Abstract:
HTML
Purpose: True-FISP is an important steady-state contrast mechanism that is
frequently used in cardiac and interventional MRI where its rapid acquisition
and high SNR are particularly important. Current gradient hardware is capable
of true-FISP acquisitions whose TR is limited, not by the hardware, but by the
risk of gradient-induced peripheral nerve stimulation (PNS). This work seeks to
design a set of true-FISP acquisitions that are Pareto-optimal with respect to
minimizing both TR and bandwidth/pixel (BW) subject to PNS constraints.
Methods and Materials: The start, ramp times, and duration of each of the eight
gradient lobes in a true-FISP pulse sequence were allowed to vary
independently. For a given duration, the gradient amplitude was calculated to achieve
the specified resolution or slice-thickness. The gradient amplitude and slew
rate constraints were then enforced. The PNS was constrained using the SAFE
(Stimulation Approximation by Filtering and Evaluation) model. The objectives
were to minimize the TR and the BW.
The NSGA II (Non-dominated Sorting Genetic Algorithm) can simultaneously
optimize multiple non-linear objectives subject to multiple non-linear
constraints. It simulates natural evolution by beginning with a random
population of pulse sequences. Their TR and BW, along with the constraints, are
compared to determine fitness for survival and reproduction. Pairs of surviving
pulse sequences are combined, using simulated reproduction and mutation, in
order to obtain new pulse sequences. The process is repeated until the NSGA II
converges. The population size was 20 and the algorithm was allowed to run for
500 generations. An archive of all tested pulse sequences was maintained. The
non-dominated sub-set of this archive was chosen as the Pareto-optimal set.
Results: 137 Pareto-optimal true-FISP pulse sequences were found. Each sequence
represents the minimum BW possible at a given TR without exceeding the PNS or
hardware constraints. Alternatively, the minimum TR can be selected for a
required BW. One example sequence was able to achieve an 18% reduction in TR
(4.2 ms v. 5.1 ms) with no loss in BW (550 Hz/pixel) relative to the true-FISP
sequence most commonly used at our institution.
Conclusion: The NSGA II can be used, in conjunction with the SAFE model for
PNS, in order to design the minimum TR and minimum BW true-FISP pulse sequences
possible without causing PNS. The multi-objective procedure results in a set of
optimal pulse sequences allowing for flexibility in choosing TR or BW while both
maintaining optimality and avoiding PNS. (B.M.D. received a Graduate Fellowship
from the Whitaker Foundation. J.L.D. received grants from Siemens Medical
Solutions and the NIH.)
Questions about this event email: duerk@uhrad.com
Dale BS, B,
Time-Optimal True-FISP Pulse Sequences with Gradient-Induced Peripheral Nerve Stimulation Constraints. Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL.
http://archive.rsna.org/2003/3102487.html