Automatic segmentation of leg bones by using active contours.

Area of Science:

• Medical Imaging
• Image Segmentation
• Computational Anatomy

Background:

• Accurate segmentation of human leg bones in computed tomography (CT) images is challenging due to inhomogeneous object properties and close proximity of anatomical structures.
• Existing segmentation methods struggle with weak or strong boundaries and distinguishing adjacent objects.

Purpose of the Study:

• To present a new active contours model for segmenting human leg bones in CT images.
• To address limitations in segmenting complex anatomical structures with varying boundary types and close spatial relationships.

Main Methods:

• A novel active contours model utilizing a variable-weighted combination of local and global intensity information.
• Development of a level set function evolution equation where the zero level set defines the contour.
• Automatic initialization of contours using a histogram-based multiphase segmentation technique.

Main Results:

• The proposed model successfully segments human leg bones in CT images.
• Demonstrated ability to split objects with weak and strong boundaries.
• Effectively distinguished very adjacent objects, showcasing its utility in complex medical imaging scenarios.

Conclusions:

• The developed active contours model offers an efficient and accurate solution for segmenting human leg bones in CT images.
• The method's robustness in handling inhomogeneous objects and close anatomical structures makes it valuable for medical image analysis.