Nonparametric intensity priors for level set segmentation of low contrast structures

Area of Science:

• Medical Imaging
• Image Analysis
• Computational Anatomy

Background:

• Low contrast object segmentation is crucial for clinical tasks like lesion and vascular wall analysis.
• Existing methods often rely on high-level information (e.g., shape priors, generative models).
• A gap exists in effectively utilizing low-level image information for low contrast segmentation.

Purpose of the Study:

• To develop an improved segmentation technique for low contrast structures.
• To incorporate a priori intensity distributions and low-level image features into segmentation.
• To apply and validate the method for positive vessel wall remodeling analysis in cardiac CT angiography.

Main Methods:

• Incorporated a priori intensity distributions and low-level image information into a nonparametric dissimilarity measure.
• Defined a local indicator function for foreground object likelihood.
• Integrated the indicator function into a level set formulation for segmentation.
• Applied the technique to cardiac CT angiography images for vessel wall analysis.

Main Results:

• The proposed method was applied to a dataset of 25 patient scans.
• Demonstrated improved segmentation performance compared to conventional gradient-based level sets.
• Successfully segmented low contrast structures in cardiac CT angiography images.

Conclusions:

• The novel approach effectively segments low contrast structures by integrating low-level image information.
• This method shows promise for enhancing clinical analysis of vessel wall remodeling.
• The technique offers an improvement over existing gradient-based level set methods.