Generalized Deep Learning Model QQ-F Enables OEF Mapping Across Diverse MRI Acquisitions Without Retraining

A new deep learning model (QQ-F) maps oxygen extraction fraction from a single MRI sequence without protocol-specific retraining, showing significantly higher lesion-to-normal tissue contrast than prior method QQ-NET in dementia patients with different echo time acquisitions.

• QQ-F uses a feature extraction unit that derives QQ model-related features instead of raw signals, enabling generalization across acquisition schemes.
• Trained on 26 ischemic stroke patients and tested on simulations, 24 multiple sclerosis, and 30 dementia patients; in dementia data with differing TEs, QQ-F significantly improved lesion contrast compared to QQ-NET.
• The model eliminates the need for retraining, enhancing clinical scalability for oxygen extraction fraction mapping.

Automated summary

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