Harmonic field extension for QSM with reduced spatial coverage using physics-informed generative adversarial network.

Area of Science:

• Neuroimaging
• Medical Physics
• Artificial Intelligence in Medicine

Background:

• Quantitative susceptibility mapping (QSM) is vital for studying brain iron in development and diseases.
• Whole-brain QSM acquisition is time-consuming, limiting clinical utility.
• Limited field-of-view (FOV) QSM reduces scan times but suffers from susceptibility underestimation due to background field removal errors.

Purpose of the Study:

• To develop and validate a novel method for accurate QSM reconstruction with limited FOV.
• To overcome susceptibility underestimation artifacts in reduced FOV QSM.
• To enable faster and clinically feasible QSM for deep gray matter (DGM) and neurological conditions.

Main Methods:

• Proposed a harmonic field extension method leveraging a physics-informed generative adversarial network (GAN).
• Implemented a GAN-based approach to correct background field removal errors at the boundaries of limited FOV.
• Validated the method against conventional techniques and full FOV QSM acquisitions.

Main Results:

• The proposed method significantly reduces susceptibility underestimation in limited FOV QSM.
• Quantitative and qualitative results demonstrate performance comparable to full FOV QSM.
• The method shows improved local field results and QSM outcomes in Parkinson's disease patients and prospectively acquired data.

Conclusions:

• The physics-informed GAN-based harmonic field extension method effectively addresses susceptibility underestimation in limited FOV QSM.
• This approach enhances the clinical feasibility of rapid, high-quality QSM for brain iron quantification.
• The method shows promise for real-world clinical applications, particularly in DGM imaging and neurodegenerative diseases.