Quantitative susceptibility mapping using plug-and-play alternating direction method of multipliers.

Area of Science:

• Medical Imaging
• Computational Neuroscience
• Biophysics

Background:

• Quantitative susceptibility mapping (QSM) is crucial for neuroimaging but often suffers from artifacts.
• Existing QSM reconstruction methods have limited flexibility in incorporating advanced denoising techniques.
• There is a need for adaptable QSM reconstruction frameworks that can leverage emerging image processing algorithms.

Purpose of the Study:

• To develop a versatile plug-and-play QSM reconstruction (PnP QSM) framework.
• To demonstrate the flexibility of PnP QSM by integrating various patch-based denoisers.
• To evaluate the performance of PnP QSM in terms of accuracy and image quality, particularly in clinical datasets.

Main Methods:

• Developed PnP QSM using an alternating direction method of multipliers (ADMM) framework.
• Integrated collaborative filtering and other patch-based denoisers into the PnP QSM framework.
• Applied and compared PnP QSM against established QSM techniques using the 2016 QSM Challenge data and glioblastoma multiforme datasets.

Main Results:

• PnP-BM4D QSM showed excellent correlation and minimal bias compared to COSMOS.
• Achieved high image quality metrics (SSIM, HFEN, CC, MI) and preserved fine features.
• Demonstrated superior performance in glioblastoma datasets compared to MEDI and FANSI-TGV methods, as assessed by neuroradiologist grading.

Conclusions:

• PnP QSM is a feasible and flexible approach for quantitative susceptibility mapping reconstruction.
• The modular design allows for easy integration and testing of novel denoising algorithms.
• This technique facilitates rapid prototyping and validation of denoisers for diverse clinical applications.