Author: Tess Armstrong, Nema Bassiri, Alonso N. Gutierrez, Michael Kasper, Natalia Lutsik, Eric Mellon, Kathryn E. Mittauer, Siamak P. Nejad-Davarani, Shyam Pokharel, Suresh Rana, Hui Wang, Joseph Weygand 👨🔬
Affiliation: Department of Radiation Oncology and Applied Science, Dartmouth Health, Miami Cancer Institute, Baptist Health South Florida, ViewRay, Inc., Miami Cancer Institute, Lynn Cancer Institute, Boca Raton Regional Hospital, Baptist Health South Florida, Department of Radiation Oncology, University of Miami 🌍
Purpose: Radiation treatments on the MR-linac (MRL) enable daily acquisition of anatomical and physiological images for adaptive treatment planning. The apparent diffusion coefficient (ADC) estimated from diffusion-weighted images (DWI) is a biomarker correlating with tumor cellularity and treatment response. Assessing ADC repeatability and reproducibility across centers is essential for multi-center diffusion studies and clinical trials on low-field MRLs.
Methods: Echo-planar Imaging (EPI) based DWI was acquired using a commercial diffusion phantom with six concentrations of polyvinylpyrrolidone. Images were obtained with diffusion weighting in three orthogonal directions and two b-value sets: [0,30,60,90,120,150,180,200,500,1000] s/mm² and [0,500,1000] s/mm². Reproducibility was assessed by measuring the same phantom on four MRLs at different institutions using brain and torso MRI coils, and gantry angle set to zero. Measurements were repeated at the gantry angle giving highest distortion for each MRL. Repeatability was assessed with four measurements at a single institution. ADC values were corrected for temperature differences from 22°C. Coefficient of variation (COV) and normalized root mean squared error (nRMSE) based on deviation from reference ADC values were calculated across all vials for each setup.
Results: The COV on a single MRL with three b-values ranged 0.28%-1.44% (head coil, nRMSE=3.13%) and 0.39%-2.52% (torso coil, nRMSE =3.17%). With ten b-values, COV ranged 0.20%-2.39% (head coil, nRMSE=2.46%) and 0.06%-4.45% (torso coil, nRMSE=2.32%). Across MRLs, COV with three b-values ranged 0.29%-2.68% (head coil, nRMSE=3.69%) and 0.82%-3.08% (torso coil, nRMSE=3.78%). With ten b-values, COV ranged 0.37%-1.57% (head coil, nRMSE=3.65%) and 0.64%-3.68% (torso coil, nRMSE=3.72%). Metrics worsened with maximum gantry distortion, increasing nRMSE by 7.55% (three b-values) and 2.52% (ten b-values).
Conclusion: Our study demonstrates excellent repeatability and reproducibility of ADC values on a phantom using EPI-based DWI on low-field MRLs, consistent across b-value sets and coil types, suggesting reliable ADC measurements in patients across treatment fractions and centers.