A deep convolutional neural network for efficient microglia detection

Developing fully automated methods for counting microglia cells from immunohistological images with no user-defined parameters is a significant challenge in the field. Deep convolutional neural networks (DCNN)-based models are a way to overcome many shortcomings of manual or semi-automated methods in cell detection17. Our research aim was to develop a fully automated tool for microglia detection that would be more accurate, efficient, and faster than existing approaches. We present an innovative algorithm for the automatic detection of microglia based on YOLOv334—a powerful DCNN platform that can be customised to deal with a range of object detection tasks (Supplementary Information). To demonstrate the effectiveness of our microglia detection, we trained YOLOv3 using its general network architecture.