Automation of Fluorescence-Activated Droplet Release by Deep-Learning-Based Droplet Detector

Droplet-based microfluidics enables miniaturized, high-throughput biochemical assays but faces challenges in selective droplet retrieval, particularly after long-term monitoring. While light-induced bubble generation offers a promising, hardware-simplified strategy for releasing individual droplets from passive traps, current implementations suffer from fabrication complexity or slow-release kinetics. To overcome these limitations, we previously developed a light-responsive fluorosurfactant using fluorinated plasmonic nanoparticles (f-PNPs) that enables millisecond-scale vapor bubble formation and efficient droplet release using a 532 nm laser, without requiring integrated photothermal materials. However, automation of this approach was limited by the need for manual droplet identification and release decisions. In this work, we introduce a fully automated Fluorescence-Activated Droplet Release (FADR) system by integrating the light-triggered release mechanism with a deep-learning-based droplet detector. This AI module autonomously identifies and localizes droplets in real-time, triggering selective release based on fluorescence intensity without human intervention. The closed-loop FADR platform offers a scalable and intelligent solution for precise droplet manipulation, enabling robust, high-throughput screening workflows with minimal hardware complexity.