Hyper-dimensional computing for enhanced label-free particle analysis in a flow-based optical detection system

Abstract

Flow-based optical detection is a versatile analytical technique widely used in high-throughput characterization of particles in microfluidic environments. However, conventional implementations often rely on fluorescent labeling or bulky imaging hardware, which can be time-consuming, costly, and potentially harmful to cell viability. To address these challenges, label-free imaging combined with brain-inspired computational approaches have emerged as promising alternatives. In this study, we present a label-free particle analysis framework that integrates Hyper-Dimensional Computing (HDC) with an event-based imaging system for fast and accurate classification of microparticles. A proof-of-concept experiment is performed using an event-based camera to capture optical interference patterns generated by microparticles of four different sizes through a polymethyl methacrylate (PMMA) microfluidic channel. HDC is then employed in the post-processing stage to classify these event-derived patterns efficiently, with a low computational overhead. To further enhance optical diversity and improve classification accuracy, a ground-glass diffuser is introduced into the optical path. Comparative experiments across multiple ground-glass diffuser configurations show that the classification accuracy can reach up to 98.67% under the best diffuser condition. These findings demonstrate the feasibility of combining HDC and event-driven photonic detection for compact, label-free classification of synthetic microparticles under controlled experimental conditions. While the current study is limited to polystyrene beads with well-defined size differences, the proposed framework provides a basis for future investigations toward more complex biological or industrial particulate systems.