A Deformable Constraint Transport Network for Optimal Aortic Segmentation From CT Images

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Aortic segmentation from computed tomography (CT) is vital for aortic intervention planning, diagnosis, and measurement.
• Challenges in aortic segmentation include variable geometry due to disease and image transformations.
• Existing methods struggle with inaccurate property definition and inappropriate topological transformations.

Purpose of the Study:

• To propose a Deformable Constraint Transport Network (DCTN) for robust aortic segmentation.
• To address challenges of variable geometry and inter-image transformations in aortic CT images.
• To improve the accuracy and reliability of aortic segmentation for clinical applications.

Main Methods:

• Developed a Deformable Constraint Transport Network (DCTN) incorporating a deformable attention extractor, geometry-aware decoder, and optimal transport guider.
• The extractor adaptively extracts features, preserving semantic integrity and long-range dependencies.
• The decoder enhances geometric and semantic feature perception, mitigating background interference, while the guider aligns raw and curved planar reformation (CPR) images.

Main Results:

• DCTN demonstrated superior performance compared to 23 existing methods across 267 subjects and four public datasets.
• Achieved significant advantages in segmenting various aortic diseases and segments.
• Showcased improved accuracy in measuring key clinical indexes.

Conclusions:

• The proposed DCTN effectively addresses geometric variability and topological challenges in aortic segmentation.
• DCTN offers enhanced accuracy and reliability for aortic segmentation and clinical index measurement.
• This network advances the potential for precise aortic interventions through improved image analysis.