Liver DCE-MRI Registration in Manifold Space Based on Robust Principal Component Analysis

Area of Science:

• Medical Imaging
• Computational Anatomy
• Biomedical Engineering

Background:

• Dynamic contrast-enhanced magnetic resonance (DCE-MR) imaging is crucial for liver diagnostics.
• Intensity variations from contrast agents challenge traditional registration methods for DCE-MR liver imaging.
• Accurate registration is essential for analyzing temporal changes in liver DCE-MR data.

Purpose of the Study:

• To develop a robust registration framework for liver DCE-MR time series that addresses intensity variations.
• To improve the accuracy and reliability of image registration in the presence of contrast agent-induced intensity changes.
• To preserve the topological structures of enhancing tissues during registration.

Main Methods:

• A manifold-based registration framework is proposed, assuming liver DCE-MR time series lie on a low-dimensional manifold.
• Registration is achieved by decomposing large deformations into series of small deformations along geodesic paths on the manifold.
• Robust principal component analysis is employed to separate motion from contrast-induced intensity changes, guiding the registration process.

Main Results:

• The proposed manifold-based method effectively reduces motion artifacts in liver DCE-MR imaging.
• The framework successfully preserves the topology of contrast-enhancing structures.
• Quantitative and visual assessments demonstrate improved registration performance compared to traditional methods.

Conclusions:

• The manifold-based registration framework offers a robust solution for liver DCE-MR imaging challenges.
• This approach enhances the accuracy of analyzing dynamic changes in the liver.
• The method holds promise for improving diagnostic capabilities using DCE-MR imaging.