3D Free-Breathing Multi-echo Acquisition for Whole-body Water/Fat Separation

Combined magnetic resonance and linear accelerator systems (MR-Linac systems) are a powerful new cancer treatment modality. MR-Linac systems promise improved patient outcomes and decreased side effects compared to conventional radiation therapy (RT) systems. These systems yield exquisite soft tissue imaging, offer imaging during RT delivery, and provide a platform for adaptive RT. However, unlike traditional RT planning with computed tomography (CT) measured in Hounsfield units, the MR signal does not correlate with electron density. Electron density information is required to calculate radiation dose maps for RT planning for adaptive RT.

The MRIdian MR-Linac is a low-field system (0.35 Tesla), which is beneficial for applications in RT because it has less effect on the radiation beam than higher field systems. However, low-field MR systems have imaging challenges compared to high-field MR systems. The resonant frequencies between water and fat at 0.35 Tesla are close and traditional methods of separating these tissues (i.e., DIXON-based methods) are more difficult. Furthermore, spectral-selection of fat is not possible, which means traditional fat saturation methods cannot be used at 0.35T. Currently, neither a fat-saturation sequence nor a multi-echo sequence for fat/water separation is available on the MRIdian MR-Linac system. We propose to implement and test a fat/water separation technique optimized for 0.35T. This sequence will enable sCT generation for MR-only simulation (i.e., RT planning without CT) and adaptive RT.

The original DIXON technique for water/fat separation depends on two signal acquisitions - when the fat and water spins are in-phase and opposed-phase. New DIXON methods are more flexible and enable fat/water separation at echo times that are not directly in- and opposed-phase. At 0.35T, the fat and water spins are slow enough that the first echo (i.e., shortest echo) is a near-in-phase echo. Additional echoes will support a 3-point DIXON reconstruction and B0 mapping for inhomogeneity correction.

The long-term goal of this study is to realize the benefits of MR-guided adaptive RT to decrease toxicity and improve patient outcomes. The specific objective of this study is to develop an MR sequence on the low-field MR-Linac for fat/water separation. For the purposes of Radiation Oncology, multi-echo gradient-echo is a fast method to acquire a 3D stack with a large FOV. The images can be reconstructed using a DIXON-based method to produce multiple image types. The resulting images can be used for sCT, which could greatly assist with auto-contouring methods and adaptive planning on MR-Linac systems. These images are also diagnostically used for functional imaging, specifically Dynamic contrast-enhanced imaging (DCE-MRI), which has shown promise at low field, as well as a non-contrast method magnetic resonance angiography (MRA).

Producing these images requires chemical shift imaging. At low fields, chemical shift imaging is difficult as the spectra of fat and water are very close (52 Hz @ 0.35T as compared to 224 Hz @ 1.5T). Traditional DIXON methods use out-of-phase and in-phase echo times (TEs) to separate fat and water. At 0.35T, these TEs are 9.86ms and 19.7ms, respectively. However, long TEs degrade the signal-to-noise ratio (SNR) and lead to long imaging times, particularly for 3D stacks. In addition, B0 inhomogeneity increases and SNR degrades with longer TEs.

The hypothesis is that at 0.35T, the fat and water spins are slow enough that the first echo (i.e., shortest echo, approximately 1ms) is a near-in-phase echo. Additional echoes will support a 3-point DIXON reconstruction and B0 mapping for inhomogeneity correction. I predict that once this multi-echo gradient echo sequence is implemented on the MRIdian system, it can be used to acquire images that will successfully produce water-only, fat-only, in-phase and opposed-phase images.