Detection and Segmentation Of Shape-Coded Particles Via Hough Transforms and Snake Active Contours

Shape-coded particles enable multiplexed diagnostics, offering advantages such as reduced reagent usage, enhanced accuracy through swarm-sensing, and decreased reliance on spectral barcoding. These particles exhibit the capability to stabilize nanoliter aqueous droplets in the inner annular region, facilitating signal amplification while maintaining independent particle-based reactions. The resulting endpoint fluorescence of these droplets correlates with the biomarker concentration in the patient sample, providing valuable information for disease diagnosis. Each distinct shape can be assigned to different patients or biomarkers, enabling the collection of more comprehensive information in each diagnostic test. In order to automate the process of measuring fluorescence, it is necessary to develop methods for detecting the particles and segmenting them while differentiating the shapes. In this paper, we use the circle Hough Transform, square template matching, and snake active contours for distinguishing and segmenting circle and square particles. Without using any deep learning, these methods can successfully identify and segment the particles under various conditions.