Whole brain susceptibility mapping using compressed sensing.

Area of Science:

• Medical Imaging
• Biophysics
• Computational Neuroscience

Background:

• Susceptibility mapping from image phase is often limited by unstable filter inversions in specific k-space regions.
• This instability leads to artifacts, particularly streaking, in the resulting susceptibility maps.
• Existing methods like direct thresholding and regularization struggle to fully address these k-space instabilities.

Purpose of the Study:

• To introduce and evaluate a compressed sensing-based approach for improving susceptibility mapping.
• To compensate for ill-conditioned k-space regions that hinder direct filter inversion.
• To compare the performance of compressed sensing against direct threshold and regularization methods.

Main Methods:

• Utilizing compressed sensing to reconstruct and stabilize data in unstable k-space regions.
• Applying the method to both simulated and in vivo datasets for validation.
• Comparing artifact levels and image quality against direct threshold and regularization techniques.
• Investigating the depiction of iron-rich regions and white/gray matter contrasts.

Main Results:

• Compressed sensing significantly reduces streaking artifacts in susceptibility maps compared to the direct threshold method.
• The compressed sensing approach demonstrates superior performance over regularization-based methods.
• The method effectively estimates only the ill-conditioned k-space regions, unlike regularization which affects the entire spectrum.
• Susceptibility maps show well-defined iron-rich regions and good contrast at white/gray matter interfaces.

Conclusions:

• Compressed sensing offers a robust solution for overcoming k-space inversion instabilities in susceptibility mapping.
• This technique improves the accuracy and quality of quantitative susceptibility mapping (QSM).
• The method enables better visualization of iron content and subtle susceptibility variations in the brain, supporting neuroscientific research.