Spatially regularized estimation of the tissue homogeneity model parameters in DCE-MRI using proximal minimization

Area of Science:

• Medical Imaging
• Biophysics
• Radiology

Background:

• Standard pharmacokinetic models like Tofts and extended Tofts in DCE-MRI lack key perfusion markers.
• Advanced models, such as the tissue homogeneity model, estimate plasma flow and permeability-surface area product but suffer from biased, uncertain voxelwise results.

Purpose of the Study:

• To improve the reliability of perfusion parameter estimates in DCE-MRI by incorporating information from neighboring voxels.
• To address the limitations of voxelwise estimation in advanced pharmacokinetic models.

Main Methods:

• Implemented spatial regularization using total variation on five perfusion parameter maps derived from the tissue homogeneity model.
• Employed proximal techniques of convex optimization to numerically solve the non-differentiable total variation problem.

Main Results:

• The proposed algorithm significantly reduces noise in estimated perfusion-parameter maps.
• Demonstrated improved accuracy, spatial consistency, and readability of perfusion maps on both numerical phantoms and real DCE-MRI data.
• Achieved these improvements without a considerable decrease in the quality of the model fit.

Conclusions:

• Spatial regularization enhances the reliability of DCE-MRI perfusion analysis using the tissue homogeneity model.
• Modern optimization techniques offer a slight increase in computational cost compared to non-regularized methods.
• The approach provides more dependable perfusion parameter estimates for clinical applications.