Magnetic Resonance Fingerprinting¹ (MRF) has been demonstrated as an efficient approach in simultaneous quantification of multiple relaxation parameters. The concept of simultaneous multi-slice (SMS) is adopted in MRF to further accelerate the acquisition along the slice direction. A main challenge for SENSE² based SMS-MRF reconstruction is the high g-factor penalty due to similar coil sensitivity profiles of collapsed slices. Therefore, we proposed a new method to create spatially controlled-aliasing³ by data sharing along the time axis to reduce the g-factor penalty. Through this method, good parameter mapping of 2 slices with nearly the same sensitivity is acquired at half the acquisition time of single-excitation (or single-band) MRF.

Acquisition 2-simultaneous-slice MRF acquisition in this work is based on an inversion recovery balanced SSFP sequence with varying flip angle and repetition time (TR) train, and highly undersampled variable density spiral read-out. Gz blips⁴ are used to provide π phase modulation in even TRs for the second slice, while no phase modulation is applied in odd TRs and the first slice. The SMS MRF scan took 9s to acquire 1000 time points. At the same slice locations, the single-excitation MRF were also acquired for reference. All brain MRF data were acquired at a 3T Philips scanner (Achieva TX scanner, Philips Healthcare) using an 8-channel phase-array headcoil.

Reconstruction Gridding reconstruction is applied to collapsed spiral data and t-blipped SMS-MRF⁵ with different slice phase pattern is used to estimate coil sensitivity profiles. Afterward, the temporal data sharing for gridded collapsed k-space data is implemented between two adjacent time points, as shown in Fig.1. In k-space, the odd k-lines of current time point are re-grouped with even k-lines of next time point. K-space of adjacent 2 TRs are combined such that an alternating phase pattern in y axis is achieved, thereby ensuring spatial controlled aliasing. In image domain,$$S_{i1}(x,y)\overline{X_{1}}(x,y)+S_{i2}(x,y+FOV/2)\overline{X_{2}}(x,y+FOV/2)\approx{I_{i}}$$ for the i^th coil is obtained after data sharing operations, where S is the slice sensitivity, $$$\overline{X}$$$ is the mean of adjacent 2 TRs and is the aliased image. Theoretical error of $$$\overline{X}$$$ in this equation is related to the slice signal difference between 2 TRs, with the assumption of negligible signal change at adjacent TRs, this error is insignificant. This data sharing method includes a sliding window process, while $$$\overline{X}$$$ has the same signal evolution as the slice image with one time point less, and after using SENSE reconstruction for slice-unaliasing, dictionary matching is applied to $$$\overline{X}$$$ series of each slice for obtaining MRI signatures.

This method dramatically improves the slice-unaliasing with reduced g-factor penalty, especially when two collapsed slices have similar coil sensitivity profiles. Fig.2a shows T1, and T2 maps of two slices reconstructed from SMS-MRF data using proposed method. Fig.2b shows T1 and T2 maps of two slices reconstructed from single-excitation MRF data. The acquisition time of SMS MRF (4.5s/slice) is half of single-excitationMRF (9s/slice).SMS-MRF significantly accelerate MRF acquisition, and temporal data sharing method has the capability in reducing the g-factor penalty of SENSE-based SMS-MRF reconstruction. Because temporal phase modulation can produce image FOV shift between collapsed SMS as well as controlled aliasing, hence the limitation of slice sensitivity does not exist. In this method, part of unalising error was caused by the slice image difference between adjacent TRs. Therefore better result would be obtained if the mentioned equation error was properly estimated. The accuracy of parameter mapping may be improved by applying optimization method, such as iterative matching.