Motivation: GRASP MRI has been adapted for quantitative T1 mapping and combined with deep learning reconstruction to improve image quality, acceleration rates, and reconstruction speed. Extending this to quantitative T2 mapping holds great clinical potential. Goal(s): In this work, we present an extended version of this technique, called DeepGrasp-T2 mapping, for accelerated quantitative T2 mapping. Approach: DeepGrasp-T2 employs a combination of self-supervised learning reconstruction and GPU-accelerated parallel fitting, incorporating a novel low-rank subspace-assisted strategy to enhance image quality and accelerate training speed. Results: DeepGrasp-T2 allows for efficient and accurate T2 mapping. Impact: This work proposed DeepGrasp-T2, a self-supervised learning based approach that allows for rapid and accurate T2 mapping without requiring reference images for network training, offerring potential for different clinical applications such as prostate T2 mapping.
Pei et al. (Tue,) studied this question.