A fully automatic vertebra segmentation method using 3D deformable fences

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Radiology

Background:

• Accurate vertebra segmentation is crucial for diagnosing spinal conditions from CT scans.
• Manual segmentation is time-consuming and prone to inter-observer variability.
• Automated methods are needed to improve efficiency and consistency in clinical practice.

Purpose of the Study:

• To develop a fully automatic method for vertebra segmentation in computed tomography (CT) volume data.
• To enhance the precision and reliability of separating individual vertebrae within CT images.

Main Methods:

• The proposed method constructs 3D fences using valley-emphasized Gaussian images to delineate adjacent vertebrae.
• Initial fence curves are extracted from intervertebral discs, leveraging anatomical characteristics.
• Deformable models and minimum cost path finding optimize curves, correcting errors.
• A fence-limited region growing method labels the final segmented volume.

Main Results:

• The automatic vertebra segmentation method demonstrated high success rates.
• The technique was validated on a dataset comprising 50 patient CT scans.
• The 3D fence construction effectively separated individual vertebrae.

Conclusions:

• The developed method provides a fully automatic and successful approach for vertebra segmentation in CT data.
• This technique holds significant potential for improving diagnostic workflows in spinal imaging.
• The automated segmentation offers a reliable alternative to manual methods.