Discontinuity Preserving Liver MR Registration with 3D Active Contour Motion Segmentation

Area of Science:

• Medical Imaging
• Image Registration
• Computational Anatomy

Background:

• Conventional non-rigid image registration assumes smooth motion, which is violated by liver respiratory motion.
• This violation compromises registration accuracy in abdominal MRI.
• Accurate registration is crucial for medical diagnosis and treatment planning.

Purpose of the Study:

• To present a novel approach, registration with 3D active contour motion segmentation (RAMS), for improved non-rigid image registration accuracy.
• To address the challenge of discontinuity-aware motion regularization in the presence of liver sliding motion.
• To enable automated segmentation of the respiratory sliding motion interface in liver MR images.

Main Methods:

• Employed Markov random field-based discrete optimization with dense displacement sampling and self-similarity context for registration.
• Utilized a graph cuts-based 3D active contour approach for segmenting the sliding interface.
• Implemented a two-pass registration strategy with initial mask-free regularization followed by fine segmentation and masked regularization to preserve motion discontinuity.

Main Results:

• RAMS-generated motion masks demonstrated high consistency with manual segmentations on in vivo abdominal MRI data.
• Automatic segmentation masks significantly improved registration accuracy for both the proposed discrete and conventional continuous non-rigid algorithms.
• The method effectively preserved motion discontinuity across the segmented sliding interface.

Conclusions:

• Automated segmentation of the respiratory sliding motion interface in liver MR images is feasible using RAMS.
• The derived motion masks effectively preserve motion discontinuity, leading to improved registration accuracy.
• RAMS offers a promising solution for enhancing non-rigid image registration in the presence of complex motion patterns.