Scalable quality control on processing of large diffusion-weighted and structural magnetic resonance imaging...

Area of Science:

• Medical Imaging
• Data Science
• Neuroscience Research

Background:

• Large-scale medical imaging datasets require rigorous quality control (QC) to prevent erroneous conclusions and ensure effective machine learning model training.
• Current quantitative outlier detection methods for QC are insufficient to identify all data processing errors.
• Scalable visual inspection of data processing outputs is needed for comprehensive QC.

Purpose of the Study:

• To design and implement an efficient quality control (QC) pipeline for large-scale diffusion-weighted and structural magnetic resonance imaging datasets.
• To enable consistent QC management across research teams.
• To facilitate rapid visualization and aggregation of QC results.

Main Methods:

• Development of a novel QC pipeline with features for team-based management and efficient data visualization.
• Comparison of the proposed visual QC method against an automated QC approach using T1-weighted MRI data.
• Inter-rater variability assessment across multiple data processing pipelines.

Main Results:

• The developed QC pipeline allows for low time and effort costs in team settings.
• Experiments demonstrated high agreement among human raters during visual QC.
• Minor discrepancies were observed when comparing the proposed method to automated QC.

Conclusions:

• The proposed visual QC pipeline effectively streamlines the quality control process for large medical imaging datasets.
• The method ensures data integrity and consistency across research teams without compromising QC quality.
• Efficient visual QC is achievable, balancing necessary rigor with practical time constraints in research.