Improving intra-voxel incoherent motion MRI quantification using wild bootstrap

Wild bootstrap resampling technique was proposed to improve parameter estimations of intra-voxel incoherent motion (IVIM) MRI, i.e. diffusion fraction (f), diffusion (D) and pseudo-diffusion (D*), without increasing scan time. It was verified via simulation and clinical scan. In simulation, estimation accuracy and uncertainty obtained from asymptotic fitting with and without wild bootstrapping were investigated and compared. Four patients with nasopharyngeal carcinoma (NPC) underwent IVIM scan. IVIM parameter maps generated with and without wild bootstrapping were compared. Simulation results showed that wild bootstrap method considerably improved estimation accuracy and uncertainty of D* at low and medium signal-to-noise ratios (SNRs). For in vivo IVIM parameter mapping, wild bootstrapping notably reduced the pixels associated with extremely high values and invalid estimates in D* map, hence better revealed the underlying true tissue heterogeneities. In conclusion, the proposed wild bootstrap method could considerably improve D* mapping at clinically achievable SNR without increasing scan time, promising for better IVIM quantification in clinical practice.