Automatic construction of statistical shape models using deformable simplex meshes with vector field convolution...

Area of Science:

• Medical imaging
• Computer vision
• Computational anatomy

Background:

• Statistical shape models (SSMs) are crucial for robust segmentation in medical imaging.
• Establishing accurate shape correspondence between training shapes is a major challenge, especially in 3D.

Purpose of the Study:

• To develop a novel, efficient, and accurate method for establishing shape correspondence in SSM construction.
• To improve the robustness and reduce the computational cost of SSM building.

Main Methods:

• A mesh-to-volume registration technique using a greedy algorithm-based deformable simplex mesh with vector field convolution.
• An automatic shape initialization using Gaussian mixture model-based registration for improved deformable model evolution.

Main Results:

• The proposed method significantly outperforms the classical spherical harmonics (SPHARM) method in femur statistical shape model construction quality.
• The new approach demonstrates substantially higher computational efficiency than the SPHARM method.

Conclusions:

• The developed method is effective for constructing high-quality statistical shape models.
• This technique offers a more accurate and computationally efficient alternative for SSM applications.