Detection and Classification of Malaria Parasite Based on Improved YOLOv5 Model

At present, malaria is still one of the major threats to human health. Timely diagnosis of malaria is critical to reducing both transmission and mortality. Malaria blood smear microscopy is the standard for malaria detection, but due to the cumbersome steps of manual evaluation, this diagnostic method is time-consuming and prone to missed and false positives, even in the hands of experienced physicians. Deep learning-based neural network models have achieved great success in object detection, but state-of-the-art models have not been widely used in biological image data. In response to this challenge, we have introduced an enhanced model based on YOLOv5 to accurately identify malaria blood smear cells and detect their infected stages. Based on the original YOLOv5, CA module and BiFPN are used. The experimental results demonstrate that the enhanced YOLOv5 model achieves an average accuracy of 79.3% on the public Plasmodium vivax (malaria) infected human blood smear dataset.