An important consideration in the design of multi-echo gradient-echo pulse sequences employed for water-fat imaging is the trade-off between spatial resolution and echo time step. As the spatial resolution increases, the echo time step also increases and can yield to an echo time selection that degrades the noise performance of water-fat separation [1]. Time-interleaved gradient echo imaging sequences using mono-polar (non-flyback) gradients can reduce the echo time step with the disadvantage of requiring more shots and increasing thus scan time. Time-interleaved gradient echo imaging can be alternatively combined with bipolar (flyback) gradients in order to achieve reasonable echo time steps at high resolution without increasing scan time [2-5]. However, bipolar gradients are associated with known phase errors problems, which can lead to fat quantification errors [4-5]. The present work develops a methodology for acquiring bipolar time interleaved multi-echo gradient echo and for correcting the relevant phase errors.

Sequence & Phase Correction:

A bipolar time-interleaved multi-echo gradient echo sequence was implemented (Fig. 1). For phase correction purposed, a reference scan was performed before the actual measurement. It acquired the center k-spacpe line twice for echo time with both polarities is played out before the actual sequence. A 1D phase correction was then performed based on the reference scan data. [6] Furthermore, the concomitant gradient coefficients [7] are calculated based on the gradient waveforms and then corrected during the reconstruction process.

Signal model:

The signal model as described by Peterson et al. [8] was used, fitting for compelx water and fat, a common T2*, fieldmap and a complex error map term:

$$ S\left(t_n\right)=\left(\rho_w+\rho_f\sum_{m=1}^{M}\alpha_me^{i2\pi{}\Delta{}f_mt_n}\right)e^{i2\pi\psi{}t_n}e^{\left(-1\right)^ni\theta}$$

Measurements:

All studies were performed on a 3T scanner (Ingenia, Philips Healthcare). Agar based water-fat phantoms [9] with nominal fat fractions of 0, 5, 10, 15 and 100 % have been used in the phantom experiment.

Used parameters in vivo pancreas experiment: The Bi-polar TIMGRE used 2x3 echoes, voxel size = 0.9 x 0.9 x 5mm3, TR = 8 ms, TE1 = 2, dTE = 0.8;

Used parameters in vivo spine experiment: The Bi-polar TIMGRE used 2x3 echoes, voxel size = 0.9 x 0.9 x 10 mm³, TR = 8 ms, TE1 = 2, dTE = 0.8; mono-polar TIMGRE used 2x3 echoes, voxel size = 2.2 x 2.2 x 10 mm³, TR = 7.9 ms, TE1 = 2, dTE = 0.8.

A good agreement was achieved in estimating both fat fraction and T2* values in the water-fat phantoms between the monopolar and bipolar sequences (Figure 2). Figures 3 and 4 compare the fat quantification results between the two sequences in the calf muscle and the lumbar spine respectively, showing a good visual agreement between the two methods. Figure 5 shows the fat quantification results in the abdomen using the bipolar sequence. Despite the relatively low SNR of the abdominal scan, the pancreas geometry can be well delineated with the bipolar sequence.A novel methodology was developed incorporating a bipolar time-interleaved multi-echo gradient echo acquisition with a reconstruction process correcting for phase error effects. The fat quantification accuracy of the developed methodology was validated in phantoms and the feasibility of the method was shown in vivo in three different body regions. The developed bipolar sequence was compared with a standard monopolar sequence, showing comparable water-fat separation and fat quantification results, but at higher spatial resolution than the monopolar sequence. The proposed methodology therefore enables reliable fat fraction mapping at high spatial resolution and within reasonable acquisition times and might be useful in measuring fat fraction changes in smaller organs (e.g. the pancreas).