Combined surface and volume processing for fused joint segmentation.

Area of Science:

• Medical imaging
• Computer-aided diagnosis
• Orthopedics

Background:

• Segmentation of rheumatoid joints from CT images is challenging due to bone density variations and irregularities.
• Existing methods often yield suboptimal results for shoulder joint segmentation.

Purpose of the Study:

• To present a novel method for accurate segmentation of rheumatoid shoulder joints from CT images.
• To address the limitations of current techniques in segmenting severely deteriorated shoulder joints.

Main Methods:

• Extraction of a joint-encircling loop using minimum curvature from a surface model.
• Connecting loop intersection points with CT slices via a path search algorithm.
• Iterative refinement of segmentation using the Hough transform.

Main Results:

• Achieved high accuracy with median Dice coefficients of 98.9% for humerus and 98.5% for scapula.
• Demonstrated low median Hausdorff distances of 3.06 mm (humerus) and 3.92 mm (scapula).
• Segmentation completed in under 2 minutes for severely deteriorated shoulder joints.

Conclusions:

• The novel method accurately segments severely damaged rheumatoid shoulder joints.
• Combining surface curvature, limited user interaction, and Hough transform refinement is effective.
• The approach satisfies application criteria for rapid and precise joint segmentation.