Purpose

Magnetic Resonance Fingerprinting (MRF) [1] enables simultaneous generation of quantitative maps of multiple tissue properties and potentially enables the development of standardized imaging biomarkers. For clinical applications, the T₁ and T₂ values must be repeatable so that any observed relaxivity difference within a tissue can be assumed to be due to differences in physiology rather than scanner instability and difference in pulse sequence or map reconstruction implementations. Previous research has shown that highly repeatable T₁ and T₂ measurements can be made over time with MRF [2] with excellent reproducibility across several scanner types [3]. The purpose of this study was to evaluate multicenter repeatability and reproducibility of T₁ and T₂ estimates of MRF using the ISMRM/NIST MRI system phantom [4] and normal prostate regions in patients.

Methods

Experiments were performed on four different 3T scanners (1 Skyra and 3 Verio, Siemens Healthcare, Erlangen, Germany) with different software versions (VE11C, VB19, and VB17) in three different medical institutions: University Hospitals Cleveland Medical Center at Case Western Reserve University (CWRU) and Brigham and Women's Hospital (BWH) in US, and Multi-Imagem Clinic (DASA) in Brazil using the MRF-FISP acquisition [5] with the following parameters: FOV 400x400mm²; matrix 400x400; flip angle 5-75°; TR 11.2-15ms; slice thickness 5mm; acquisition time 39 s/slice. During MRF acquisition, the raw data were accumulated and passed on to the Gadgetron MRF reconstruction pipeline [6] along with a pre-calculated dictionary [7]. The accuracy of the reconstruction was validated using the T₂ layer of ISMRM/NIST MRI system phantom with T₁ values between 307.7 and 2360ms and T₂ values between 19.1 and 479.2ms. The phantom was placed in the magnet for at least 20 minutes before the acquisition to decrease the motion effects. No B₀ and B₁ maps were collected in this study. Six MRF measurements were made during one session with a delay of 5 secs between measurements. The phantom was then moved out of the magnet and placed again in the magnet, and the phantom was again allowed to settle for at least 20 minutes followed by another MRF acquisition for same-day test-retest reproducibility. The repeatability of T₁ and T₂ estimates was assessed using the coefficient of variation (CV), defined as the ratio of the standard deviation to the mean T₁ and T₂ values over 6 measurements. The CV of T₁ and T₂ estimates between the four clinical scanners is shown to demonstrate inter-scanner variation. The same-day test-retest normal prostate mean T₁ and T₂ values from right peripheral zone (RPZ), left peripheral zone (LPZ), right transitional zone (RTZ), and left transitional zone (LTZ) is also shown.

Results

Over the wide ranges of T₁ and T₂ values found in the ISMRM/NIST system phantom, intra-scanner MRF T₁ estimates had less than 2% variation and T₂ estimates had less than 4.7% variation, with the exception of T₂ relaxation of 19.1ms, which showed a variation of 8.9% (Fig.1a,b). The short T₂ relaxation times were on the order of the TR used for the MRF measurement. The test-retest reliability coefficient for both T₁ and T₂ estimation was above 0.99. Inter-scanner variation of mean from all 6 measurements and only measurement #5 showed T₁ variation less than 4.9% and T₂ variation less than 8.1% between multicenter scanners, and the variation increased when T₂ was lower than 28.8ms (Fig.2). The same-day test-retest in in-vivo normal prostate T₁ and T₂ values showed good agreement in four different regions (Fig.3). Figure 4 shows representative prostate MRF T1 and T2 maps from the four different scanners.

Discussion

The T₁ and T₂ values measured with MRF-FISP using the ISMRM/NIST MRI system phantom and from normal prostate regions in patients were highly repeatable and reproducible over multiple repetitions and different scanners. High repeatability is also observed in the inter-scanner data, even with different position, room temperature and inhomogeneous B₁. The lower test-retest agreement in NPZ at very high T₂s (~300 ms) is likely due to dictionary coarseness; T₂ step size was set to 10ms from 160 to 300ms and 50ms above 300ms. These settings can be optimized for future applications. These results demonstrate that T₁ and T₂ values are highly comparable between multicenter scanners, and that quantitative multiparametric imaging with MRF for multicenter clinical studies may be feasible.

Conclusion

MRF measurements of T₁ and T₂ are highly repeatable and reproducible between MRI scanners at different centers on different continents.

Acknowledgements

Siemens Healthineers, 1R01EB016728, 1R01DK098503, 1R01CA208236, P41 EB 015898, R01EB020667

References

[1] Ma et al. Nature. 2013;495:187-192. [2] Jiang et al. MRM. 2017;78:1452-1457. [3] Körzdörfer et al. Proc. of ISMRM Workshop on MRF. 2017. [4] Russek et al. Proc. of ISMRM, 2012. #2456. [5] Jiang et al. MRM. 2015;74:1621-1631. [6] Lo et al. Proc. of ISMRM Workshop on MRF. 2017. [7] McGivney et al. IEEE TMI. 2014;33:2311-2322.