TrueFISP Cardiac MR Multitasking for T1-T2 Mapping at 1.5 T

Background: Cardiac MR Multitasking [1] is a promising framework for free-breathing, non-ECG quantitative CMR. It has previously only been demonstrated for GRE [2] sequences, primarily at 3T. However, at 1.5T, the workhorse field strength for CMR, TrueFISP is generally preferred over GRE and may additionally provide more suitable contrast for simultaneous cine imaging. Here we evaluate a TrueFISP multitasking sequence for T1-T2 mapping at 1.5 T and introduce an AI eddy current correction (ECC) module to remove artifacts from large k-space jumps in self-gated TrueFISP.

Methods: Four healthy volunteers were scanned on a Siemens Avanto 1.5T system. A prototype TrueFISP multitasking sequence cycled through five T2-prepared inversion recovery (T2IR) modules (durations 0 ms [IR-only for 0 ms] to 60 ms) and two flip angles (18°/55°) to further enable B1+ mapping. The scan was run twice to assess repeatability. Raw k-space data were acquired and reconstructed with and without the proposed ECC module, a scan-specific conditional variational autoencoder (CVAE) which learns the effects of preceding k-space jumps on the self-gating data used for cardiac/respiratory motion state identification and for multitasking temporal basis estimation; the trained decoder was then used to regenerate self-gating data without eddy current effects from k-space jumps (Fig. 1). The normalized dot-product between magnitude spectra of the k-t space sampling scheme (which induces the eddy currents) and the signal and image spectra was used to quantify the reduction of eddy currents. Repeatability was assessed by Bland-Altman analysis and paired t-tests of mean septal T1 and T2 values.

Results: Fig. 2a depicts images and time-resolved spatiotemporal profiles showing strong T1 and T2 weightings generated by the sequence, as well as the successful removal of eddy current fluctuations using the ECC module. Fig. 2b shows the spectrum of the k-t space sampling scheme (top), with matching peaks in the spectrum of the uncorrected self-gating data (middle), removed in the corrected data (bottom). A similar effect can be seen on the spectra of the time-resolved images (Fig. 2c) reconstructed using uncorrected and corrected self-gating data.

The ECC module significantly reduced eddy current energy in TrueFISP data by a factor of 17 (p=0.026, Fig. 2d) and reduced the Bland–Altman limits of agreement width by 71.6% for T1 and 47.6% for T2, indicating substantially improved repeatability. The multitasking TrueFISP maps provide qualitatively better blood–myocardium delineation than the GRE maps (Fig. 3).

Conclusion: TrueFISP Multitasking is feasible at 1.5 T, especially in concert with an ECC module, which improves repeatability of T1/T2 values. Further study in patients, comparison to reference methods, and extension of TrueFISP Multitasking to other field strengths is warranted.

Figure 1. Reconstruction method overview. The eddy current correction (ECC) module removes eddy current effects from the self-gating data as a pre-processing step to the conventional multitasking reconstruction pipeline.
Figure 2. a) Two separate contrast weightings and spatiotemporal profiles showing the removal of eddy-current temporal fluctuations and artifacts. b) Frequency spectra of the tiny golden angle k-t sampling schedule (top) which is the cause of eddy current artifacts, the self-gating data (middle) showing peaks corresponding to the same frequencies as the sampling schedule, and the self-gating data after ECC (bottom) which no longer has peaks at the eddy-current associated frequencies. c) Frequency spectra of the reconstructed images from conventional Multitasking (top) and ECC Multitasking (bottom) showing the downstream impacts of removing eddy currents d) Dot plot of eddy current signal energy for Conventional Multitasking and ECC Multitasking.
Figure 3. a) GRE and TrueFISP Multitasking maps at 1.5 T. b) Bland–Altman plots for repeatability of TrueFISP T1 and T2 values using Conventional Multitasking (left) and ECC Multitasking (right).