Quantitative effects of using compressed sensing in dynamic contrast enhanced MRI

Area of Science:

• Radiology and Medical Imaging
• Biomedical Engineering
• Quantitative MRI

Background:

• Dynamic contrast-enhanced MRI (DCE-MRI) requires rapid imaging for quantitative modeling, often compromising image quality.
• Compressed sensing (CS) MRI offers a potential solution to improve temporal or spatial resolution and signal-to-noise ratio.

Purpose of the Study:

• To evaluate the accuracy of pharmacokinetic parameters extracted from DCE-MRI data reconstructed using compressed sensing.
• To assess the performance of CS in DCE-MRI across various undersampling factors (2x, 3x, 4x) in pre-clinical and clinical studies.

Main Methods:

• Retrospective undersampling of fully sampled DCE-MRI data in Fourier space by factors of 2, 3, and 4.
• Reconstruction of undersampled data using compressed sensing (CS) MRI techniques.
• Extraction and comparison of pharmacokinetic parameters (Ktrans) between CS-reconstructed and fully sampled data.

Main Results:

• High voxel-level concordance correlation coefficients for Ktrans were observed between 2x accelerated and fully sampled data (0.92 for mouse, 0.90 for human).
• Concordance decreased with higher acceleration factors (0.79 for 3x, 0.64-0.70 for 4x).
• Mean errors in tumor Ktrans remained relatively low, particularly at lower acceleration factors (e.g., 1.8% for mouse, -4.2% for human at 2x).

Conclusions:

• Compressed sensing (CS) MRI is an effective method for reconstructing undersampled DCE-MRI data.
• CS allows for improved image quality in DCE-MRI without significant loss of quantitative accuracy, especially at lower acceleration factors.
• The findings support the use of CS with reduced acquisition schemes to enhance DCE-MRI performance in research and clinical settings.