Recovery of Damped Exponentials Using Structured Low Rank Matrix Completion

Area of Science:

• Medical Imaging
• Applied Mathematics
• Signal Processing

Background:

• Magnetic Resonance Imaging (MRI) parameter mapping requires high-quality data.
• Undersampled measurements pose challenges for accurate image reconstruction.
• Existing low-rank matrix completion methods can be computationally intensive.

Purpose of the Study:

• To develop an efficient structured low-rank matrix completion algorithm for image recovery.
• To improve the accuracy and reduce the computational complexity of MRI parameter mapping.
• To enable large-scale 3-D imaging applications.

Main Methods:

• Exploiting exponential signal behavior and spatial smoothness for annihilation relations.
• Constructing a structured matrix from Fourier samples and enforcing low-rank property.
• Utilizing an iterative re-weighted least squares algorithm with novel 2-D Fast Fourier Transform approximations.

Main Results:

• Demonstrated significant improvement in MR parameter mapping.
• Drastically reduced memory demand and computational complexity.
• Enabled extension of structured low-rank methods to large-scale 3-D problems.

Conclusions:

• The proposed algorithm offers a computationally efficient and effective solution for image reconstruction from undersampled data.
• This method advances the field of MRI parameter mapping and large-scale imaging.
• The algorithm shows promise for future applications in medical imaging analysis.