Metric-based Automatic Quality Assessment for Quantitative susceptibility mapping in UK Biobank study

Motivation: Quantitative susceptibility mapping (QSM) is susceptible to motion artifacts, leading to potentially unreliable data in large-scale studies like the UK Biobank. Goal(s): The primary goal is to develop an automated, machine-learning-based method to efficiently detect and assess motion artifacts in QSM images, thus enhancing data reliability for large cohort studies. Approach: A machine learning classifier using image metrics was subsequently developed and tested on the UK Biobank QSM dataset to automate the detection and assessment of motion artifacts based on the validated metrics. Results: The model effectively differentiated between motion grades, significantly reducing manual review requirements and improving the dataset's overall quality Impact: This automated, metric-based quality assessment approach for QSM QC evaluation has proven to be efficient and effective to reduce review rate in cohort study