4557
Accelerated Multi-band Magnetic Resonance Fingerprinting Using Spiral in-out with additional kz Encoding and Modified Sliding Window Reconstruction1Diagnostic Radiology, The University of Hong Kong, Hong Kong, China, 2Philips Healthcare, Hong Kong, China
Synopsis
Multi-band Magnetic Resonance Fingerprinting can be achieved using UNFOLD-like acquisition and dictionary matching without using parallel methods. However, the MR parametric maps after dictionary matching in one slice suffers from artifacts due to the high frequency components of other simultaneously acquired slices. In this work, a new acquisition strategy was proposed for the multi-band acquisition, where spiral-in-out trajectory was used to provide extra kz encoding. A modified sliding window reconstruction was also proposed to reduce the high frequency oscillations.
Purpose
Magnetic Resonance Fingerprinting (MRF) [1] is a novel approach to efficiently quantify multiple MR parameters, and multi-band (MB) has been implemented in MRF for further acceleration. UNFOLD [2] is an effective method to decouple the simultaneously acquired slices in multi-band MRF, given the different frequencies of their signal evolution. A major problem in UNFOLD-based MB-MRF is that the desired low frequency component of target slice is usually collapsed with high frequency artifacts from other slices due to undersampling and signal evolution. Here we proposed spiral-in-out MB-MRF to alleviate this problem in two aspects. First, one additional blip gradient was added in spiral in-out sequence, leading to better separation of frequency components using UNFOLD. Second, a modified sliding window [3] reconstruction was proposed to remove the high frequency artifacts before dictionary matching.Methods
A FISP-MRF [4] sequence was used and the readout gradient of our proposed sequence is shown in Fig 1a and the associated spiral-in-out trajectory in Fig 1b. The flip angle (FA) and repetition time (TR) trains are respectively shown in Figs 1c and 1d. The additional Gz blip (the green trapezoid in Fig 1a) is inserted in between the spiral-in and spiral-out readout portion, thereby introducing a $$$\Delta k_{z}$$$ shift between the two readouts. The additional Gz blip is equivalent to phase encoding step along kz. The encoding pattern of our proposed MRF acquisition is better visualized in the kz-t domain in Fig 2. Figs 2a and 2c show the interleaved acquisition pattern, which is similar to the stack of spiral 3D MRF [6]. Figs 2b and 2d show two representative sampling patterns using the additional Gz blip in our proposed acquisition strategy. Note that the density should be compensated along kz before reconstruction or matching for the sampling strategy in Fig 2b.
Given our proposed sampling strategy, the image series of different slices are encoded with unique phase pattern that favors UNFOLD. An example for MB factor of 3 is illustrated in Fig 3. The dictionary matching process is to search the maximum inner product between the acquired MR signal and dictionary $$$\sum_{t=0}^N m_{i,j}(t) = \sum_{t=0}^N s_{i}(t)d_{j}(t)$$$ where $$$\textbf{s}_i$$$ and $$$\textbf{d}_j$$$ respectively indicate the signal evolution vector for the ith voxel and the jth dictionary vector, $$$\textbf{m}_{i,j}=\textbf{s}_i\cdot\textbf{d}_j$$$, and $$$s_{i}$$$, $$$d_{j}$$$ and $$$m_{i,j}$$$ are their tth elements, and N is the number of TR. If $$$\textbf{S}_{i}$$$, $$$\textbf{D}_{j}$$$ and $$$\textbf{M}_{i,j}$$$ are Fourier transform of $$$\textbf{s}_i$$$, $$$\textbf{d}_j$$$ and $$$\textbf{m}_{i,j}$$$, then $$$\textbf{M}_{i,j}=\textbf{S}_{i}\circledast\textbf{D}_{j}$$$, where $$$\circledast$$$ indicates convolution, so $$$\sum_{t=0}^N m_{i,j}(t) = M_{i,j}(0)=\sum_{\omega=0}^N S_{i}(\omega)D_{j}(-\omega)$$$. Then the target low frequency component in signal is chosen by $$$D_{j}(\omega)$$$. Due to the high in-plane undersampling factor, the high frequency components of signal are not neglectable. The low frequency component is then collapsed with the high frequency components of other slices. With the extra encoding of kz, the frequency of different slices became more sparsely distributed, thus rendering less effect on each other.
To further reduce the high frequency artifact, a modified sliding window reconstruction method was proposed. As shown in Fig 2e, sliding window gridding was performed for each kz so that the combined k-space data still followed the desired sampling pattern in kz-t, and that in-plane artifacts and, more importantly, high frequency components of signals were subsequently reduced.
All data were acquired with a 3T MRI scanner (Achieva TX, Philips Healthcare) with 8-channel head coil. An in-house FISP-MRF sequence was used with imaging parameters: TE = 6ms, TI = 20ms, FOV = 300 x 300 mm2, image resolution = 1.17 x1.17 x 5 mm3, acquisition matrix size = 256x256. The acquisition window is 8.4ms, with a rotation increment of gold angle 222.5 degree for different dynamics, and TR varied between 12~14 ms.
Results and Discussion
Figs 4 and 5 show the parameter maps with MB = 3 and MB = 4 respectively. For 2D MRF, spiral-in-out trajectory can reduce the error caused by off-resonance effect, especially with long acquisition window. For MB-MRF, the use of our proposed spiral-in-out trajectory in conjunction with an additional Gz blip allows two different kz encoding within one TR, which helps unfold the slices in subsequent procedures. During the reconstruction of dynamic images, a modified sliding window method was proposed to remove the high frequency artifacts from irrelevant slices. Our results showed the proposed method improved the quality of UNFOLD-like MRF matching, thereby permitting higher MB factor (Figs 4 and 5).Conclusion
A new acquisition strategy and reconstruction method were proposed for MB-MRF that consist of an additional Gz blip for kz encoding and modified sliding window for improving the UNFOLD performance for MB-MRF.Acknowledgements
No acknowledgement found.References
[1] Ma, D., Gulani, V., Seiberlich, N., Liu, K., Sunshine, J. L., Duerk, J. L., & Griswold, M. A. (2013). Magnetic resonance fingerprinting. Nature, 495(7440), 187.
[2] Madore, B., Glover, G. H., & Pelc, N. J. (1999). Unaliasing by Fourier‐encoding the overlaps using the temporal dimension (UNFOLD), applied to cardiac imaging and fMRI. Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine, 42(5), 813-828.
[3] Cao, X., Liao, C., Wang, Z., Chen, Y., Ye, H., He, H., & Zhong, J. (2017). Robust sliding‐window reconstruction for Accelerating the acquisition of MR fingerprinting. Magnetic resonance in medicine, 78(4), 1579-1588.
[4] Jiang, Y., Ma, D., Seiberlich, N., Gulani, V., & Griswold, M. A. (2015). MR fingerprinting using fast imaging with steady state precession (FISP) with spiral readout. Magnetic resonance in medicine, 74(6), 1621-1631.
[5] Zahneisen, B., Poser, B. A., Ernst, T., & Stenger, V. A. (2014). Three‐dimensional Fourier encoding of simultaneously excited slices: generalized acquisition and reconstruction framework. Magnetic resonance in medicine, 71(6), 2071-2081.
[6] Ma, D., Jiang, Y., Chen, Y., McGivney, D., Mehta, B., Gulani, V., & Griswold, M. (2018). Fast 3D magnetic resonance fingerprinting for a whole‐brain coverage. Magnetic resonance in medicine, 79(4), 2190-2197.
Figures
