MR plays an important role in evaluating the integrity of silicone implants and detecting their rupture. This is commonly performed with saturation or inversion pulses that suppress one or two of the three chemical species (water, fat, and silicone) that can be imaged with MR. However, such techniques require long acquisition times and suffer from poor signal. Chemical shift dependent saturation pulses also tend to fail in areas of high susceptibility. In this work we propose a two-step Dixon method that produces separate images of water, fat, and silicone from a triple-echo sequence without the use of suppression pulses. Moreover, this data can be acquired with a single-pass Dixon sequence [1] to quickly and efficiently acquire all the required data without excessive extension of scan time.

Fast Triple Echo Dixon (FTED) is a Fast Spin Echo technique that acquires three echoes per phase encode line with a single bipolar readout. The center echo is timed in-phase with the spin echo, while the outer echoes are timed such that water and fat are out-of-phase. To separate water from fat, the first two echoes are processed with a 2-point Dixon algorithm [2], as are the last two echoes (second and third echoes). The water images from each process can be magnitude combined to form a single water image, as can the fat images.

Since silicone has a chemical shift relatively close to that of fat, it is also out of phase with water in the out-of-phase images and hence remains in the fat image. Fat and silicone are thus also more difficult to separate. The proposed method accomplishes this in a second step by using a phase map from the first and third echoes. With an echo spacing twice that of the usual echo spacing used for water-fat separation, the outer echoes allow for twice the phase evolution between species, and can better separate fat and silicone based on chemical shift. Instead of magnitude combining the fat/silicone images from the first Dixon process, the original phase from the first and third echoes are restored to generate complex fat and silicone images with the water removed. Depending on the exact chemical shift of the silicone compound and the echo spacing achieved in the sequence, the phase evolution of silicone from the first to third echo could even be out of phase with fat. If that is the case then fat-silicone separation can be performed using conventional 2-point Dixon methods. However, we found that the chemical shift of our silicone samples to be approximately 1.2 times that of fat, and a modified, flexible TE 2-point Dixon algorithm was applied to better separate the two species, by modifying the region growing routine to detect discontinuities of 60-90 degrees in phase instead of 180 degrees.

The triple echo acquisition was implemented in a 3D Cube sequence [3], and was run on a fat/water/silicone phantom and a volunteer subject with silicone breast implants, using a 3.0T whole body scanner (GE Healthcare, Waukesha, WI) and an 8-channel breast coil. Sequence parameters were TR = 2000, TE = 100, ETL = 110, FOV = 32cm, matrix = 224x224, slices = 160, slice thickness = 1.4mm, bandwidth = ±125kHz, Dixon echo spacing = 1.232 msec, ARC acceleration = 2x1.5, total acquisition time = 4:25.

Acquired data was reconstructed with 3-species separation and evaluated. Resulting images are shown in figures 2 and 3. 3-species separation was successful in both phantom and human data.

These results demonstrate the feasibility of separating three chemical species with an efficient triple-echo acquisition and a two-step Dixon algorithm. Historically, susceptibility has been a major problem with chemical shift dependent saturation pulses in breast imaging, due to the air interfaces around the breasts and their close contact with surface coils. Hence saturation failures can often occur near areas that are prone to rupture. By avoiding saturation pulses, Dixon separation methods are able to utilize all of the signal available from the sequence, improving the overall SNR efficiency of the technique and reducing sensitivity to magnetic field inhomogeneity. While Dixon water-silicone separation may be combined with a STIR technique to provide water- and silicone-only images [4], the STIR pulse again reduces available signal and extends acquisition time. The FTED readout utilized in our study is symmetric about the spin echo and therefore maximizes the readout efficiency for a fast spin echo based Dixon technique. Since in-phase and out-of-phase information are all acquired in a single pass, a 3D acquisition can be acquired in clinically feasible scan times.