Fast and accurate Slicewise OutLIer Detection (SOLID) with informed model estimation for diffusion MRI data.

Area of Science:

• Medical Imaging
• Neuroscience
• Biophysics

Background:

• Diffusion MRI is essential for characterizing tissue microstructure.
• Image artefacts, particularly subject motion, challenge accurate diffusion MRI analysis.
• Existing voxelwise outlier detection methods struggle with slicewise intensity errors after image registration.

Purpose of the Study:

• To develop an automated tool (SOLID) for detecting outlier slices in diffusion MRI data before geometrical transformations.
• To integrate data uncertainty information from SOLID into diffusion MRI model estimation frameworks.
• To improve the robustness and accuracy of diffusion MRI analysis by addressing slicewise artefacts.

Main Methods:

• Development of SOLID, an automated tool for slicewise outlier detection using a straightforward intensity metric.
• Implementation of a framework to incorporate SOLID's uncertainty information into diffusion MRI model estimation.
• Validation through comprehensive simulations and analysis of neonatal and in vivo human diffusion MRI datasets.

Main Results:

• SOLID demonstrates high sensitivity and specificity in detecting outlier slices in simulations, outperforming other methods.
• The SOLID-informed estimation framework effectively downweights uncertain measurements, improving iterative algorithm convergence.
• Validation on neonatal data confirmed SOLID's ability to identify problematic slices in large population studies.

Conclusions:

• SOLID provides an effective and efficient solution for detecting slicewise outliers in diffusion MRI data prior to transformation.
• The integration of uncertainty information enhances the reliability and accuracy of diffusion MRI model estimation.
• This approach supports more robust characterization of diffusion processes in challenging datasets, including large-scale studies.