Quality Evaluation and Multicenter Comparison of MR Scanners Using Standard Breast Diffusion-Weighted Imaging Phantom

Abstract

Purpose: The purpose of this study was to compare apparent diffusion coefficient (ADC) value, spatial resolution, and signal-to-noise ratio (SNR) of diffusion-weighted imaging (DWI) obtained from nine 3-T magnetic resonance (MR) scanners from six institutions, using standard breast DWI phantoms for quality evaluation of quantitative DWI. Materials and methods: Nine MR scanners from six institutions were evaluated using a standard breast DWI phantom (QalibreMDTM). The DWI scan protocol was standardized as much as possible, given the constraints of using multiple scanner platforms. DWI was performed before and after phantom repositioning using three b-values (0, 800, 1200 s/mm2), single-shot or readout-segmented echoplanar imaging (EPI). In addition, further DWI was performed using two b-values (0 and 1000 s/mm2) immediately before repositioning in order to assess the effect of different b-value combinations. We assessed artifacts, fat suppression, SNR, and ADC values. We also evaluated circularity and areas of variable-sized pattern holes on the spatial resolution plate. Repeatability and reproducibility were evaluated using a Bland–Altman plot, within-subject coefficient of variation (wCV), intraclass correlation coefficient (ICC), agreement index (AI), and repeatability coefficient (RC). Results: Each ADC value obtained from test tubes with variable polyvinylpyrrolidone concentrations was matched with the corresponding reference ADC value. Mean ADC values showed no significant difference (p = 0.456) across three vendors and across different institutions. ADC values were not significantly different between two groups of b-value combinations (set 1 as test b = 0, 800, 1200 vs. set 2 b = 0, 1000, p = 0.826: and set 2 vs. retest, p = 0.525). Temperatures at the time of acquisition were 20℃ (n = 5) and 21℃ (n = 4), and mean ADCs were not significantly different according to temperature (p = 0.262). SNRs were significantly different according to institutions and vendors (p < 0.001 for both). As b-values increased, SNR decreased. ADC repeatability between test and retest was excellent, with an ICC of 0.98 (95% confidence interval [CI] 0.97, 0.99) and wCV of 8.1% (95%CI 6.2, 11.6 %). However, ICC was 0.23 (95%CI 0.13, 0.42) and wCV was 113.2% (95%CI 87, 161.6%) for SNR repeatability, indicating poor repeatability. SNR differed significantly within the same vendor (vendor B, p < 0.05). In terms of spatial resolution, there was a significant difference according to the location from the nipple (anteroposterior direction) and central plate (right-to-left direction). On the spatial resolution plate, circularity of the posterior pattern holes was better than that of the anterior pattern holes (p < 0.001) and circularity of central pattern holes was better than that of the peripheral pattern holes (p < 0.05). Circularity differed statistically significantly within the same vendor, for both the axial and the sagittal set (p < 0.05). Conclusion: In the multicenter setting, mean ADC can be measured with excellent repeatability and reproducibility using a standardized DWI protocol and a standard phantom. SNR may be affected by b-values and vendors, while spatial resolution, including circularity and area, is not affected by vendor, but varies by the location within the breast coil. In our multicenter test using standard DWI phantom and a standard acquisition protocol, ADC values and spatial resolution was not affected by vendors, while SNR was affected by vendors and b-values.