YOLO-Based Microglia Activation State Detection

To enhance the precision of identifying microglia activation states, this study proposes a detection network utilizing YOLOv5, which leverages advanced deep learning techniques and convolutional neural networks (CNNs) for microglia activation state recognition. The decoupled head is seamlessly included in the Head network, accompanied by the introduction of novel feature extraction modules derived from DenseNet, namely the DenseNet-C2f module and the DenseNet-SimCSPSPPF module. Subsequently, the Wise-IoU is employed as the loss function. The network's performance is evaluated using the Microglia dataset. The experimental findings indicate that the Mean Average Precision of the enhanced network has risen from 59.6–65.6%. Additionally, the recall has increased from 56.3–71.5%. These improvements result in more effective detection performance, better fulfilling the requirements for identifying the activation state of microglia in the field of medicine.