Fast reconstruction for multichannel compressed sensing using a hierarchically semiseparable solver.

Area of Science:

• Medical Imaging
• Computational Imaging
• Magnetic Resonance Imaging

Background:

• Clinical adoption of multichannel compressed sensing (CS) in MRI requires fast and accurate image reconstruction from undersampled data.
• Combining CS with SENSE parallel imaging often leads to computationally intensive iterative reconstruction methods.

Purpose of the Study:

• To introduce a novel technique for fast computation of a compact inverse model solution as an alternative to iterative CS+SENSE reconstruction.
• To address the computational demands of CS+SENSE MRI reconstruction.

Main Methods:

• Utilized a hierarchically semiseparable (HSS) solver to accurately represent the inverse of the CS+SENSE encoding matrix.
• Compared the computational efficiency of the HSS-Inverse method against state-of-the-art reconstruction techniques.
• Evaluated the method using in vivo 3T brain data with varying contrasts, resolutions, acceleration factors, and channel counts.

Main Results:

• The HSS-Inverse method achieved over a 6x speedup compared to current state-of-the-art methods with equivalent accuracy.
• Demonstrated efficient computational scaling with respect to image size for CS+SENSE.
• Showed minimal dependency of the HSS-Inverse method on the number of parallel imaging channels or acceleration factor.

Conclusions:

• The proposed HSS-Inverse method offers high computational efficiency for CS reconstruction.
• This approach facilitates real-time CS reconstruction capabilities on standard MRI hardware.
• The method is suitable for clinical applications requiring rapid image acquisition and reconstruction.