Patient and globe motion, high susceptibility, and high spatial resolution requirements make orbits a challenging application for MR imaging. A high resolution, fat suppressed 3D T1 weighted sequence is necessary to properly discern whether contrast enhancement is within the optic nerve, the perineural space, the dural sheath, or extrinsic to both the optic nerve and sheath. While IR-prepared gradient echo sequences are typically used in such applications, recent advances in the acquisition speed and image contrast of 3D FSE based sequences have made them a viable alternative. Similarly, Dixon fat-water separation is now robust enough to provide water-only images in areas of high susceptibility. In this study, we investigate the combination of these techniques for assessing the integrity of the optic nerve and enhancement of orbital tumors.

A 3D Cube sequence [1] was modified to acquire in-phase and out-of-phase echoes for 2-point water-fat separation [2]. The echoes were acquired in separate passes for high resolution imaging. Imaging was performed using a 3.0T whole body imager and 32-channel head coil (GE Healthcare, Waukesha, WI). After gadolinium contrast injection was applied to a subject with a lesion around the orbits, three 3D T1-weighted sequences were acquired: an IR-prepared FSPGR with fat sat, a Cube with fat sat, and a Cube with 2-point Dixon. All sequences were acquired with the following parameters: FOV=16cm, matrix=256x256 (zero interpolated to 512x512), slice thickness=1mm (zero interpolated to 0.5mm), number of slices=64 (interpolated to 128). This resulted in an acquired pixel size of 0.625x0.625x1mm, interpolated to 0.32x0.32x0.5mm. The IR-prepared FSPGR was also acquired with: TE=2.3, TR=5.1, Flip=13, Bandwidth=+/-41.67kHz, TI=15. T1 Cube sequences were also acquired with: TE=16.3 (fat sat) or 18.4 (Dixon), TR=700, Bandwidth=+/-83.33kHz, ETL=24, ARC acceleration factor=2x1. Total scan times were 2:55 (IR-FSPGR), 3:26 (Cube fat sat), 6:46 (Cube Dixon).

Subject images are shown in figures 1 through 3. Overall, the full extent of the optic nerve was best visualized with the Cube Dixon technique. The IR-FSPGR best depicted the details of the nerve and dural sheath in the area immediately posterior to the orbits, but the area around the optic chiasm was more difficult to visualize due to the signal loss from susceptibility and bright vessel signal from fresh blood inflow (arrows, figure 3). The Cube Dixon also had overall higher and more consistent signal through the entire volume than the Cube with fat sat.

We have shown certain advantages for applying 2-point Dixon fat-water separation for high resolution MR imaging of the orbits. It’s possible that the benefits can be extended to the rest of face as well, as demonstrated previously with a 2D FSE acquisition in the head and neck [3]. Chemical shift dependent fat saturation pulses are not robust in areas of high susceptibility, and also risk saturation of water signal, as shown in both fat sat acquisitions. Thus Dixon water-fat separation may also offer some signal improvement by utilizing all of the acquired signal, instead of relying on signal suppression techniques

Motion management plays an important role for imaging the orbits. The subjects on this protocol are instructed to focus on a visual point placed directly in front of the eyes on the scanner bore. 3D sequences provide the thin slice, high resolution geometry required for imaging of the orbits and can also better disperse motion artifacts compared to 2D sequences, though they cannot resolve blurring due to motion. The Dixon sequence acquires the two echoes in separate passes and hence requires twice the scan time of a conventional acquisition. However, since this also results in double the number of averages and maintains the same SNR efficiency, parallel imaging acceleration could be increased slightly to reduce extension of scan time. Future improvements utilizing single-pass Dixon acquisitions with flexible echo times [4,5] may enable the desired resolution with little or no increase in scan time.