An Enhanced YOLOv5-seg Based Digital LAMP for Polydisperse Droplets

Digital nucleic acid amplification technology plays a pivotal role in medical diagnostics. Polydisperse droplet-based nucleic acid detection may be compromised by factors such as varying droplet sizes, impurities and uneven lighting. This paper proposes a enhanced YOLOv5-seg model for polydisperse droplets. The Shuffle Attention mechanism is introduced into the network to improve feature extraction. Wise-IoU (WIoU) is used as the bounding box regression loss function, improving both convergence speed and detection accuracy. Pruning techniques are used to optimize and streamline the model. The enhanced model accurately distinguishes between positive and negative droplets and pillars in complex scenarios, with a mean average precision (mAP) of 97.7% and a 3.3ms reduction in detection time. This improved model offers a highly efficient and cost-effective solution for real-time nucleic acid detection in droplet-based systems.