Multi‐Site Quality Assurance of ADC and PDFF Measurements: A Multi‐Vendor Phantom Study

ABSTRACT Background Reliable quantitative imaging biomarkers are important for multi‐site studies. However, scanner‐ and site‐specific variations can compromise the clinical utility of many measures including apparent diffusion coefficient (ADC) and proton density fat fraction (PDFF). Phantom‐based quality assurance provides a foundation for multi‐site studies. Purpose To evaluate cross‐site reproducibility, intra‐session reliability, and accuracy of ADC and PDFF measurements across four sites. Study Type Prospective phantom study. Phantoms A NIST diffusion phantom (CaliberMRI, Model 128) and a fat phantom (Calimetrix, Model 300). Field Strength/Sequence 3 T/diffusion‐weighted spin‐echo echo‐planar imaging for ADC and 3D gradient‐echo imaging for PDFF. Assessment ADC and PDFF were measured to assess cross‐site reproducibility and intra‐session reliability. Accuracy was evaluated against reference values. Statistical Tests Intra‐session repeatability was assessed using the intraclass correlation coefficient (ICC). Intra‐session measurement variability was quantified using the coefficient of variation (CoV), and agreement between repeated measurements was assessed using Lin's concordance correlation coefficient (LCCC). NRMSE (normalized root mean square error) assessed accuracy relative to reference values. A linear mixed‐effects (LME) model assessed site differences in ADC and PDFF, with p < 0.05 considered statistically significant, while two one‐sided tests (TOST) assessed equivalence within prespecified margins of ±10% for ADC and ±5% for PDFF. Results ICC demonstrated excellent repeatability across repeated scans for both ADC and PDFF (ICC > 0.99). Mean NRMSE between measured and reference values was < 10.0% for ADC and < 7.0% for PDFF, suggesting high accuracy. Mean intra‐session CoV was < 1.0% for ADC and < 2.0% for PDFF, while LCCC exceeded 0.99 for both measurements, demonstrating strong agreement. LME models showed significant site effects for both ADC ( F (3, 36) = 6.05) and PDFF ( F (3, 33) = 5.07), while pairwise differences met the prespecified equivalence criteria for both measures. Data Conclusion Diffusion and fat phantoms provide reliable quality assurance tools for quantitative ADC and PDFF measurements across multiple MRI sites, supporting consistency despite statistically significant site effects. Evidence Level 4. Technical Efficacy Stage 1.