Optimized Hough Circle Transform for Automated Microparticle Counting in Microfluidic Platforms

Accurate identification and enumeration of microscopic particles are important for microfluidic analysis, electrokinetic studies, and microscopy-based characterization of microfabricated systems. This study presents an optimized Hough Circle Transform (HCT) workflow for automated particle detection, sizing, and counting. Gold interdigitated electrode arrays (IDEAs) were fabricated on wafer substrates to generate electroosmotic flow, and 3 μm and 5 μm polystyrene microbeads were used as model particles. The final workflow incorporates parallelized multicore parameter optimization and composite statistical metrics based on detection accuracy and frame-to-frame standard deviation, enabling a small manually counted calibration set to be converted into locked detection parameters. In the final validation workflow, 10 manually counted calibration frames were used to optimize HCT parameters for each of four scenarios, and the locked parameters were then validated on 50 new frames per scenario (200 validation frames total) with two independent annotators. Mean validation success rates were 85.1% for 3 μm beads, 90.0% for 5 μm beads, 86.1% for 3 μm beads in mixed suspensions, and 87.2% for 5 μm beads in mixed suspensions, corresponding to object-level error rates of 17.9%, 10.9%, 19.9%, and 13.7%, respectively. Compared with the historical Generation I serial workflow, the optimized workflow reduced parameter-selection time from 24-48 h to 1-2 h, and the runtime image-processing time was approximately 45 ms per frame during offline analysis. These results show that parameter optimization is essential for robust HCT-based particle enumeration and that the workflow provides a practical analytical tool for microfluidic device characterization and electrokinetic experiments.