XGBoost predicts cardiac MRI motion artifacts with AUC 0.875 in retrospective cohort

An XGBoost model using 15 clinical features predicted cardiac MRI motion artifacts with AUC 0.875 in 514 adults; SHAP showed state anxiety, breath-hold duration, and heart rate variability drove risk, with risk rising sharply when STAI-State exceeded 45.

• SHAP analysis identified STAI-State anxiety, maximum breath-hold duration, and heart rate variability as the main predictors of artifact risk.
• Non-linear threshold effects emerged: artifact risk rose sharply when STAI-State exceeded 45, and physical capacity interacted with claustrophobia.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.