An automated endpoint control platform for a cyclone concentration pipette

In the point-of-care, microfluidic, and biochemistry areas, the sample volume is becoming as small as µL order. In this study, we introduce a cyclone concentration pipette that integrates quantitative concentration and sample ejection functions for liquid volume conversion (mL to µL). The concentration pipette is a three-scale polypropylene (PP) tube comprising a cm-scale cyclone concentration chamber, a mm-scale channel for keeping the 10 µL liquid, and a µm-scale orifice (105 µm ± 6 µm) for ejecting liquid. The cyclone concentration system requires continuous observation to determine the endpoint (10 µL). This work introduces an automated control system designed for precise control of a 10 µL volume. The system is built on a Raspberry Pi platform with a Logitech C920 USB camera (1920 × 1080). It monitors the tube in real-time using three parallel detection methods: a convolutional neural network (CNN) for liquid state classification, a MOG2 background subtractor for droplet movement detection, and temporal pixel variance analysis across a 40-frame rolling window for flow state monitoring. When no activity is detected by any method for 30 consecutive seconds, the Raspberry Pi automatically closes a KOGANEI proportional pneumatic valve via a PWM-to-voltage converter (0-10 V). The system automatically terminated concentration without manual intervention. For the water concentration, the mean final volume was 10.6 μL with a standard deviation of 1.65 μL, corresponding to a coefficient of variation (CV) of 15.6% and a bias of + 6.0% relative to the target volume of 10 μL (n = 10). The optimized imaging conditions improved the stability of endpoint detection and volume control, resulting in a final volume of 10.4 μL (CV = 8.2%) with a bias of + 4.0% relative to the 10 μL target, demonstrating acceptable endpoint control accuracy. For the Sunset Yellow solution, a recovery of > 90% and a concentration factor of approximately 90-fold were achieved, indicating that the automated control strategy maintained high sample recovery while achieving efficient concentration.