High-Precision Detection and Classification of Pneumonia Cases Based on Optimized ResNet

This paper attempts to classify images of pneumonia patients using Convolutional Neural Networks (CNN) [2]. The classification model is constructed based on a large number of real X-ray images of both pneumonia patients and healthy individuals. In this study, the model structure is optimized based on the Residual Neural Network (ResNet), resulting in an improvement in test set accuracy from 78.7% to 93.1% [4]. Additionally, a comparative analysis was conducted on CNN, ResNet, and Vision Transformer (VIT) in terms of accuracy, precision, recall, and AUC [5]. The improved ResNet achieved an accuracy of 98.5% on the training set and 93.1% on the test set, demonstrating excellent performance.