Comparison of Deep Learning on Four-Class Classification About Lung Cancer
Ruibin Wang · 2024
Image classification is an important task that involves categorizing images into different classes. In the digital age, deep learning models have shown tremendous potential in image classification. By leveraging deep learning models for image classification, we can save a significant amount of time and manpower, thereby enhancing business productivity and efficiency. This study focuses on the classification of lung medical images using deep learning models. We selected a dataset consisting of normal lung CT images, lung adenocarcinoma CT images, large cell carcinoma CT images, and squamous cell carcinoma CT images, aiming to accurately classify these images using deep learning models. We chose several classical deep learning models for our classification experiments, including VGG16, VGG19, ResNet50, and ViT models. These models have been widely applied in the field of image classification and have demonstrated exceptional performance. By comparing the performance of these models, we observed certain differences in their performance in the lung medical image classification task. Specifically, the VGG16 model exhibited a high accuracy rate of 88.9%, demonstrating its potential in lung disease classification. The ResNet50 model performed slightly better than the VGG16 model, showcasing improved classification capabilities. The most impressive performance was observed in the ViT model, which outperformed the others in this study. The ViT model exhibited high accuracy and excelled in metrics such as precision, recall, and Fl score. The research findings reveal the immense potential of deep learning models in the classification of lung medical images. This provides valuable insights for the diagnosis and treatment of lung diseases.