Image quality assessment model for 3D ASL in children

Motivation: Arterial spin labeling (ASL) is vulnerable to motion and off-resonance, which may result in unstable image quality, particularly in children. Goal(s): To propose an automatic image quality assessment model for 3D ASL in children. Approach: The proposed model was trained and validated on 51 3D ASL scans from children, and was compared to a previously developed reference method. The performance was evaluated using AUC and 5-fold cross-validation tests. Results: The proposed model yielded 8%-11% higher AUC, accuracy, and F1 score compared to the reference method. Impact: The proposed model for assessing image quality in children's ASL may offer a valuable tool for clinics and studies on brain development with a large cohort.