AI-optimized evaporation control in digital microfluidics for enhanced biochemical reaction stability

Evaporation poses a significant challenge in digital microfluidics (DMF) systems, especially when microdroplets are exposed to air. Evaporation can increase substance concentration, resulting in distorted detection outcomes and even causing cell apoptosis. In our research, we systematically evaluated the effects of various factors on droplet evaporation rates, including path length, encapsulation (gap vs. encapsulated), reagent type, temperature, airflow, humidity, and positioning within an incubator. Our findings demonstrate that under optimized conditions (37 °C temperature, 0 m/s wind speed, 90 % humidity, and an encapsulated chip), the evaporation rate can be reduced to 1/105 of the rate observed under less favorable conditions (65 °C temperature, 2 m/s wind speed, 50 % humidity, and a gap-type chip). Building on these insights, we propose an AI-optimized evaporation control approach that uses deep learning (DL) models for real-time droplet area detection and regulates evaporation through intelligent replenishment strategies, including rapid and precise replenishment. Our results show that rapid replenishment enhances the accuracy of lysine detection by 5 times compared to ignoring evaporation. Additionally, Normal human dermal fibroblast (NHDF) cells exhibited normal growth for 4 days with precise replenishment, whereas cells lacking such measures did not survive beyond 20 h. We believe our approaches provide a crucial advancement in the stable maintenance of biochemical reaction systems and precise detection in DMF, offering valuable guidance for overcoming evaporation challenges in micro-reaction applications.