Accelerating quantitative susceptibility imaging acquisition using compressed sensing.

To evaluate the feasibility of compressed sensing (CS) for accelerating quantitative susceptibility mapping (QSM) acquisition in MR histology, control, demyelination, and remyelination mice were scanned using a modified 3D gradient echo (GRE) pulse sequence (allowing undersampling the k-space in two phase dimensions) at 9.4 T. Fully sampled and CS data at various acceleration factors (AF) (4.0-16.0) were acquired with high spatial resolution of 22.5 µm3 isotropic. For prospective scans, the local phase was reconstructed slice-by-slice using a nonlinear reconstruction algorithm with phase cycling, and the QSM was calculated using the iLSQR method. Susceptibility values of major white matter tracts at various CS AF were compared with the fully sampled data as ground truth. The reconstructed QSM preserved major information with negligible artifacts, even at CS AF of 8.0 and 16.0. The maximum variation of susceptibility value of control mice was 5.8% in the fimbria (FI) area with the acceleration factor of 16.0, compared to the ground truth. The maximum variation was 9.5% in corpus callosum (CC) area of the demyelination mouse brain with the acceleration factor of 8.0. The maximum variation was 8.2% in AC area of remyelination mouse brain with 16.0 acceleration factor. This preliminary study shows that CS can reduce the acquisition time of GRE MRI in MR histology without losing significant accuracy in quantitative susceptibility values.

Duke Scholars

Author G. Allan Johnson Radiology