Diffusion imaging quality control via entropy of principal direction distribution

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Medical Physics

Background:

• Diffusion MRI offers insights into white matter microstructure beyond conventional MRI.
• Diffusion MRI is prone to artifacts (motion, eddy-currents) due to longer acquisition times and multiple directional acquisitions.
• Existing quality control (QC) methods in DWI or voxel domains often fail to detect all artifacts, especially those visible in derived models like DTI.

Purpose of the Study:

• To propose a novel regional QC measure for diffusion MRI data.
• To address limitations of existing QC methods in detecting residual artifacts.
• To improve the accuracy and reliability of diffusion tensor imaging (DTI) studies.

Main Methods:

• Developed a regional QC measure based on the entropy of the principal diffusion direction (PD) distribution within DTI data.
• The proposed PD entropy is invariant to patient position.
• Quantified scattering and spread of PDs to identify artifacts causing directional bias.

Main Results:

• The PD entropy method reliably detects residual artifacts, particularly those from scanner table vibrations.
• Low entropy values indicate clustered PDs, signaling dominant direction artifacts.
• High entropy values suggest uniform PD distribution, indicating good data quality.

Conclusions:

• The novel regional QC measure using PD entropy complements existing methods for diffusion MRI data.
• This approach enhances the detection and potential correction of artifacts in DTI studies.
• The method is valuable for general quality assessment in diffusion MRI research, improving neuroimaging findings.