Quantum Optimization for MRF Framework

Magnetic Resonance Fingerprinting (MRF) is a quantitative MR technique that enables simultaneous mapping of multiple tissue properties from a single scan within a clinically feasible time. An MRF scan typically consists of hundreds to thousands of time points where RF pulses are applied to tip the proton spins in tissues. Each pulse employs varying flip angles and is separated with varying TRs to generate distinct signal evolutions for different tissues. Longer TRs and more acquisition time points allow for more accurate and robust parameter estimation. Various techniques have been adopted in MRF to speed up each single acquisition, such as k-space undersampling with variable-density trajectories, resulting in a scan time of approximately 13 seconds per slice. However, the scan time is still too long for large volumetric coverage, which leads to patient discomfort, higher expenses, and greater chance of motion artifacts. Although fully quantitative scans have long been the goal of magnetic resonance imaging, the tradeoff between maximizing precision and minimizing scan time has been the main challenge that prevents quantitative MRI from entering widespread clinical use. Despite the overall tradeoff relation between short scan duration and high measurement accuracy, pulse sequence optimization could potentially enable truncated MRF scans to achieve equivalent quality of tissue property mapping as that of long MRF scans.
Here we apply the quantum-inspired optimization methods to accelerate the MRF scans, while reducing the random and systematic error in measurements. We want to obtain sequences which are intrinsically robust against the well-known shading artifacts that otherwise arise in MRF images due to image undersampling and field inhomogeneity