acceleration) cardiac cine dataset and reconstruct it with the proposed method.

Note
The demo requires to download and install the packages KSVD-Box v13 and OMP-Box 10, which can be obtained from Ron Rubinstein's website.

DLTG demo v2

The following is an older version (October 2012) of the software which was used for the results in MICCAI 2012;1:256-263 [pdf]. The main difference with v2 is that v1 will only process magnitude real-valued data, but not complex-valued data.

DLTG demo v1