Sparse reconstruction of compressive sensing MRI using cross-domain stochastically fully connected conditional random...

Area of Science:

• Medical Imaging
• Radiology
• Biomedical Engineering

Background:

• Magnetic Resonance Imaging (MRI) is vital for cancer screening but suffers from long acquisition times and patient motion.
• Reducing MRI scan duration can enhance patient comfort and minimize motion artifacts.
• Compressive sensing MRI (CS-MRI) accelerates scans by sparsely sampling k-space, but requires advanced reconstruction algorithms.

Purpose of the Study:

• To propose a novel reconstruction algorithm for compressive sensing MRI.
• To improve the quality and reliability of images reconstructed from sparsely sampled k-space data.
• To address the limitations of existing reconstruction methods in CS-MRI.

Main Methods:

• A new reconstruction approach termed cross-domain stochastically fully connected conditional random fields (CD-SFCRF) is presented.
• The CD-SFCRF model integrates constraints in both k-space and spatial domains.
• A stochastically fully connected graphical model is utilized for improved MRI reconstruction.

Main Results:

• The CD-SFCRF method demonstrated strong performance in reconstructing T2-weighted (T2w) and diffusion-weighted imaging (DWI) of the prostate.
• Fine details and tissue structures were well-preserved in reconstructed images, even at low sampling rates.
• The proposed method outperformed other tested reconstruction techniques.

Conclusions:

• The CD-SFCRF framework effectively reconstructs T2w and DWI images in reduced acquisition times.
• It preserves crucial image details, showcasing its potential as a viable CS-MRI reconstruction algorithm.
• This approach offers a promising solution for accelerating MRI scans while maintaining diagnostic image quality.