Dual-beam frequency-division multiplexing for high-throughput single-pixel microscopy

Imaging flow cytometry (IFC) is a pivotal technology for comprehensive single-cell analysis, integrating the statistical power of traditional flow cytometry with the spatial resolution of quantitative microscopy. However, conventional IFC systems are constrained by the inherent throughput limits of single-axis interrogation architectures. While the advent of frequency-division multiplexing (FDM) in single-pixel microscopy has effectively addressed fundamental imaging speed barriers by bypassing camera frame-rate limitations, the absolute analytical throughput remains fundamentally restricted by the volume of a single microfluidic interrogation channel. To overcome this critical bottleneck, we propose and experimentally validate a dual-beam parallel FDM architecture. In the implemented system, short-pass dichroic mirrors are used for epifluorescence collection. This design elegantly divides a digitally synthesized, radio-frequency (RF)-tagged excitation comb into two parallel and independent optical arms. By doing so, the system enables the simultaneous interrogation of two separate microfluidic channels, effectively doubling the overall throughput without sacrificing the ultrafast sub-millisecond scanning speeds characteristic of FDM systems. Experimental results demonstrate that the dual-path system exhibits robust and highly consistent imaging capabilities across both branches. The architecture achieves an ultrahigh line-scan rate of 0.7 MHz and a diffraction-limited lateral resolution of approximately 3 µ m, and demonstrates a maximum blur-free flow velocity of 2.31 m/s, structurally supporting a peak theoretical throughput of up to 154,000 events per second (eps) via dual-beam parallel acceleration. Also, through an optimized multi-frequency driving and digital flat-field correction strategy, the linear signal-to-noise ratio (SNR) is significantly improved by 54 % . Crucially, this system inherently captures pixel-level spatially co-registered multimodal signals—morphological light-loss and biochemical fluorescence—thereby entirely bypassing the need for computationally expensive elastic registration algorithms. Ultimately, this highly scalable and modular framework supports independent multiresolution configurations, establishing a robust hardware foundation for the next generation of ultra-high-throughput, multiparametric IFC.