Objectives: The aims of this study were to implement and systematically evaluate the performance of a new parameter-free segmented algorithm for analysis of diffusion imaging data using a combined intravoxel incoherent motion and diffusion kurtosis imaging (IVIM-DKI) model of spin diffusion in comparison with the simpler intravoxel incoherent motion (IVIM) model.
Materials and methods: A multistep algorithm was implemented intended to separate diffusion kurtosis from IVIM effects in multi-b-value diffusion measurements using an adaptive b-value threshold technique. For each possible b-value threshold (separating diffusion and perfusion effects), diffusion kurtosis analysis of high b-values is followed by IVIM analysis keeping kurtosis parameters fixed. The b-value threshold with smallest Akaike information criterion is chosen as best model solution. The algorithm was tested in diffusion data sets of the upper abdomen from 8 healthy volunteers with 16 different b-values and compared with a standard multistep IVIM analysis.
Results: The proposed algorithm could successfully be applied to all data sets and provided a significantly better fit of the observed signal decay in all assessed organs (all P < 0.03). Using the proposed IVIM-DKI model of diffusion instead of an IVIM model had a systematic impact on the resulting IVIM parameters: The pure diffusion coefficient and the pseudodiffusion coefficient were significantly increased (P < 0.03 in all assessed organs), accompanied by a decrease in the perfusion fraction in liver, pancreas, renal cortex, and skeletal muscle (all P < 0.02). Optimal b-value thresholds separating diffusion from perfusion effects had a tendency to lower values when the IVIM-DKI model was used.
Conclusions: The proposed algorithm provides a new approach for separation of IVIM and kurtosis effects of diffusion data without organ-specific adaptation.