Multiquadric Spline-Based Interactive Segmentation of Vascular Networks

Area of Science:

• Biomedical image analysis
• Computational biology
• Medical imaging

Background:

• Interactive image segmentation tools like active contours and scribbles are common but struggle with thin vessel structures.
• Segmenting thin, faint structures in microvasculature images, such as the dura mater, is challenging for quantitative analysis due to imaging and tissue preparation issues.

Purpose of the Study:

• To introduce a novel interactive segmentation method using a sparse set of user-defined seed points for complex biomedical images.
• To develop a precise, quick, and robust segmentation approach for challenging vascular network images.

Main Methods:

• A multiquadric spline-based binary classifier was developed for interactive segmentation.
• Features used include color values and the spatial location of user-defined seed points.
• The method was tested on epifluorescence imagery of dura mater microvasculature.

Main Results:

• The proposed seed point-based method achieved fast and accurate interactive segmentation.
• Experimental results on twenty epifluorescence images demonstrated significant benefits over existing methods.
• The approach proved effective for segmenting thin vessel shapes in complex biomedical images.

Conclusions:

• A sparse set of seed points provides a precise, quick, and robust solution for interactive biomedical image segmentation.
• This method overcomes limitations of traditional tools in segmenting challenging thin structures like microvasculature.
• The developed classifier offers a unique and effective approach for quantitative analysis of vascular networks.