Beomgu Kang1, Byungjai Kim1, Hye-Young Heo2,3, and Hyunwook Park1
1Department of Electrical Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Korea, Republic of, 2Russell H Morgan Department of Radiology and Radiological Science, Johns Hopkins University, Baltimore, MD, United States, 3F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States
1Department of Electrical Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Korea, Republic of, 2Russell H Morgan Department of Radiology and Radiological Science, Johns Hopkins University, Baltimore, MD, United States, 3F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States
We proposed a learning-based
optimization framework to improve quantification accuracy and accelerate data
acquisition for magnetization transfer contrast MR fingerprinting.
A schematic overview of the learning-based optimization of the
acquisition schedule (LOAS). MTC-MRF signals are synthesized using randomly initialized
scan parameters, noise, and tissue parameters (Input) and fed to the fully
connected neural network (FCNN). The FCNN outputs tissue parameter estimates
(Output). A loss function is a mean square error between the ground-truths and
estimated tissue parameters. The calculated loss was back-propagated with an
ADAM optimizer to update scan parameters.
Optimized MRF schedules consisting of
four scan parameters (B1, Ω, Ts, and Td) with 40 dynamic scans from
the LOAS, CRLB, IP strategies and PR,
and Linear schedule. CRLB strategy generates an acquisition schedule by minimizing CRLB
values of the MTC-MRF signal model. IP strategy generates an acquisition schedule using an IP optimization algorithm by maximizing signal difference between
different tissue types. PR schedule was generated by increasing spectral
and temporal incoherence between dynamic scans. PR schedule was the same as that
used in our previous studies.