Accelerated MR imaging using compressive sensing with no free parameters.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Signal Processing

Background:

• Compressive sensing MRI (CS-MRI) accelerates image acquisition but requires robust reconstruction methods.
• Traditional reconstruction methods often need manual parameter tuning, limiting clinical efficiency.
• Existing iterative methods can be computationally intensive or require empirical adjustments.

Purpose of the Study:

• To introduce and evaluate a data-driven iterative soft thresholding (IST) framework for CS-MRI reconstruction.
• To demonstrate a robust and parameter-free approach for improved MRI image reconstruction.
• To enhance the clinical applicability of CS-MRI through automated, high-quality reconstruction.

Main Methods:

• Implemented a data-driven IST framework combining Nesterov optimal gradient schemes with wavelet-based adaptive denoising.
• Compared the IST method against a nonlinear conjugate gradient (NL-CG) approach with empirical tuning.
• Evaluated reconstruction performance on T₂-weighted brain and 3D phase contrast vascular data.

Main Results:

• The data-driven IST method achieved image reconstruction quality comparable to empirically tuned NL-CG.
• Statistical analysis indicated improved robustness and image quality with the IST approach.
• The IST method demonstrated a leaner implementation than NL-CG.

Conclusions:

• The presented data-driven IST framework offers a robust and parameter-free solution for CS-MRI reconstruction.
• This method shows potential for improving image quality and reliability in clinical settings.
• Automated, data-driven reconstruction is key to advancing the clinical utility of CS-MRI.