Spiral MRI acquisitions^1,2 can achieve a long sampling duration τ vs. standard Cartesian methods, allowing them to concurrently decrease scan time while also increasing image SNR³. However, increasing τ comes at the expense of increased sensitivity to B0-field inhomogeneities, which cause image blurring. This sensitivity is further compounded by the scanner's inherent B0 drift, which can be observed as a time-dependent change in center-frequency ΔF0(t) under gradient loading. Nevertheless, if an accurate B0 map is known at the time of the spiral scan, this may be used to remove the blurring artifact during image reconstruction⁴.

The current work is part of a larger spiral project whose goal is to achieve a comprehensive, routine brain exam with spiral MRI. For this spiral exam, we propose a two-stage deblurring strategy: (stage 1) a reference B0 map, B0_ref, is acquired at the start of the exam. However, due to B0 drift, B0_ref may not accurately reflect the B0 field observed by a later spiral scan; deblurring using B0_ref would then result in residual blurring artifact. It would be time prohibitive to acquire a B0 map prior to every spiral scan, therefore (stage 2) F0 navigators (F0-nav) are used to quickly measure the DC component of B0 drift over time in order to estimate a corrected B0 map, B0_corr, for deblurring of the current spiral scan.

The F0-nav (~ 3 s) consists of STIR fat suppression, RF excitation, followed by four adiabatic refocusing pulses (two each in two orthogonal directions⁵) to volumetrically excite and measure the water resonance frequency. Experiments were performed on a Philips Ingenia 3T scanner.

Experiment 1 was designed to validate F0-nav measurements in the presence of B0 drift. A block of 3 scans (B0 map, s1, s2) was repeated 9 times on a Philips resolution phantom, with an F0-nav inserted at the end of each scan. B0 maps (3D-FFE mDIXON, res = 2 mm³, acq = 3, ΔTE = 0.7 ms, T = 86 s) were used to independently verify the accuracy of the F0-navs, while s1, s2 were spiral scans used to establish heavy gradients duty cycle conditions. Gradient temperature was also recorded in real-time via temperature probes at 6 locations.

Experiment 2 demonstrates F0-navs for improved deblurring in a typical spiral brain exam. B0_ref was acquired at the start of the exam, followed by a series of anatomical spiral scans based on spherically-distributed spiral (SDS) trajectories⁶, with an F0-nav inserted at the end of each scan. For each scan t, ΔF0(t) = F0(t) – F0_ref, where F0(t) is the F0-nav measurement at the end of scan t, and F0_ref is the center frequency for B0_ref. We then have B0_corr(t) = B0_ref + ΔF0(t), where B0_corr(t) is the B0-drift corrected B0-map. B0_corr(t) is then used in a conjugate-gradient algorithm⁴ for deblurring and water/fat separation of scan t.

Data from experiment 1, 2 are shown in Fig. 1, 2, respectively. In Fig. 1, B0 drift measured by F0-navs (ΔF0) and B0-maps (ΔB0) are in good agreement, with error = 2.4±2.0 Hz (mean±std). Increase in gradient temperature and B0 drift directly correlate with gradient activity (green bars), while the time to return to resting values is on the order of ~ 10 min for gradient temperatures and ~ 1 hr for B0 drift. In Fig. 2, an example spiral scan is shown ~ 30 min into the spiral exam where ΔF0 = 32 Hz. Image quality is significantly improved with spiral deblurring using B0_corr (bottom) vs. B0_ref (top).

Short F0-navs successfully track B0 drift over time (Fig. 1), enabling a spiral deblurring strategy where only an initial B0 map needs to be acquired. B0 maps showed only small spatial variations over time, suggesting that B0 drift can largely be characterized by the F0-nav’s DC measurement alone. We observed a B0 drift ~ 2.2 Hz/min under the max gradient load evaluated (Fig. 1, 0–25 min), which is consistent with other MR vendors⁷, and necessitates compensation especially for spiral MRI due to its sensitivity to off-resonances.

A B0 map with B0 drift correction, B0_corr, is shown to improve spiral deblurring (Fig. 2) using data from an F0-nav at the end of the scan. This strategy may be extended by inserting an F0-nav at both scan start and end, and interpolating between them to estimate the F0 for each spiral arm; intra-scan correction may then be performed by demodulating each spiral arm by its corresponding F0. These B0 drift correction methods will be particularly beneficial for long τ scans, and heavy gradient duty-cycle applications.