Optimal b-values for diffusion kurtosis imaging of the liver and pancreas in MR examinations

Abstract

Purpose: The objective was to optimise the number of b-values for diffusion kurtosis imaging (DKI) of the liver and pancreas in MR examinations to ensure reliable results with the shortest possible acquisition time.

Methods: Twenty healthy volunteers underwent DKI at 3.0 T Siemens Magnetom Skyra using 7 b-values (b = 0, 200, 500, 750, 1000, 1500, 2000 s/mm²). The regions of interest (ROIs) were placed in the liver (right lobe, left lobe) and pancreas (head, tail). DKI parameters (D_app, K_app) for ROIs were calculated for 7 b-values utilising the nonlinear least-squares (NLLS) Marquardt-Levenberg algorithm. All calculations were repeated for ten subsets of data, with the number of b-values reduced to 4. DKI parameters calculated for subsets were compared with parameters calculated for all 7 b-values.

Results: The correlation coefficient between DKI parameters calculated for 7 b-values and subsets ranged from 0.65 to 1.00. The coefficient of variation (CoV) of DKI parameters calculated for a group of volunteers varied from 8% to 42% and was not affected by the reduction of the b-values number. Only one subset of data (b = 0, 500, 1500 and 2000 s/mm²) simultaneously met two criteria: no statistical difference (p < 0.05) from results obtained for 7 b-values and very good correlation with them.

Conclusions: DKI acquisition with 4 b-values (b = 0, 500, 1500 and 2000 s/mm²), compared to DKI acquisition utilising 7 b-values, allowed for the reduction of acquisition time by 36%, without affecting the calculated DKI parameters.

Keywords: Diffusion kurtosis imaging; Magnetic resonance.