Click here to

Session: Imaging BLUE RIBBON [Return to Session]

Reducing Parametric Maps’ Error by Sliding Window Reconstruction and Inter-Phase Data Sharing in Low-Rank Four-Dimensional Magnetic Resonance Fingerprinting (4D-MRF)

Y Wong1*, C Liu2, T Li3, J Cai4, (1) The Hong Kong Polytechnic University, ,,(2) The Hong Kong Polytechnic University, Hong Kong, Hong Kong, (3) The Hong Kong Polytechnic University, Hong Kong, ,HK, (4) The Hong Kong Polytechnic University, Hong Kong, ,CN


SU-I400-BReP-F1-3 (Sunday, 7/10/2022) 4:00 PM - 5:00 PM [Eastern Time (GMT-4)]

Exhibit Hall | Forum 1

Purpose: Retrospective phase sorting of 4D-MRF data results in inconsistency of the accuracy of parametric maps. This study demonstrates a strategy to reduce the parametric maps’ error by utilizing sliding window (SW) reconstruction and inter-phase data sharing (IDS) before the low-rank projection of MRF data and dictionary for the subsequent iterative denoising.

Methods: The proposed strategy was tested on a single slice MRF data of a health volunteer acquired on a 3T scanner (GE Healthcare) with a 2D inversion-recovery FISP sequence and a total of 2000 frames. All parametric maps in this experiment were generated by low-rank back-projection (LBP) method. The accuracy and quality of parametric maps reconstructed by the SW+IDS+LBP, SW+LBP and LBP only strategy were evaluated by normalized root-mean-square error (NRMSE), peak signal-to-noise ratio (PSNR), and structural similarity (SSIM) index.

Results: The T1, T2, and PD maps generated by the proposed strategy (SW+IDS+LBP) have better accuracy and quality as compared to other methods. The quantitative results show that both T1, T2, and PD maps generated by SW+IDS+LBP method have lower NRMSE (T1: 0.176 ± 0.0242; T2: 0.147 ± 0.0213; PD: 0.0913 ± 0.0118), higher PSNR (T1: 53.8 ± 1.19; T2: 58.8 ± 1.33; PD: 20.8 ± 1.05) and higher SSIM (T1: 0.63 ± 0.0873; T2: 0.666 ± 0.0516; PD: 0.7 ± 0.0385) than that generated by SW+LBP, and LBP only methods. Meanwhile, the qualitative results suggested that the respiratory motion-resolved parametric maps retain high motion fidelity.

Conclusion: Both qualitative and quantitative results suggest that the proposed SW+IDS+LBP can reduce the parametric maps’ error induced by the data segmentation in the phase sorting process of 4D-MRF. The enhanced quality of the low-rank subspace images could potentially benefit the convergence of subsequent iterative reconstruction.

Funding Support, Disclosures, and Conflict of Interest: This work was supported by research funding from the GRF (15102219).


MRI, Quantitative Imaging


IM- MRI : Fingerprinting, synthetic MRI, Parametric mapping

Contact Email