The stability of diffusion measurements with simultaneous multi-slice imaging: A multi-scanner, multi-site, test-retest study using the NIST/QIBA CaliberMRI diffusion phantom.

Abstract

This study evaluated how simultaneous multi-slice (SMS) acceleration affects the accuracy and repeatability of apparent diffusion coefficient (ADC) measurements across multiple MRI platforms using a traceable diffusion phantom. A multi-centre, multi-scanner test-retest study was conducted across seven MRI systems from three major vendors using a CaliberMRI diffusion phantom. Diffusion-weighted images with SMS acceleration factors of 0, 2, 3, and 4 were acquired. Measured ADC values were compared to temperature matched NIST reference standards. The signal-to-noise ratio (SNR) was calculated for each acquisition. Bias was measured using percentage change from reference values and multivariable regression. Lin's coefficient was used to summarise the concordance between repeat acquisitions. On average, scanners overestimated ADC by 4%. SMS factor 2 did not significantly bias measurements compared to non-SMS acquisitions. However, SMS factors 3 and 4 introduced systematic biases, with mean ADC overestimations of 4.62% and 9.22%, respectively. Biases were most pronounced in media with ADC values outside the physiological range. Despite reduced SNR at higher SMS factors, intra- and inter-session repeatability remained high, suggesting intrinsic bias rather than random noise. These findings demonstrate that higher SMS factors can bias ADC measures and degrade image quality highlighting the need for caution when absolute quantification is required. In contrast, low SMS acceleration appears suitable for reducing scan times in clinical and research protocols without materially compromising ADC accuracy across platforms. These findings have implications for the adoption of SMS in quantitative diffusion imaging protocols across platforms.