Multiscale Opening of Conjoined Fuzzy Objects: Theory and Applications

Area of Science:

• Medical imaging
• Computer-assisted diagnosis
• Image processing

Background:

• Accurate separation of conjoined vessels is crucial for diagnosing conditions like intracranial aneurysms.
• Existing methods for artery/vein separation often lack precision or robustness.

Purpose of the Study:

• To establish theoretical properties of a novel multi-scale opening (MSO) algorithm.
• To introduce an extension of the MSO algorithm for separating conjoined objects with differing intensity properties.
• To evaluate the algorithm's application in artery/vein separation for pulmonary CT imaging and carotid vessel segmentation in CT angiograms (CTAs).

Main Methods:

• The MSO algorithm combines fuzzy distance transform (FDT) and fuzzy connectivity.
• It iteratively opens conjoined objects from large to finer scales.
• Applications include pulmonary CT artery/vein separation and carotid vessel segmentation in patient CTAs.

Main Results:

• The algorithm achieved high average accuracy (96.3%), sensitivity (95.1%), and specificity (97.5%) in patient CTA data.
• Reproducibility was high, with 94.2% average agreement between two independent users.
• Segmentation required 25-35 seeds and averaged 30 minutes per CTA using a custom interface.

Conclusions:

• The MSO algorithm demonstrates a robust and accurate method for separating conjoined vessels in medical imaging.
• Its high accuracy and reproducibility support its clinical utility in diagnosing vascular diseases.
• The algorithm offers a significant advancement in medical image segmentation for vascular applications.