Compressed sensing MRI via two-stage reconstruction.

Area of Science:

• Medical Imaging
• Signal Processing
• Applied Physics

Background:

• Compressed sensing (CS) accelerates magnetic resonance imaging (MRI) data acquisition.
• Existing CS methods often result in image artifacts, especially at high reduction factors.

Purpose of the Study:

• To propose a novel two-stage reconstruction scheme for CS-MRI.
• To leverage k-space data properties and under-sampling patterns for improved image reconstruction.

Main Methods:

• The proposed method segments under-sampled k-space data into low- and high-frequency domains.
• Stage one reconstructs the low-frequency region using dense measurements.
• Stage two combines the reconstructed low-frequency data with high-frequency data for whole k-space reconstruction.

Main Results:

• The two-stage approach reduces the effective under-sampling rate and normalizes signal magnitudes.
• This leads to lower residual errors and improved reconstruction accuracy.
• Achieved 2-4 dB improvement in peak signal-to-noise ratio for cardiac cine, brain, and angiogram imaging at reduction factors up to 6.

Conclusions:

• The proposed two-stage CS reconstruction scheme effectively reduces artifacts in MRI.
• It offers a more accurate and robust method for accelerated MRI data acquisition.
• Demonstrates significant improvements over conventional CS reconstruction techniques.