Rapid whole-brain T2* and susceptibility mapping using 3D multiple overlapping-echo detachment acquisition and missing modality synthesis embedded simulation.

Abstract

Purpose: To develop a 3D multiple overlapping-echo detachment (3D-MOLED) imaging technique, along with data generation and reconstruction strategies, for rapid whole-brain T₂* and QSM.

Methods: MOLED encoding was extended to a 3D multi-shot acquisition and combined with dual-echo blip-reversed EPI trains to simultaneously acquire T₂* and QSM signals while reducing image distortion. To enable Bloch simulation for training data generation, a deep learning-based missing modality synthesis approach was employed to produce co-registered multi-parametric templates. In addition, a pseudo-3D Bloch simulation was proposed to accelerate synthetic data generation for network training. A cohort of healthy volunteers and clinical participants were recruited to evaluate the motion robustness of the proposed method in comparison with conventional 3D-GRE.

Results: Compared to 3D-GRE, 3D-MOLED achieved significant improvements in both scan speed and motion robustness, with over 70% of scans rated as good image quality in both healthy and clinical cohorts. The missing modality synthesis approach generated high-quality 3D multi-parametric maps. Combined with the pseudo-3D Bloch simulation framework, it enabled efficient generation of paired training data with acceptable computational cost, thereby facilitating accurate quantitative mapping.

Conclusion: 3D-MOLED enables simultaneous whole-brain T₂* and QSM mapping at 1 mm isotropic resolution in 50 s, offering superior motion robustness compared to conventional 3D-GRE.