Good quality of single droplet image acquisition in imaging flow cytometry (IFC) is crucial for a wide range of biological analyses. Recently, there have been significant advances in droplet microfluidic data analysis; however, acquiring blur-free single object images is still a great challenge because of the tradeoff between high flow rate and hardware setup complexity and cost. State-of-the-art hardware setups for blur-free single image acquisition are often complex, cumbersome, and not portable, limiting their suitability for point-of-care diagnostics. Moreover, motion blur and duplicate droplet image acquisition can occur with flow rate variation. To address these issues, this paper proposes a lightweight imaging pipeline for acquiring blur-free single droplet images for portable applications; this pipeline is capable of acquiring every single droplet image. While most of the existing literature focuses on complex hardware setups, utilizing high frame rate cameras that are not cost effective and complex optical solutions for droplet focusing, our pipeline utilizes minimum hardware and a lightweight algorithm for detecting, counting, and acquiring single object images from the video stream. The proposed pipeline was evaluated experimentally using videos of fast-moving droplets in which the input fluid flow rate was as high as 67.7 mu L/min. The proposed pipeline achieves 100% counting accuracy on the tested videos and 2 ms, 25 ms and 10 ms processing time for each droplet on a desktop PC, single-board computer Raspberry Pi-4, and Nvidia Jetson Nano, respectively. This yields a maximum of 500, 40, and 100 blur free detected droplets per second (DPS), respectively. The Jetson Nano implementation, achieving 100 DPS with processing time of 10 ms, is faster than existing similar studies and fast enough for the target application. The results suggest that the proposed lightweight pipeline is suitable for efficient single object image acquisition in IFC on an embedded portable platform.
