Multi-object Model-Based Multi-atlas Segmentation Constrained Grid Cut for Automatic Segmentation of Lumbar Vertebrae...

Area of Science:

• Medical Imaging
• Computational Anatomy
• Radiology

Background:

• Accurate segmentation of lumbar vertebrae is crucial for diagnosing spinal conditions.
• 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 and evaluate an automatic multi-object model-based segmentation method for lumbar vertebrae.
• To achieve precise separation of the five lumbar vertebrae from each other in CT images.
• To validate the method's performance against manual segmentations.

Main Methods:

• A two-step approach combining affine atlas-target registration-based label fusion and bone-sheetness assisted multi-label grid cut.
• Utilizing a multi-object model and multi-atlas segmentation strategy.
• Employing constrained grid cut for automatic separation of individual vertebrae.

Main Results:

• The proposed method achieved a high average Dice coefficient of 93.9% ± 1.0%.
• An average symmetric surface distance of 0.41 mm ± 0.08 mm was obtained, indicating excellent boundary accuracy.
• The method demonstrated effective automatic separation of the five lumbar vertebrae.

Conclusions:

• The developed method provides accurate and robust automatic segmentation of lumbar vertebrae from CT images.
• This technique offers a significant advancement for quantitative analysis in spinal imaging.
• The high performance suggests potential for clinical application in diagnosis and treatment planning.