LAUSR

The most commonly applied model for the description of diffusion‐weighted imaging (DWI) data in perfused organs is bicompartmental intravoxel incoherent motion (IVIM) analysis. In this study, we assessed the ground truth of underlying diffusion components in healthy abdominal organs using an extensive DWI protocol and subsequent computation of apparent diffusion coefficient ‘spectra’, similar to the computation of previously described T2 relaxation spectra. Diffusion datasets of eight healthy subjects were acquired in a 3‐T magnetic resonance scanner using 68 different b values during free breathing (equidistantly placed in the range 0–1005 s/mm2). Signal intensity curves as a function of the b value were analyzed in liver, spleen and kidneys using non‐negative least‐squares fitting to a distribution of decaying exponential functions with minimum amplitude energy regularization. In all assessed organs, the typical slow‐ and fast‐diffusing components of the IVIM model were detected [liver: true diffusion D = (1.26 ± 0.01) × 10–3 mm2/s, pseudodiffusion D* = (270 ± 44) × 10–3 mm2/s; kidney cortex: D = (2.26 ± 0.07) × 10–3 mm2/s, D* = (264 ± 78) × 10–3 mm2/s; kidney medulla: D = (1.57 ± 0.28) × 10–3 mm2/s, D* = (168 ± 18) × 10–3 mm2/s; spleen: D = (0.91 ± 0.01) × 10–3 mm2/s, D* = (69.8 ± 0.50) × 10–3 mm2/s]. However, in the liver and kidney, a third component between D and D* was found [liver: D′ = (43.8 ± 5.9) × 10–3 mm2/s; kidney cortex: D′ = (23.8 ± 11.5) × 10–3 mm2/s; kidney medulla: D′ = (5.23 ± 0.93) × 10–3 mm2/s], whereas no third component was detected in the spleen. Fitting with a diffusion kurtosis model did not lead to a better fit of the resulting curves to the acquired data compared with apparent diffusion coefficient spectrum analysis. For a most accurate description of diffusion properties in the liver and the kidneys, a more sophisticated model seems to be required including three diffusion components.