A novel methodology for modelling images using level-sets with application to atopic asthma fibrin networks

This paper introduces a novel methodology for image understanding, specifically focusing on binary dataset classification. This method makes use of fuzzy level-sets to decompose the images for analysis. With the growing need for interpretable image analysis techniques in medical imaging and other fields where complex models often lack transparency, the need for an interpretable method to understand images arises. The proposed approach combines fuzzy level-sets, graphical models, and a bag-of-visual-words approach to not only train a logistic regression model but also to infer how images content contribute to the understanding of images. The methodology is applied to an illustrative example, followed by an investigation of images of fibrin networks of patients diagnosed with atopic asthma. The purpose of this application is to evaluate fibrin networks in clots of platelet poor plasma derived from healthy controls and asthmatic patients.