MRI is increasingly incorporated into the workflow of radiotherapy. Recently, a real time MRI-guided radiotherapy system (ViewRay™) has been commercially introduced, which is capable of MRI and radiotherapy simultaneously while the patient is in the treatment position. Such a system can potentially enable a scientifically and logistically feasible adaptive radiotherapy strategy, wherein the treatment plan is adapted throughout the course of therapy based on tumor response assessment based on diffusion MRI. However, the most commonly used spin echo single-shot echo-planar-imaging (SS-EPI) diffusion sequence suffers from low spatial resolution and severe geometric distortion, which is particularly problematic for radiotherapy because any geometric inaccuracy directly translates to mis-calculated radiation dose and potentially radiation target miss. The purpose of the current study was to develop a reliable, distortion-free DW sequence that is practicable for tumor response evaluation and radiation treatment re-planning.

Sequence: The DWI sequence used in this study combines a diffusion preparation module with a segmented 3D steady-state free precession (3D-SSFP) readout module, which has been previously proposed for diagnostic imaging at higher field strength (1.5T or higher) [1, 2]. The diffusion module, which was inserted at the beginning of each k-space segment, used a twice-refocused spin-echo (TRSE) with diffusion gradient to generate diffusion weighting. At the end of diffusion encoding, the magnetization is restored using a 90° RF pulse, which is followed by a spoiler. Table 1 lists the imaging protocol parameters that were used in this study.

ADC Accuracy: A commercially available diffusion phantom (High Precision Device, Inc.) was used to validate the ADC accuracy of our DW-SSFP sequence. The phantom consists of 13 vials filled with aqueous solutions of polymer polyvinylpyrrolidone (PVP) at 0%, 10%, 20%, 30%, 40% and 50% concentrations (Fig. 1a), and weighted mean apparent diffusion coefficient (ADC) for each vial was provided as a reference. Our TRSE DW-SSFP sequence was compared with the standard SS-EPI sequence on a Siemens Prisma 3T scanner. A total of twelve DW-SSFP measurements and six SS-EPI measurements with the same phantom positioning was conducted on the same/multiple days to verify the short-term/long-term reproducibility.

Geometric Reliability: The standard SS-EPI and our DW-SSFP sequences were compared against Turbo Spin Echo (TSE) for geometrical reliability in both phantom and healthy volunteer (three brain scans and three liver scans). In the phantom study, the angle between three fiducial markers was measured and compared to quantify the shape invariability. In the volunteer study, contours of the brain and liver were drawn in the three types of images, and the area differences between SS-EPI, DW-SSFP and TSE were calculated.

Tumor Detection: A pilot study was performed on the ViewRay system, which includes a tri-Cobalt 60 radiotherapy system and a 0.35T MRI system with Siemens pulse programming interface. The TSE, SS-EPI and DW-SSFP images of one liver metastasis patient, one glioblastoma (GBM) patient, and one peritoneal sarcoma patient were acquired.

ADC accuracy measurements are summarized in Table 2. Mean ADC values calculated from DW-SSFP agreed with both SS-EPI and reference values, indicating ADC reliability of the DW-SSFP sequence. Extensive sequence development and optimization are underway to further improve ADC accuracy and stability.

Phantom and in-vivo geometric reliability results are shown in Fig 1. Distortion from EPI is apparent by visual inspection. Quantitative fiducial marker angle measurements from the three type of sequences were: TSE: 89.37°±0.21°, SS-EPI: 87.57°±0.48°, DW-SSFP: 89.33°±0.20°. Absolute area difference between SS-EPI and TSE was 253.43±57.45 (mm2), and absolute area difference between DW-SSFP and TSE was 38.07±23.96 (mm2). The DW-SSFP-based contour matched well with TSE whereas the SS-EPI-based contour had large derivation from TSE (Fig. 1(d-f)). All these suggest that DW-SSFP is substantially more accurate and consistent than SS-EPI in terms of geometric reliability.

For all three cancer patients, ADC maps from SS-EPI and DW-SSFP were concordant, and tumor regions exhibited different ADC patterns compared with normal surrounding, suggesting the feasibility of detecting and observing tumor response based on our DW-SSFP. The peritoneal sarcoma case is shown in Fig. 2, where the tumor region was highlighted on both ADC maps due to surrounding muscle gave low signal given long TE and low field strength of the ViewRay. Comparing with our DW-SSFP, distortion of the SS-EPI was severe which made it unacceptable for treatment planning.

Our DW-SSFP sequence provides desirable ADC accuracy and high geometric reliability. Pilot study on ViewRay demonstrates the feasibility of quantitative longitudinal evaluation of tumor response and consequently treatment re-planning using the DW-SSFP sequence.