Cardiac T1, T2, T2*, and PDFF Mapping at 0.55T with Rosette MR Fingerprinting and a Deep Image Prior Reconstruction (RF_TH_)

Background: Given that cardiovascular disease can result in changes to multiple magnetic properties^1,2, techniques which enable the quantification of multiple properties in a single scan such as cardiac MR Fingerprinting (cMRF) may be beneficial^3–6. Recently, Si, et al. proposed a 3D cMRF sequence for T₁ and T₂ mapping on 0.55T systems⁷; at higher field strengths, previous work has extended cMRF to simultaneously map T₁, T₂, T₂*, and PDFF⁶. We build on these previous works, introducing a 2D cMRF sequence for 0.55T systems that can simultaneously collect T1, T2, T2*, and PDFF maps using the rosette trajectory and a Deep Image Prior reconstruction⁸.

Methods: The ISMRM/NIST MRI system phantom, an in-house oil/water phantom, and three healthy subjects (with IRB approval) were scanned to assess the quantitative accuracy and quality of the tissue property maps measured with rosette cMRF. An ECG-gated cMRF sequence with the following parameters was used: 24 heartbeats , 8 TRs per HB, 18 echoes, TR/TE1/ΔTE = 33.0/3.08/1.57 ms, FA range = 5.3-20°, 1.6x1.6x8.0mm resolution, 300x300 mm2 FOV. Data were collected on a 0.55T MRI system (MAGNETOM Free.Max, Siemens Healthineers, Erlangen, Germany) using an external gating device (Expression MR400, Phillips N.V.). Image reconstruction was performed using a Deep Image Prior method⁸, which was extended in this project for rosette T₁, T₂, T₂*, and PDFF mapping. T₂* estimation was performed by fitting the images to an exponential decay model⁹. Fat/water separation was performed using graph-cut IDEAL¹⁰. Fat and water separated T₁ and T₂ maps were generated by matching images from the DIP reconstruction to the MRF dictionary. PDFF maps were calculated with the noise-robust method described by Liu, et al¹¹. Phantom values were compared to inversion recovery T₁, spin echo T₂, and gradient echo T₂*/PDFF maps.

 


Results: Figure 1 shows measurements made in the ISMRM/NIST phantom, with excellent agreement between rosette MRF and reference measurements. Maps acquired using rosette cMRF in all three healthy subjects are shown in Figure 2. Table 1 summarizes the quantitative measurements in each volunteer. The average myocardial values over all subjects were T₁=592.1±17.3 ms, T₂=55.6±7.9ms, T₂*=40.4±4.2ms, and PDFF=1.41±5.36%.

Compared to previous studies at 0.55T using conventional mapping sequences^12,13, rosette cMRF yielded shorter T₁ values, similar T₂ values, shorter T₂* values, and lower PDFF values. The previous 0.55T cMRF study also observed a lower T₁ compared to MOLLI⁷. The T₂ values measured in this study are similar to both those reported in prior 0.55T cMRF studies⁷ and those obtained with T2-prep bSSFP¹². For T₂*, the SCMR recommends using echo times designed to target pathologic T₂* values, which may not always accurately capture healthy tissue values¹. A readout duration of 30 ms was targeted in this work, which is not long enough to capture the T₂* values reported for the healthy population¹² . One limitation of the technique is the long breathhold duration (24 heartbeats), as well as the lack of direct comparison to conventional mapping sequences.



Conclusion: In conclusion, this study has demonstrated the feasibility of simultaneous collection of T₁, T₂, T₂*, and fat fraction maps in the myocardium using rosette cMRF on a 0.55T system.