Deep Learning and Hybrid Deep Neural Network to Quantify Agglutination in Dual-Phase Lab-in-Tubing Microfluidics

In this letter, a hybrid deep neural network is developed to quantify agglutination assays in dual phase lab-in-tubing microfluidic system. The system generates 9674 droplet images over five orders of magnitude of analyte concentration, from 0.0128 to 5000 μg/mL. A comparative analysis of five different deep learning models was completed on an already existing dataset. From that, a hybrid deep neural network is formulated based on DenseNet121 and Xception by using the parallel concatenation stacking technique yielding an accuracy of 99.24%.