Introduction

In recent years, MR-guided radiation therapy (MRgRT) has gained growing interest since the introduction of MR-Linac as this modality can potentially guide on-board treatment replanning and tracking target movement in real time However, MR sequences commercially available today are limited for these goals. For example, 4D MR sequences for pre-beam motion characterization is based on a fixed contrast weighting that may be suboptimal for resolving different organs-at-risk (OARs) and target, separate sequences to provide multiple contrast weightings are often misaligned because of different motion compensation strategies during acquisition, and real-time imaging during beam-on is based on a 2D acquisition that lacks 3D information and also suffers from the drawback of single contrast weighting^1-4. To overcome the above limitations and improve treatment precision, we present an integrated adaptive MRgRT imaging framework, named multi-task MR (MT-MR), technique to generate multi-contrast, volumetric, motion-resolved or real-time images for both adaptive radiation planning pre-beam replanning scan and enabling treatment beam-on real-time monitoring from a single “pre-beam” scan. This MT-MR technique was first developed on a clinically 3T scanner. However, some commercially available MR-Linacs are at lower field strengths. In this work, we also present the initial results and then adapted to a low-field system.

Method

Sequence design:
As shown in Fig. 1, each repetition time (TR) of our multi-task (MT-) MR sequence starts with a saturation recovery (SR) pulse followed by a group of FLASH readouts during T1 recovery; Then a lower flip angle (FA) FLASH for signal recovery (Benefits shown in fig. 3); and a T2 Prep followed by additional FLASH readouts. On 0.55T, two gradient echoes are applied between FLASH RF pulses for Dixon-based water-fat separation. K-space is sampled with a Cartesian spiral-in pattern (Fig. 1B) with 10 lines in each spiral arm and the 10th line being the k-space center and used as the navigator. This sampling pattern is used for 1) denser k-space center coverage and 2) reduced eddy-current effects before the center line is sampled ^6,7. FAs and length of each segment within TR are optimized for both the 3T and the 0.55T system.
Image reconstruction workflow:
Before treatment, a “pre-beam” scan is conducted to generate the volumetric, multi-contrast motion-resolved images for adaptive planning and retrospective real-time images. Using MR Multitasking⁵ (MT), In this case,

A = U x Phi

A is the tensor representing a multidimensional image A(r ,t_motion, t_contrast) with r being the physical dimensions [x,y,z], and t_motion, t_contrast being the temporal dimensions. The combined temporal factor Phi is determined from Bloch-constrained low-rank tensor completion and motion binning from the navigator data. In the case of real-time imaging Phi can be extracted directly from SVD of the navigator data.

With a learned U from retrospective real-time application of the “pre-beam” scan, only navigator data (center k-space line) is needed during treatment to generate the real-time Phi and thus the multi-dimensional images. Furthermore, the real-time generated Phi can be projected onto a Bloch simulated subspace to “freeze” the real-time image to a specific contrast while preserving real-time respiratory motion.

To make the real-time monitoring more robust, we calculate an adjustment matrix R of size (Ncoils x Ncoils) from the time-averaged signal of the “pre-beam scan” and the first 20 seconds of the treatment scan to account for any inter-scan coil settings change and reshimming.

Experiments
Ten healthy volunteers at 3T (MAGNETOM VIDA, Siemens Healthineers) and 3 volunteers at a whole body 0.55T system (prototype Magnetom Aera, Siemens Healthineers, Erlangen, Germany) equipped with high-performance shielded gradients (45 mT/m amplitude, 200 T/m/s slew rate) were included in the study. Following imaging parameters were used: On Both 3T and 0.55T: FOV = 256(SI) x 358(LR) x 256(AP) mm3, resolution = 1.6 x 1.6 x 3.2mm3. On 3T: Water Excitation, TR/TE = 6.0/2.13 ms, total scan time = 10 mins.On 0.55T: TR/(TE1,TE2) = 9.4/(1.54,7,71)ms, total scan time = 14 mins.

Results

Figure 3 demonstrates the use of low-FA between T1 and T2 acquisition phases is advantageous over the use of gap in providing high-quality images as well as the results of applying the coil adjustment matrix. From 3T, the volumetric, motion-resolved and multi-contrast images are shown in figure 1D, where user can choose to view any contrast, slice and motion states that are available. Real-time monitoring capabilities are demonstrated in figure 2 where not only are we able to generate the real-time images during treatment, we can also freeze at any contrast, keeping only respiratory motion during treatment. Similarly, From 0.55T, the real-time water images are shown in figure 4 and the volumetric, motion-resolved and multi-contrast images are shown in figure Figure5.

Discussion

We can demonstrated a solid capability of MT-MR as an integrated platform framework for adaptive abdominal radiation therapy on 3T where multidimensional images for adaptive planning can be generated while enabling volumetric and simultaneous multi-contrast treatment monitoring from a single “pre-beam” scan. We also demonstrated an initial experience of using the MT-MR technique on a low-field scanner. As expected, the results from 0.55T has lower SNR and susceptibility artifacts, they can be overcome by better B0 correction strategies and more advanced acquisition trajectories enabled by the high-performance gradient system on such low-field scanners.

Acknowledgements

We acknowledge grant support from the National Science Foundation (#1828736) and research support from Siemens Healthineers.