The purpose of this study was to quantify the influence of factors of variability on apparent diffusion coefficient (ADC) estimation in the normal prostate peripheral zone (PZ).

Fifty healthy volunteers underwent in 2017 (n = 17) or 2020 (n = 33) two-point (0, 800 s/mm²) prostate diffusion-weighted imaging in the morning on 1.5 T scanners A and B from different manufacturers. Additional five-point (50, 150, 300, 500, 800 s/mm²) acquisitions were performed on scanner B in the morning and evening. ADC was measured in PZ at midgland using ADC maps reconstructed with various b-value combinations. ADC distributions from 2017 and 2020 were compared using Wilcoxon rank sum test. ADC obtained in the same volunteers were compared using Bland Altman methodology. The 95% confidence interval upper limit of the repeatability/reproducibility coefficient defined the lowest detectable ADC difference.

Forty-nine participants with a mean age of 24.6 ± 3.8 [SD] years (range: 21–37 years) were finally included. ADC distributions from 2017 and 2020 were not significantly different and were combined. Despite high individual variability, there was no significant bias (10 × 10⁻⁶ mm²/s, P = 0.58) between ADC measurements made on both scanners. On scanner B, differences in lowest b-values chosen within the 0–500 s/mm² range for two-point ADC computation induced significant biases (56-109 × 10⁻⁶ mm²/s, P < 0.0001). ADC was significantly lower in the morning (bias: 33 × 10⁻⁶ mm²/s, P = 0.006). The number of b-values had little influence on ADC values. The lowest detectable ADC difference varied from 85 × 10⁻⁶ to 311 × 10⁻⁶ mm²/s across scanners, b-value combinations and periods of the day.

The MRI scanner, the lowest b-value used and the period of the day induce substantial variability in ADC computation.