Self-adaptive weighted level set evolution based on local intensity difference for parotid ducts segmentation.

Area of Science:

• Medical Imaging
• Computational Anatomy
• Radiology

Background:

• Parotid ducts (PDs) are crucial for diagnosing parotid lesions.
• Accurate segmentation of PDs in Cone beam computed tomography (CBCT) images aids parotid gland pathological analysis.
• Existing level set methods (LSMs) struggle with noise, intensity inhomogeneity, and vague boundaries in PD segmentation.

Purpose of the Study:

• To propose a novel Self-adaptive Weighted level set method via Local intensity Difference (SWLD) for precise and adaptive PD segmentation.
• To address the limitations of existing LSMs in segmenting PDs from CBCT images.
• To improve the accuracy and automation of parotid duct segmentation.

Main Methods:

• Developed a new adaptive weighted operator using local intensity variance difference for automatic segmentation.
• Incorporated local intensity mean difference into the energy function to enhance curve evolution efficiency.
• Utilized a local similarity factor with dual neighborhood sizes to mitigate noise, intensity inhomogeneity, and boundary blur.

Main Results:

• The proposed SWLD algorithm achieved a mean Dice score of 91.3% and a mean Hausdorff distance (HD) of 1.746.
• SWLD demonstrated superior performance compared to existing LSM algorithms on the same dataset.
• The algorithm accurately and automatically segmented PDs, even with complex gradient boundaries.

Conclusions:

• The SWLD method offers a significant improvement over existing LSMs for parotid duct segmentation.
• The proposed approach provides accurate and automatic segmentation of PDs in CBCT images.
• SWLD effectively handles challenges like noise, intensity inhomogeneity, and vague boundaries.