In abdomen imaging, several problems such as motion and B₀ field inhomogeneity hinder accurate diagnosis and cause imaging artifacts. Often, the fat signal is separated based on techniques originating from the Dixon method¹. The gating methods² or fast imaging methods³ with single breath-hold are used to overcome the motion artifact. Herein, to acquire water/fat fraction and T1, T2 quantification simultaneously without these limitations, a dual TR FISP with random rotating golden angle radial acquisition while free breathing was proposed.

Figure 1 shows the proposed algorithm. In the data acquisition step, dual TR pattern with alternating in-phase TR (4.6 ms) and out-phase TR (6.9 ms) and varying sinusoidal flip angle pattern were applied with FISP (unbalanced SSFP) acquisition scheme which is robust to B₀ field inhomogeneity⁴. Random rotating golden angle radial trajectories⁵ were also applied to grant incoherency of undersampling artifacts and motion robustness. A total of 600 highly undersampled (16 spokes for each images, reduction factor 8) 2D images were acquired in 3T Siemens scanner (Tim Trio) with 2x2mm² resolution, TE=TR/2, 256x256 mm² FOV, 5 mm slice thickness and 1560 Hz/Px bandwidth for a scan time 54 sec/slices.

To generate water/fat signal combined dictionary, fat's T1, T2 and Δφ_fat (phase difference between in-phase TR and out-phase TR due to chemical shift of fat) were pre-determined with an assumption that the fat's T1, T2 and Δφ_fat values are uniform in the subject. Fat dominant region mask was generated from the difference between the average phase images acquired from in-phase TR and out-phase TR as shown in Fig. 1. An average fat signal was acquired using the fat dominant region mask, then T1, T2 and Δφ_fat of the subject's fat were pre-determined. Finally, water/fat signal combined dictionary was generated based on a water signal dictionary applying complex sum of the fat signal evolution which was pre-determined. The water/fat dictionary has water fraction : [0:0.05:1], water T1 : [500:25:1980, 2000:200:3000] ms and water T2 : [5:5:145, 150:50:250] ms.

To validate the proposed method, single breath-hold (exhale) multi-echo GRE was acquired with TE : 2.2, 3.0, 3.8 ms, TR : 6.1 ms and FA : 18˚ with same FOV and resolution from the above. For water/fat separation, a graph cut based method⁶ was applied. Furthermore, an additional breath-hold experiment was performed using the proposed dual TR FISP pulse sequence to validate the motion robustness of random rotating golden angle trajectories. The data was collected during 16 breath-hold (exhale) states (breath-hold duration 600 TRs=3.4 sec) which were synchronized with the pulse sequence. All experiments were performed with a healthy volunteer (IRB-approved) in both axial and sagittal slices.

Figure 2 and 3 shows the results of the proposed method for each axial and sagittal slices. Water/fat fraction maps from the proposed method are in good agreement with the fraction maps from the multi-echo gradient echo data in both axial and sagittal slices. Signal evolutions in point 'x' shows that there are no meaningful difference between the signal evolution pattern of breath-hold and free breathing. This signal evolution results and comparison results shown in Figures 2 and 3 between breath-hold and free-breathing shows that the random rotating golden angle trajectories can successfully reduce the motion artifacts due to the respiratory motion. Furthermore, ROI-averaged T1 and T2 values in the liver from the proposed free-breathing method are T1 : 726 ms and T2 : 52 ms, which are in good agreement with previous known T1, T2 values in liver³.

The proposed method based on the random rotating golden angle radial acquisition and dictionary matching method can successfully suppress motion artifacts arising from respiration. However, further analysis for non-periodic or hard-breathing respiratory motion is required, due to all experiments being performed with a healthy volunteer in this study.

Streaking artifact due to strong flow signal from the aorta was shown in the results. Additional flow suppress preparation technique may also required.

We have presented a simultaneous water/fat separation and T1, T2 quantification method using dual TR FISP with random rotating golden angle radial trajectories while free breathing in the abdomen. The method would be useful for patients who cannot maintain breath-hold and also for quantification research in the abdomen.