Confidence-Controlled Local Isosurfacing.

Area of Science:

• Medical Imaging
• Computer Vision
• Computational Geometry

Background:

• Isosurface modeling of X-ray computed tomography (CT) data offers subvoxel precision but suffers from geometric distortion due to CT artifacts.
• Existing segmentation techniques are often not robust to noisy data and computationally intensive for 3D CT artifacts, leading to flawed surface geometries.

Purpose of the Study:

• To present a novel local isosurfacing framework designed to generate high-fidelity surface models from CT data.
• To address the limitations of traditional isosurfacing and segmentation methods in handling CT artifacts and data noise.

Main Methods:

• Employs a Canny edge detection approach to identify surface candidate boundary points and assess their confidence levels.
• Utilizes screened Poisson optimization, incorporating confidence terms, to fit a smooth surface to the detected boundary points.
• Includes an intuitive user interface for simplified parameter selection in confidence computation.

Main Results:

• The proposed framework successfully generates high-fidelity surface models from CT data.
• Demonstrates improved robustness against noise and artifacts compared to conventional methods.
• Experimental results validate the effectiveness of the local isosurfacing approach.

Conclusions:

• The novel local isosurfacing framework effectively produces accurate and high-fidelity surface models from CT data.
• The combination of Canny edge detection and screened Poisson optimization offers a robust solution for challenging CT data.
• The developed framework provides a significant advancement in 3D surface modeling for medical imaging applications.