Rajiv G Menon1, Marcelo V.W. Zibetti1, and Ravinder R. Regatte1
1New York University Langone Health, New York, NY, United States
1New York University Langone Health, New York, NY, United States
Data-driven optimization
of sampling pattern provided significant improvement in the performance of 3D-T1ρ mapping for brain applications. A
significant reduction in the time required for the acquisition of 3D-T1ρ MRI data can be achieved.
Figure
3: Comparison of SPs at different AFs. (a) shows FS k-space and
SENSE reconstruction. (b) shows Poisson-disc SP at different AFs (4, 10, 20, 30)
and the corresponding Low-rank reconstruction. (c) shows the optimized SP at
the same AFs as (b) and resulting low-rank reconstructions. The improvement in
performance is highlighted by the arrows.
Figure
5: Comparison shows improved performance for optimized SP. (a)
shows NRMSE errors across iterations for k-space training and validation sets
(b) shows NRMSE errors across iterations for image training and validation sets
(c) shows lower k-space NRMSE errors at different AFs for the optimized SP
compared to PD SP, both SP using low-rank reconstruction (d) shows lower image
NRMSE errors at different AFs for optimized SP vs PD (e) the improvement of the optimized SP vs PD
in T1ρ mapping compared to FS reference at different AFs.