Automatic Global Level Set Approach for Lumbar Vertebrae CT Image Segmentation.

Area of Science:

• Medical Imaging
• Computer-Aided Surgery
• Image Segmentation

Background:

• Accurate vertebrae segmentation is crucial for image-guided minimally invasive spine surgery.
• Current methods struggle with vertebrae CT image challenges like topological variation, irregular boundaries, and noise, necessitating human intervention.

Purpose of the Study:

• To develop an automatic global level set approach (AGLSA) for robust lumbar vertebrae CT image segmentation.
• To address limitations of existing methods in handling complex image characteristics and noise.

Main Methods:

• Proposed an automatically initialized level set function (AILSF) using hybrid morphological filter (HMF) and Gaussian mixture model (GMM) for precise initial contour generation.
• Introduced a regularized level set formulation with region correlation of histograms to prevent weak boundary leaking.
• Employed a gradient vector flow (GVF) based edge-stopping function for fast convergence and to prevent oversegmentation.

Main Results:

• The AGLSA demonstrated superior accuracy in segmenting lumbar vertebrae CT images with irregular boundaries.
• The approach proved robust against various levels of salt-and-pepper noise.
• Quantitative comparisons validated the effectiveness of the proposed method on 115 patient CT volumes.

Conclusions:

• The developed AGLSA effectively automates lumbar vertebrae CT image segmentation, improving upon existing techniques.
• This method offers a reliable solution for image-guided spine surgery by overcoming common segmentation challenges.