A study on the feasibility of active contours on automatic CT bone segmentation

Area of Science:

• Medical Imaging
• Computer Vision
• Image Segmentation

Background:

• Accurate bone segmentation in computed tomography (CT) is crucial for image-guided surgery.
• Existing segmentation methods often lack sufficient accuracy or require extensive user interaction.
• Level-set-based active contour (AC) models offer potential for automatic segmentation due to their adaptability.

Purpose of the Study:

• To evaluate the feasibility of five level-set-based active contour (AC) approaches for automatic CT bone segmentation.
• To compare the performance of these AC models against commercial software and expert segmentation.
• To introduce and validate a novel augmented Chan-Vese (Aug. CV) AC model for enhanced CT bone segmentation.

Main Methods:

• Tested five level-set-based AC models: geometric AC, geodesic AC, gradient vector flow fast geometric AC, Chan-Vese (CV) AC, and the proposed Aug. CV AC.
• Utilized both synthetic and real CT images for qualitative and quantitative evaluations.
• Compared segmentation results with standard commercial software and a medical expert.

Main Results:

• Geometric AC variants showed contrast robustness but degraded with increased noise.
• Chan-Vese (CV) AC was robust to noise but sensitive to image contrast.
• The proposed Aug. CV AC demonstrated superior robustness to both noise and contrast, outperforming commercial software on real CT data.

Conclusions:

• The Aug. CV AC model is highly suitable for automatic bone segmentation from CT images.
• This novel approach offers improved accuracy and robustness compared to existing methods.
• The Aug. CV AC model shows promise for enhancing image-guided surgical procedures.