Axisymmetric diffusion kurtosis imaging with Rician bias correction: A simulation study.

Area of Science:

• Neuroimaging
• Diffusion Tensor Imaging
• Biomedical Engineering

Background:

• Standard diffusion kurtosis imaging (DKI) is susceptible to noise-induced signal variations.
• Axisymmetric DKI offers improved robustness due to a reduced parameter space.
• The impact of Rician noise bias on axisymmetric DKI at low signal-to-noise ratios (SNR) was previously unknown.

Purpose of the Study:

• To evaluate the estimation accuracy of axisymmetric DKI combined with Rician bias correction (RBC).
• To compare the accuracy of axisymmetric DKI with RBC against standard DKI.
• To assess the performance across five axisymmetric DKI tensor metrics (AxTM) in varying tissue microenvironments.

Main Methods:

• A noise simulation study using synthetic data with varied fiber alignment was conducted.
• In-vivo data from white matter was analyzed to assess estimation accuracy.
• The five axisymmetric DKI tensor metrics (AxTM) were investigated: parallel/perpendicular diffusivity and kurtosis, and mean kurtosis.

Main Results:

• Rician bias correction (RBC) significantly improved accuracy for parallel AxTM in highly to moderately aligned fibers.
• Axisymmetric DKI without RBC showed slightly better performance for perpendicular AxTM compared to with RBC.
• The combination of axisymmetric DKI with RBC demonstrated superior performance across all five AxTM in white matter, and improved accuracy in tissues with low fiber alignment.

Conclusions:

• Axisymmetric DKI combined with RBC enables accurate DKI parameter estimation at very low SNRs ($$) in white matter.
• This approach holds potential as a valuable tool for neuroscience and clinical research, particularly when scan time is limited.
• The utilized tools are accessible through the open-source ACID toolbox for SPM.