This work targets those interested in quantitative and free-breathing abdominal MRI.Quantitative T₁ and T₂ mapping is extremely challenging in the abdomen due to the deleterious effects of respiratory motion. Recently, an MR Fingerprinting (MRF) technique has been introduced for quantitative abdominal imaging, which can provide simultaneous T₁ and T₂ quantification in a single breath-hold¹. While this technique holds great potential for lesion detection and characterization in abdomen, it still requires ~19 sec to acquire a 2D slice and thus is not realistic for routine clinical applications that typically require volumetric coverage with a preference for a completely free-breathing scan. Here, an extension of the technology to provide a navigated free-breathing 3D acquisition with volumetric coverage is presented.

The acquisition scheme of the proposed 3D abdominal MRF pulse sequence is shown in Fig. 1. The 3D encoding was performed sequentially through partitions and the data for each partition were acquired within a cycle comprised of 16 segmented acquisition blocks. This is similar to a 2D MRF method previously proposed for cardiac imaging². Each acquisition block was preceded by a navigator module, a magnetization preparation module and a fat saturation module. Within each acquisition block, 48 uniform-density spiral interleaves were acquired in 48 TRs with a FISP readout and variable flip angles ranging from 4° to 15°. TR was minimized and held constant at 5.84 ms to maximize the number of images acquired. As in the original MRF implementation, a high in-plane acceleration factor of 48 was used, so only one spiral interleaf was acquired for each partition within a volume. At the end of each cycle, a pause of 5 sec was applied for the recovery of the spins to the equilibrium. This acquisition through all partitions was repeated until all the data within each partition were accepted by the navigator. To reduce the acquisition time during this process the cycle for a given partition was not repeated once all 16 segments were accepted by the navigator. For magnetization preparation, either a non-selective inversion pulse with TI=21/100/250/400ms or an MLEV composite T₂-preparation³ module with TE=50/90ms was used. Navigator related parameters were: module type=spin-echo based cross-pair; position: dome of the left hemidiaphragm; acceptance window= ±3mm. Other imaging parameters included: FOV, 420×420 mm; matrix, 224×224; partition thickness, 3mm; number of partitions, 8.

As described previously¹, a dictionary including signal evolution for all possible combinations of T₁ (100~3000 ms) and T₂ (10~500 ms) values was generated using Bloch simulations⁴. The signal intensity timecourse of each voxel within the highly undersampled volumes (R=48) was then matched to this dictionary to extract tissue properties. The proposed method was tested with three normal volunteers (M:F, 2:1; mean age, 30.3 years) and the average acquisition time for 8 partitions was ~7.8 ± 2.6 min. A 2D breath-hold scan (~16 sec) with the same FOV and matrix size (slice thickness, 5 mm) was also performed using the same flip-angle pattern and preparation modules as a comparison.

Fig. 2 presents the T₁, T₂, and proton density maps from one slice of a 3D acquisition. Quantitative maps acquired from a 2D breath-hold scan at the same slice location are also presented for comparison. The free-breathing 3D scan provides quantitative maps with no motion artifacts, which matches well with the maps from the 2D breath-hold scan.

Fig. 3 shows representative 3D T₁, T₂, and proton density maps acquired from a different volunteer. The summary of T₁ and T₂ relaxation times for multiple abdominal organs acquired with the proposed technique is presented in Table 1 and the numbers are in close agreement with the literature values^5-7.

In this proof-of-concept study, a free-breathing quantitative abdominal imaging method was developed using the MRF technique in combined with navigators, which allows simultaneous and volumetric quantification of multiple tissue properties in abdomen. While only 8 partitions were acquired in approximately 8 min with the current protocol, the method can be accelerated by undersampling data in the partition direction. In addition, presently the entire 16-segment acquisition for a same partition was repeated if any of the segments was not accepted. This can be modified to combined data acquisition of unaccepted segments from multiple partitions, which can greatly improve scan efficiency. In addition, data acquisition for MRF can also be accelerated by using simultaneous excitation of multiple slices⁸. Finally, self-navigated approaches are readily feasible and can help remove the deleterious effects of the navigator seen in the maps, on the left side of the subject.