BCdetection-YOLOv8: An Improved Blood Cell Detection Algorithm Based on YOLOv8n

In clinical applications, blood cell detection is a crucial method for maintaining health and guiding treatment. In practice, detection technologies must not only ensure high accuracy but also meet the demand for rapid testing. Therefore, this paper proposes a detection algorithm based on an improved YOLOv8n model, named BCdetection-YOLOv8. In terms of model architecture, all C2f modules in the Backbone are replaced with DynamicConv modules to reduce the computational load while maintaining strong feature extraction capabilities. Regarding the dataset, the publicly available BCCD dataset is selected, which includes three labels: white blood cells, red blood cells, and platelets. Compared to the original YOLOv8n, the proposed BCdetection-YOLOv8 model shows a 1.7% increase in Precision, a 0.1% increase in mAP50, a 13.3% reduction in Params, and a 17.3% reduction in FLOPs. The results demonstrate that the model significantly enhances its lightweight nature while maintaining accuracy, making it highly relevant for blood cell detection tasks requiring high real-time performance.