Synopsis

Keywords: Diffusion/other diffusion imaging techniques, Radiotherapy

Diffusion weighted imaging (DWI) allows for the evaluation of tumor cellularity. Its application on a 0.35 T MRI-guided linear accelerator (MRL) would facilitate integration of this information into radiotherapy planning and potentially allow for online biologically-guided plan adaption. This study demonstrates the capability of a DWI protocol both in phantom and in vivo. In particular, it is demonstrated that quantitively accurate, repeatable, and geometrically precise ADC maps can be produced in phantom on the 0.35 T MRL. Additionally, this technique was applied in vivo on one sarcoma patient receiving same-day diagnostic diffusion scans before, during, and after radiotherapy.

Introduction

Diffusion weighted imaging (DWI) allows for the noninvasive interrogation of tissue cellularity¹, which is a surrogate for cellular proliferation. Its integration into adaptive MRI-guided radiotherapy on a 0.35 T (ViewRay MRIdian) MRI-guided linear accelerator² (MRL) through the production of apparent diffusion coefficient (ADC) maps would allow for the identification of cellular subpopulations with restricted diffusion for which dose escalation strategies and/or biologically-guided plan adaptation may be clinically advantageous. In this study, the accuracy, repeatability, and geometric precision of ADC maps produced using an echo planar imaging (EPI)-based DWI protocol on the MRL system is illustrated, and its in vivo potential for longitudinal patient imaging is demonstrated in an 82 year-old female sarcoma patient receiving same-day diagnostic diffusion scans before, during, and after radiotherapy.

Methods

Images were acquired of a National Institute of Standards and Technology (NIST)-traceable diffusion phantom (CaliberMRI) containing differing amounts of polyvinylpyrrolidone (PVP) which controls the ADC in each vial. Images were acquired using a multi-slice EPI sequence (matrix size=100x100x21, FOV=350x350x190 mm³, TR=3200 ms, TE=120 ms, BW=1352 Hz, α=90°, 6 averages). The gantry of the linear accelerator was kept at 0° for all scans to minimize eddy current-induced geometric distortion. The b-values of 0, 200, 300, 500, and 800 s/mm² were used with diffusion weighting applied along each of the three principal directions. ADC maps were produced by fitting an exponential in each voxel along each diffusion direction in MATLAB Version R2021a (The Mathworks, Inc) and then averaging over all three directions. ADC accuracy was assessed by measuring ADC values in each vial at two timepoints and comparing these values to their known NIST-traceable ADC. Repeatability of the measurement was assessed by measuring the mean repeatability coefficient, averaged over each vial in the phantom. System-dependent geometric distortion was interrogated by measuring the spatial distance between 93 pairs of phantom features in both a central slice and a slice 5.4 cm off-center and comparing these distances to those measured in a spatially precise CT scan of the same phantom. This analysis was performed by three independent observers on both ADC maps taken on the 0.35 T MRL system and on a 3 T MAGNETOM Vida (Siemens Healthineers). DWI on the 3 T system was performed using the institution’s clinical protocol which utilized b-values of 100 and 1000 s/mm². Additionally, for a sarcoma patient receiving radiotherapy on the MRL, same-day in vivo ADC maps were acquired on both systems at three timepoints: prior to treatment (day 0), mid-treatment (day 21), and 3 weeks post-treatment (day 69).

Results

In the phantom experiments, ADC quantification was accurate with discrepancies beyond measurement uncertainty only seen in vials approaching free diffusion (see Figure 1). ADC quantification was constant over repeat measurements with small discrepancies only seen at high ADC. A mean repeatability coefficient of 0.068 mm²/s was measured on the MRL compared to 0.057 mm²/s on the diagnostic scanner. In the central slice, average geometric distortions of 0.35 (±0.02) mm and 0.85 (±0.02) mm were noted for the MRL and the diagnostic system, respectively. Similarly, in the slice 5.4 cm off-center, average geometric distortions of 0.66 (±0.04) mm and 2.14 (±0.07) mm were noted on the two systems, respectively (see Figure 2). In the sarcoma patient images, comparable image quality was observed on both systems at both timepoints. Additionally, the general region of enhanced ADC signal on the images acquired with the 0.35 T MRL demonstrates marked conformity with the tumor contours delineated on the anatomical True FISP (TRUFI) images and similarity with tumor ADC features visualized in the diagnostic ADC map (see Figure 3).

Discussion

Although DWI using an integrated 0.35 T MRL has been studied in the past³, until now, the accuracy and repeatability has been lacking. The work presented here demonstrates vastly superior accuracy and repeatability in quantifying ADC. Performing DWI on the 0.35 T MRL has the obvious inherent limitation of lower sensitivity than DWI at diagnostic field strengths. This limited the acquisition to lower b-values than typical diagnostic scans and slightly degraded the repeatability relative to 3 T. However, these minor limitations are greatly outweighed by the following advantages. Particularly, in radiation oncology, where geometric precision is extremely important, ADC maps produced on the 0.35 T MRL demonstrated greater than two-fold improvement in spatial precision compared to ADC maps acquired on a 3 T scanner. Additionally, the ability to acquire accurate ADC maps on the same system that the patient is being treated with is logistically beneficial, enables a more geometrically reproducible setup, and will potentially allow for on-line biologically-guided adaptation and dose escalation for personalized radiotherapy.

Conclusion

The acquisition of accurate, repeatable, and geometrically precise (sub-millimeter distortion; >2x improvement over 3T) ADC maps is possible at 0.35 T with an EPI approach. This enables tracking of longitudinal changes in tumor cellular density during MRL treatment and will facilitate personalization/adaptation not only to patient anatomy but dynamic tumor physiology as well.

Acknowledgements

*Dr. Rosenberg and Dr. Redler contributed equally as senior authors.