Automatic Initialization Active Contour Model for the Segmentation of the Chest Wall on Chest CT

Area of Science:

• Medical Image Processing
• Computer Vision
• Computational Anatomy

Background:

• Active contours (snakes) are vital for object boundary detection in computer vision and medical imaging.
• Traditional snake methods face challenges with initialization and converging to concave boundaries.
• Gradient Vector Flow (GVF) offers a novel external force to overcome these limitations.

Purpose of the Study:

• To present an automatic initialization method for snake algorithms using Gradient Vector Flow (GVF).
• To enhance the segmentation of the chest wall in medical images, specifically computed tomography (CT) scans.
• To address the manual initialization requirement of conventional snake algorithms.

Main Methods:

• The proposed method utilizes the mean shape as an automatic initialization value for GVF snakes.
• GVF is computed by diffusing gradient vectors from an edge map of the medical image.
• The algorithm involves three phases: landmark identification, Procrustes shape analysis, and shape alignment.

Main Results:

• The mean shape initialization effectively segments the chest wall in CT images.
• GVF demonstrates a significant capture range and capability to handle boundary concavities.
• Experimental results confirm the algorithm's good performance in chest wall segmentation.

Conclusions:

• The developed algorithm provides a robust automatic initialization for GVF snakes.
• GVF-based active contours show superior performance compared to traditional methods for image segmentation.
• This approach enhances the utility of active contours in medical image analysis, particularly for complex structures like the chest wall.