Introduction

Magnetic Resonance Fingerprinting (MRF) is a robust and efficient MR technique for the estimation of MR parameters, such as T₁, T₂ and proton density ^1,2. That these MR parameters are quantitative and largely do not depend on MRI hardware and sequence configuration renders MRF an ideal technique for longitudinal studies, such as treatment effect on cancer patients. It is therefore imperative that the repeatability of MRF be evaluated. The purpose of this study is to assess the repeatability of MRF based on the inversion-recovery unbalanced steady-state (IR-FISP) sequence using the ISMRM/NIST MRI system phantom³.

Methods

All MRI experiments were performed using a 3.0T human scanner (Achieve TX, Philips, Best, Netherland). The ISMRM/NIST MRI phantom (Fig 1) was used for this repeatability study. It contains fiducial spheres filled with well-characterized T₁ and T₂ relaxation times. The phantom was scanned using the IR-FISP MRF sequence over 15 sessions throughout 65 days. The IR-FISP MRF sequence with spiral-in-spiral-out readout trajectory (acceleration factor = 58.4) was developed in-house. Detailed imaging parameters were: TR = 8.5ms to 10.1ms, flip angles = 5° to 60°, matrix size = 256 x 256, field of view = 30 x 30 cm², slice thickness = 5 mm, and 1000 dynamics per slice. An addition of a minimum 5 seconds delay was added before each acquisition to ensure the net magnetization to be in its thermal equilibrium state. The resulting scan time per slice was 55s. MRF was performed 5 different times in each session. MRF dictionary was generated using the extended phase graph algorithm using MATLAB (Mathworks Inc, Natick, MA). The dictionary contained T₁ values ranging from 300 to 6000 ms in 15 ms increment and T₂ values ranging from 10 to 6000 ms in 5 ms increment. Region-of-interest analysis of each fiducial sphere from all repetitions of each session was performed. In particular, the mean of each ROI on T₁ and T₂ maps were measured, and the median and standard deviation of the means amongst the 5 repetitions were recorded for all 15 sessions. Bland-Altman (B&A) analyses were performed to assess the agreement between the T₁ and T₂ estimated from MRF sequence and those provided by ISMRM-NIST (Table 1).

Results

There was a strong correlation between the T₁ and T₂ estimated from MRF versus the reference values (R² = 0.999, and R² = 0.981, respectively; Fig 2). In Bland-Altman analysis, the T₁ estimated from MRF has a bias of 74.8 ms, and the range of limit of agreement was between 8.37 and 141.2 ms (Fig 3a). All measurements were within the limits of agreement. The T₂ estimated from MRF has a bias of -32.2 ms with range of limits of agreement between -94.3 and 29.9 ms (Figure 3b). There was a measurement that lied above the limit of agreement (sphere T2-1) with a difference of 67.9 ms shorter than the reference value.

Conclusion

The MR parameters estimated from MRF display high repeatability over the 15 sessions. The T₁ estimated from MRF has a significantly higher degree of repeatability than T₂.

Acknowledgements

None