Bi-planar image segmentation based on variational geometrical active contours with shape priors

Area of Science:

• Medical Imaging
• Computer Vision
• Computational Anatomy

Background:

• Image segmentation is crucial for analyzing medical images.
• Challenges exist in segmenting poorly contrasted or incomplete anatomical structures.
• Existing methods often require re-initialization, adding complexity.

Purpose of the Study:

• To develop an advanced image segmentation model using active contours.
• To enhance segmentation robustness by incorporating prior shape information.
• To improve handling of challenging regions in medical imaging.

Main Methods:

• A level set approach is employed for image segmentation.
• A variational formulation integrates four energy terms, including prior shape information.
• The model avoids classical re-initialization using a signed distance function property.
• Euler-Lagrange equations are solved for functional minimization.

Main Results:

• The proposed model demonstrates efficiency and robustness across synthetic, reconstructed, and real radiographic images.
• Quantitative evaluations confirm the significant impact of prior shape information on segmentation accuracy.
• The method successfully segments anatomical structures in challenging low-contrast and incomplete regions.

Conclusions:

• The active contour model with integrated prior shape information offers superior image segmentation performance.
• This approach is particularly effective for medical imaging tasks with difficult-to-segment structures.
• The findings highlight the value of incorporating anatomical priors in segmentation algorithms.