A ResNet Based Pneumonia Classification Algorithm for Chest X-Ray Images

X-rayexamination is an effective and reliable method for early detection and diagnosis of pneumonia, because the spatial structure of human lung organs presented by pneumonia X-ray images is complex and the gray scale distribution is uneven, and it will take a lot of manpower and time to identify pneumonia X-ray images. This study proposes a resNet based pneumonia classification algorithm for chest x-ray images. The upgraded ResNet-18 model is trained and developed on the chest X-ray 2017 dataset. In this study, the original normalization method Batch Normalization (BN) in the ResNet-18 model is replaced with Layer Normalization (LN) to mitigate the drawbacks of BN affected by Batch Size. Furthermore, the activation function ReLu is substituted with Leaky ReLu to prevent the complete death of negative weights, retaining some negative weights and thus constructing an improved ResNet-18 model. Finally, a three-fold cross-validation method is adopted, where 67% of the data is used for training and 33% for testing. The performance of the proposed improved ResNet-18 model is evaluated on the test data, demonstrating superior binary classification performance compared to the unimproved ResNet-18 and other benchmark models (including AlexNet, DenseNet-121, MobileNet-V2, and VGG-16), with accuracy, precision, recall, Kappa, and F1 score values of 91.14%, 87.38%, 91.85%, 78.34%, and 89.13%, respectively. These results indicate that the improved ResNet-18 model can better classify pneumonia X-ray images, reducing diagnosis time for doctors and conserving medical resources.