Evaluation of diffusion‐weighted MRI and geometric distortion on a 0.35T MR‐LINAC at multiple gantry angles

Abstract Diffusion‐weighted imaging (DWI) provides a valuable diagnostic tool for tumor evaluation. Yet, it is difficult to acquire daily MRI data sets in the traditional radiotherapy clinical setting due to patient burden and limited resources. However, integrated MRI radiotherapy treatment systems facilitate daily functional MRI acquisitions like DWI during treatment exams. Before ADC values from MR‐RT systems can be used clinically their reproducibility and accuracy must be quantified. This study used a NIST traceable DWI phantom to verify ADC values acquired on a 0.35 T MR‐LINAC system at multiple gantry angles. A diffusion‐weighted echo planar imaging sequence was used for all image acquisitions, with b‐values of 0, 500, 900, 2000 s/mm2 for the 1.5 T and 3.0 T systems and 0, 200, 500, 800 s/mm2 for the 0.35 T system. Images were acquired at multiple gantry angles on the MR‐LINAC system from 0° to 330° in 30° increments to assess the impact of gantry angle on geometric distortion and ADC values. CT images, and three fiducial markers were used as ground truth for geometric distortion measurements. The distance between fiducial markers increased by as much as 7.2 mm on the MR‐LINAC at gantry angle 60°. ADC values of deionized water vials from the 1.5 T and 3.0 T systems were 8.30 × 10‐6 mm2/s and −0.85 × 10‐6 mm2/s off, respectively, from the expected value of 1127 × 10‐6 mm2/s. The MR‐LINAC system provided an ADC value of the pure water vials that was −116.63 × 10‐6 mm2/s off from the expected value of 1127 × 10‐6 mm2/s. The MR‐LINAC also showed a variation in ADC across all gantry angles of 33.72 × 10‐6 mm2/s and 20.41 × 10‐6 mm2/s for the vials with expected values of 1127 × 10‐6 mm2/s and 248 × 10‐6 mm2/s, respectively. This study showed that variation of the ADC values and geometric information on the 0.35 T MR‐LINAC system was dependent on the gantry angle at acquisition.


| INTRODUCTION
As MR-guided radiotherapy (MRgRT) continues to develop, additional tools are being adapted from diagnostic MRI systems to combined MRI-Radiotherapy treatment (MR-RT) systems to provide more information to clinical teams. Currently, combined MR-RT systems provide imaging in the treatment position with superior soft tissue contrast, compared to x-ray imaging, real-time tumor tracking with CINE imaging, beam gating, and the ability to perform daily adaptive radiotherapy (ART). [1][2][3][4] Diffusion-weighted imaging (DWI) is an important imaging biomarker for tumor identification and assessment of response to radiotherapy that can indicate changes in tumor function before tumor size or morphology changes appear on traditional imaging methods. [5][6][7][8][9] Changes in the apparent diffusion coefficient (ADC) of tumors were correlated with local tumor control and radiotherapy treatment outcomes. [10][11][12][13] Therefore, DWI is an excellent tool for ART, allowing a patient's treatment plan to be adapted based on improved visualization of the tumor with DWI, and quantitative information from the calculated ADC.
However, performing imaging studies to monitor treatment response requires significant dedication of imaging resources and increases the patient burden. Very few studies have been performed to monitor changes in ADC values over time and often are performed at different time points during treatment. [14][15][16] MR-RT systems allow daily imaging during treatment, including DWI, without significant changes to the standard treatment protocol. Direct application of ADC values from dedicated MRI systems to MR-RT systems may be difficult due to substantial differences in the design of the MRI components compared to conventional diagnostic systems. Therefore, the difference between diagnostic MRI and MR-RT system ADC values must be assessed before DWI can be used as a functional imaging tool for treatment response monitoring or ART.
The feasibility of DWI acquisition for a 0.35 T MR-60 Co system and for a 1.5 T MR-LINAC system was shown in prior studies. 17,18 An added challenge presented by the MR-LINAC system is the motion of the LINAC gantry around the MRI. Although there is significant magnetic and RF shielding between the MRI and LINAC subsystems, changes in the gantry angle cause changes in the imaging isocenter and spatial integrity. 19 These changes could adversely impact the accuracy of ADC values and reduce their utility for ART or disease monitoring.
In this study, we perform a phantom-based accuracy and reproducibility study of ADC values calculated from DWI MRIs acquired on clinical 1.5 T and 3.0 T MRI systems, and on a 0.35 T MR-LINAC MR-RT system with multiple gantry angles.

| ME TH ODS
DWI MRIs and ADC values were quantified using a phantom-based approach. A diagnostic 3.0 T Siemens Vida scanner (Erlangen, Germany) and a radiation oncology MR-simulator 1.5 T Philips wide bore Ingenia scanner (Amsterdam, Netherlands) were used for comparison and to confirm ADC value calculation accuracy. The same phantom was then imaged on the ViewRay MRIdian 0.35 T MR-LINAC system (Mountain View, CA, USA). 20,21
Immediately prior to imaging additional ice was added to the phantom and it was placed into a cooler with ice for transport to ensure the phantom remained at or near 0°C during imaging. The water temperature was acquired with a NIST traceable long stem digital thermometer (Thomas Scientific, precision = 0.01°C, accuracy = AE0.005°C) through the top fill port. Temperature was taken before and after all image acquisitions with the maximum temperature after image acquisition being less than 1°C for all acquisitions.

2.B | Image acquisition
As the vendor of the phantom recommended, images were acquired with diffusion-weighted echo-planar-imaging (DW-EPI). The highest b-values were reduced on the 0.35 T system due to gradient limitations. Three independent imaging sessions were performed on each scanner. The diffusion phantom was also imaged on a Philips wide bore CT scanner to provide a ground truth geometric dataset using the following parameters: coronal orientation, field of view (FOV) = 373 × 373 × 221.5 mm 3 , matrix size = 512 × 512 × 443, slice spacing = −0.5 mm.

2.B.1 | Siemens and Philips MRI systems
The diffusion phantom was imaged at 1.5 and 3 T to verify the ADC calculation methodology and ensure that the calculated values

2.C | Geometric distortion quantification
To assess geometric distortion, the CT images were taken as the ground truth. The DWI phantom includes three fixed fiducial markers, attached to the central plate of the phantom. The fiducial markers were labeled A, B, and C as shown in Fig. 1. The distances from A to B, B to C, and A to C were measured using the Philips multimodality DICOM Viewer software (Philips Medical Systems, Netherlands).

2.D | ADC value calculation
After image acquisition, the same image processing was performed on images from all three systems with in-house scripts using MATLAB 2019b (MathWorks, Natick, MA, USA). The noise floor was removed to reduce the impact of noise on the ADC calculation using the noise subtraction method proposed by Dietrich et al. 23 using the following equation: where S nc and S n are the noise corrected and noisy images, respec- where S nc (b) is the noise corrected pixel signal intensity, b is the image b-value, and S 0 is the signal intensity for b = 0 s/mm 2

| RESULTS
On CT images the distances between the fiducials were 104.5, 60, and 120.4 mm from A-B, B-C, and A-C, respectively. The average differences between fiducial distances measured on CT vs MRI are shown in Table 1. The difference in fiducial distance for different gantry angles measured on the MR-LINAC system is shown in Fig. 2.
The DWI images acquired on each machine are shown in Fig. 3.     This study had some limitations, including the phantom itself.
The phantom required the addition of ice to the water in the internal basin, this resulted in significant susceptibility-related heterogeneities. The heterogeneities produced large artifacts on 1.5 T and 3.0 T systems. However, they were greatly reduced on the 0.