Quantitative diffusion MRI is widely utilized in oncologic research studies and is of increasing clinical interest as a tool to predict or assess treatment response. Quantitative diffusion phantoms are urgently needed to help clinicians and researchers to compare data acquired using different scanner platforms, imaging protocols, and field strengths. Such phantoms would ideally include samples with known apparent diffusion coefficients (ADC) spanning the entire physiological range (600-2600 μm²∙s^-1) ^[1] at an easily reproducible temperature (e.g. using an ice water bath at 0°C) with a simple NMR spectrum (single peak) and Gaussian diffusion. Currently available diffusion phantoms that utilize an aqueous Polyvinylpyrrolidone solution ^[2,3] provide ADC values that cover only a portion of the physiological range when scanned at 0°C (<1120 μm²∙s^-1)^[2]. A recently proposed acetone-D₂O phantom, where D₂O modulates the ADC of acetone without producing MR signals, provides ADC values covering the entire physiological range at 0°C ^[4,5]. The purpose of this study was to test the reproducibility of quantitative diffusion measurements of the acetone-D₂O phantom across three sites with different vendors at both 1.5T and 3T, and using two different acquisition protocols.

Phantom construction and shipment: An acetone-D₂O phantom was constructed, which included 5 cylindrical vials (radius=2.75cm, height=9.5cm) filled with acetone-D₂O mixtures with D₂O concentrations of 0%, 5%, 10%, 20%, 40% (v/v). To test reproducibility across sites, vendors and platforms (Table 1), the phantom was scanned at Site 1, then sequentially shipped to and scanned at Site 2 and Site 3. At each site the phantom was stored under 0°C.To test the integrity of the phantom at the end of the study, the phantom was shipped back to Site 1 and rescanned using the same acquisition protocols.

Imaging protocol: All scans were performed with the phantom placed in an ice-water bath, where the water signal was eliminated with MnCl2 (5mM), after 30 minutes of equilibration. Cylinder vials were aligned in the S/I direction. A 2D single spin echo DW-EPI sequence was acquired in the axial plane without parallel imaging. Two different sets of b-values were applied to test the robustness of ADC measurements to protocol variations: (1) 0-2000s/mm², and (2) 0-500s/mm². Remaining acquisition parameters are listed in Table 2. Data obtained at sites 2 and 3 were transferred to Site 1 for centralized processing. At each site and field strength, each DW-EPI series was acquired twice to test the intra-exam repeatability of ADC measurements. Further, each imaging experiment was repeated in a different session to test the inter-exam reproducibility.

Diffusion quantification: For each DW-EPI scan, one slice with minimum image artifact such as EPI distortion was chosen for quantitative analysis. An ADC map was generated for this slice using a single-exponential model on a voxel-by-voxel basis. For each vial, a circular ROI (diameter=8mm) was chosen in the center of the vial while avoiding apparent artifacts. The ADC for the vial was calculated by averaging the ADC of each voxel inside the ROI. The measured ADC from all intra- and inter-exam repetitions, choice of b-values, sites, and field strengths were compared using intra-class correlation coefficient (ICC). In addition, Bland-Altman tests were performed to assess the individual effects of intra-exam repetitions, inter-exam repetitions, choice of b-values, sites/vendors, and field strengths, respectively.

ADCs were measured successfully from all datasets, with mean ADCs of 3411.4, 2686.4, 2093.7, 1337.7, 632.3μm²∙s^-1 in ascending order of D₂O concentrations. ICC was 0.9983 with 95% CI=[0.9950,0.9998] across all acquisitions. Bland-Altman analyses of ADC measurements between intra- and inter-exam repetition, field strengths and choices of b-values (Figure 1), as well as between sites (Figure 2) demonstrated good agreement. The close agreement between the ADC measured at Site 1 at the beginning and end of this project (Figure 3) validates the stability of the phantom during the entire study (37 days). The acetone-D₂O phantom was successfully built and shipped (three sites overall), scanned with similar protocols at each site, and returned to Site 1 where it was scanned again. Overall good agreement was observed across intra- and inter-exam repetition, sites/vendors, field strengths, and choices of b-values. This indicates good repeatability and reproducibility of ADC measured in the acetone-D₂O phantom over a wide range of ADCs. Some of the variations across exams and sites may be due to temperature deviations (imperfect ice water bath) or residual imaging artifacts. These factors highlight the need for careful control of the temperature and imaging setup in diffusion MRI phantom studies. In conclusion, the acetone-D₂O phantom is a promising phantom for future multi-center validation and quality assurance of quantitative diffusion MRI techniques.