Combination of ESPIRiT and back-projection reconstruction for 3D MR fingerprinting within 2.5 minutes
Xiaozhi Cao1, Qing Li1, Huihui Ye1,2, Hongjian He1, and Jianhui Zhong1,3

1Center for Brain Imaging Science and Technology, Department of Biomedical Engineering, Key Laboratory for Biomedical Engineering of Ministry of Education, Zhejiang University, Hangzhou, China, 2State Key Laboratory of Modern Optical Instrumentation, College of Optical Science and Engineering, Zhejiang University, Hangzhou, China, 3Department of Imaging Sciences, University of Rochester, Rochester, NY, United States


A spiral projection acquisition scheme was implemented for 3D MR fingerprinting to achieve isotropic resolution of 1x1x1 mm3 in whole brain T1 and T2 mapping within 2.5 minutes by using efficient L1SPIRiT reconstruction (ESPIRiT) and back-projection reconstruction.


Magnetic Resonance Fingerprinting (MRF) is a novel quantitative imaging technique for estimating multiple parameters simultaneously 1. To realize 3D MR fingerprinting, stack-of-spiral (SOS) and spiral-projection-imaging (SPI) trajectories have been used in some previous studies 2-4. In this work, by combining the ESPRiT and back-projection reconstruction methods, the acquisition time of 3D MRF with SPI trajectory was reduced by half to 2.5 minutes.


The acquisition and reconstruction process were shown in Figure 1.

a. A spiral trajectory was rotated about x-axis along time points (TP) dimension and simultaneously about z-axis along acquisition group (AG) dimension.

b. By combining the spiral interleaves acquired from the same time point but different acquisition groups, spiral interleaves formed an undersampled disk-like k-space coverage in a same plane.

c. Undersample the “disk” with acceleration factor R=2, which means it could reduce the number of acquisition groups by half, resulting in a two-fold reduction of acquisition time.

d. ESPIRit 5 reconstruction method was applied on the undersampled disk-like k-space data. Since the “disk” was rotated along time points dimension, the reconstructed image of each time point was actually a parallel-beam projected image from a specific angle.

e. By using a sliding-window method 6 to select adjacent projected images and applying a back-projection reconstruction method 7 on these images, a series of 3D images with varying mixed contrast were obtained along the time point dimension.

f. Reconstructed images were template matched with a pre-calculated dictionary voxel-by-voxel to generate T1 and T2 maps by using the extended phase graph (EPG) 8 method. The dictionary entries were also implemented with sliding-window method in the same way as we did it in step e.

Figure 2 shows a typical pulse sequence design of a 3D MRF fingerprinting with fast imaging with steady state precession (FISP) readout 9 and SPI trajectory. A slab-selective gradient was used to achieve a 240-mm slab thickness and a dephase gradient was used to achieve a 4- dephase. Each acquisition group includes 500 time points in which the flip angles vary in a preset way as Figure 2b shows and the TRs were set at 16 msec constantly. Between two adjacent acquisition groups, a waiting time of 2sec was inserted for the recovery of longitudinal magnetization. A total of 30 acquisition groups were acquired for fully-sampled data as reference and 15 of them were used for the validation of the proposed method. The spiral rotates about x-axis at tiny golden angle (23.63°) along time points dimension. The isotropic spatial resolution of 1x1x1 mm3 was achieved for the identified parametric maps in a field of view (FOV) of 240x240x240 mm3. The measurements of a phantom and in vivo brains were performed on a Siemens 3T Prisma scanner with a 64-channel head coil.


The projected images by using full-sampled and under-sampled data are shown in Figure 3. By using ESPIRiT reconstruction, the quality of images from under-sampled data can approach to those from fully-sampled data with less RMSE than a direct 2D INUFFT reconstruction (6.25% compared to 19.53%). Figure 4 shows the in-vivo results by using original 3D INUFFT reconstruction method on fully-sampled data, back-projection reconstruction on full-sampled/under-sampled data, and combination of ESPIRiT and back-projection methods on under-sampled data, from left to right respectively. The acquisition time Tacq is (TR×NTP+TwaitNAG=(16ms×500+2000ms)×15=150s, where NTP is the number time points for each acquisition group, NAG is the number of acquisition groups and Twait is the waiting time between adjacent acquisition groups.

Discussion and Conclusion

We proposed a reconstruction scheme by combining the ESPIRiT and back-projection reconstruction on 3D MRF with SPI trajectory. Compared to original 3D SPI MRF using 3D INUFFT method [4], the reconstruction was split into two part, 2D INUFFT which transforms k-space data to projected images and back-projection which transforms projected images to 3D images. With a simple 2D INUFFT replaced by ESPIRiT, acquisition acceleration could be achieved by undersampling the “disk” data with less spiral interleaves. Therefore, the T1 and T2 maps of 1-mm isotropic resolution and FOV = 240x240x240mm3 could be obtained within half the acquisition time of original 3D SPI MRF, namely 2.5 minutes. A drawback of the proposed method is that the images seem to be smoothed which may be caused by the rudimental algorithm of back-projection method we are using currently. In the future work, we will utilize a better back-projection reconstruction scheme to fix this problem.


This work is supported by National Key R&D Program of China (2017YFC0909200), NSFC (81871428, 91632109) and Shanghai Key Laboratory of Psychotic Disorders (13dz2260500).


1. Ma D, Gulani V, Seiberlich N, Liu K, Sunshine JL, Duerk JL, Griswold MA. Magnetic resonance fingerprinting. Nature 2013;495(7440):187-192.

2. Ma D, Jiang Y, Chen Y, McGivney D, Mehta B, Gulani V, Griswold M. Fast 3D magnetic resonance fingerprinting for a whole-brain coverage. Magnetic resonance in medicine 2017; DOI:10.1002/mrm.26888.

3. Liaoc C, Bilgic B, Manhard M, Zhao B, Cao X, Zhong J, Wald L, Setsompop K. 3D MR fingerprinting with accelerated stack-of-spirals and hybrid sliding-window and GRAPPA reconstruction. NeuroImage 2017;162:13-22.

4. Cao X, Liao C, Li Q, Ye H, He H, Zhong J. Fast 3D MR fingerprinting with spiral projection acquisition for whole brain quantification imaging. In Proceeding of the 26th Annual Meeting of the ISMRM, Paris, France, 2018. p. 1017.

5. Uecker M, Lai P, Murphy M, Virtue P, Elad M, Pauly J, Vasanawala S, Lustig M. ESPIRiT-an eigenvalue approach to autocalibrating parallel MRI: Where SENSE meets GRAPPA. Magnetic resonance in medicine 2014;71(3):990-1001.

6. Cao X, Liao C, Wang Z, Chen Y, Ye H, He H, Zhong J. Robust sliding-window reconstruction for Accelerating the acquisition of MR fingerprinting. Magnetic resonance in medicine 2017;78(4):1579-1588.

7. Herman GT. Image reconstruction from projections. Real-time imaging 1995;1(1):3-18.

8. Weigel M. Extended phase graphs: dephasing, RF pulses, and echoes - pure and simple. Journal of magnetic resonance imaging : JMRI 2015;41(2):266-295.

9. Jiang Y, Ma D, Seiberlich N, Gulani V, Griswold MA. MR fingerprinting using fast imaging with steady state precession (FISP) with spiral readout. Magnetic resonance in medicine 2015;74(6):1621-1631.



Process of acquisition and reconstruction for the proposed method.


(a) Pulse sequence of 3D MRF with spiral projection acquisition.

(b) Flip angles pattern. TR was kept at 16 ms constantly for whole scan.

(c) Acquisition groups distribution.


The projected images reconstructed from full-sampled (a) and under-sampled using a simple 2D INUFFT (b) and ESPIRiT algorithm (d) as well as corresponding difference map compared with the full-sampled data (c,e).


T1 (a) and T2 (b) maps using different reconstruction schemes with full-sampled and under-sampled (acceleration factor R=2) data.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)