Effect of Attention Mechanism on the Performance of YOLOv8n Model in X-Ray Image Recognition of Pneumonia

Pneumonia is a common and serious respiratory disease that affects millions of people worldwide. With the rapid development of deep learning techniques, numerous studies have focused on using deep learning models to recognize pneumonia from X-ray images. However, research on using YOLOv8 models with added attention mechanisms for this task remains limited. This paper investigates the impact of different attention mechanisms on the performance of the YOLOv8n model in recognizing normal lung, viral pneumonia and COVID-19 pneumonia from X-ray images. Six attention mechanisms—CBAM, CPCA, DA, ECA, GAM, and SE—are integrated into the YOLOv8n model, along with a baseline model. The models are evaluated on training and test datasets using metrics such as F1-score, accuracy, precision, and recall. The results show that the addition of the attention mechanism improves the performance of the model in general, among which, on the balanced test set, ECA has the highest f1 score of 0.92. The accuracy of ECA model on both training set and test set reaches 88.46%.