Feature Guided Motion Artifact Reduction with Structure-Awareness in 4D CT Images

Area of Science:

• Medical Imaging
• Image Processing
• Computational Anatomy

Background:

• 4D CT imaging is crucial for visualizing anatomical structures over time.
• Artifacts in 4D CT can obscure details and hinder accurate analysis.
• Existing methods for artifact reduction may compromise anatomical integrity.

Purpose of the Study:

• To develop a novel method for reducing 4D CT artifacts.
• To preserve local anatomical structures during artifact reduction.
• To improve the image quality of 4D CT scans.

Main Methods:

• A data-driven regularization constraint is used for stitching two images.
• An interface seam is computed in the smoothest region of the first image.
• Sparse 3D features and Markov Random Fields guide seam matching and displacement propagation.

Main Results:

• Reduced landmark distance error from 2.9 mm to 1.3 mm on simulated data (55% improvement over registration-based methods).
• Clinical 4D CT images showed significantly fewer artifacts compared to a reference method.
• Medical experts confirmed improved image quality and reduced artifacts.

Conclusions:

• The proposed stitching method effectively reduces 4D CT artifacts.
• The method preserves local anatomical structures, enhancing diagnostic utility.
• This technique offers a significant improvement in 4D CT image quality.