Integration and evaluation of several droplet imaging quantification methods for a compact fluorescent digital bead imaging device

Digital droplet analysis divides a liquid biopsy sample into twenty-five thousand of nanoliter beads. Specific molecular recognition and amplification reactions are spaced into each droplet. To achieve sensitive molecular detection, we need to group droplets by different fluorescent colors and count the number of sorted beads. To expedite the entire molecular omic data analysis, we use Roboflow to label positive/negative droplets, and train the labeled images through several YOLOv5 models. We are developing and integrating an efficient droplet identification method into a point-of-use Raspberry Pi bead imaging device for counting the copy number of specific gene expression in biofluids. We particularly focus on the specific gene expression on extracellular vesicles of a malignant brain tumor.