Introducing Willmore flow into level set segmentation of spinal vertebrae.

Area of Science:

• Medical Imaging
• Computational Anatomy
• Biomedical Engineering

Background:

• Spinal vertebrae segmentation is vital for diagnosing spinal diseases.
• Image artifacts like noise, gaps, and inhomogeneity complicate accurate segmentation.
• Existing methods struggle with incomplete or degraded image data.

Purpose of the Study:

• To develop an accurate spinal vertebrae segmentation method.
• To address challenges posed by noisy images and missing information.
• To improve the reliability of 3-D spinal imaging analysis.

Main Methods:

• Introduced a novel level set segmentation framework.
• Incorporated an edge-mounted Willmore flow for local geometry capture and smoothing.
• Utilized a prior shape kernel density estimator for incorporating prior knowledge and handling missing data.

Main Results:

• Achieved high accuracy with Dice similarity coefficient of 89.32±1.70%.
• Demonstrated strong performance with Hausdorff distance of 14.03±1.40 mm.
• Showcased excellent inter- and intra-observer agreement (92.11-94.94% and 3.32-3.80 mm).

Conclusions:

• The proposed method significantly enhances spinal vertebrae segmentation accuracy.
• The approach is robust in segmenting images with noise and missing information.
• This technique offers a reliable tool for clinical applications in spinal disorder studies.