Early Lung Cancer Detection with a Fusion of Inception V3and Vision Transformers: A Binary Classification Study

Lung cancer is a significant global health issue, causing around 13 % of newly diagnosed cancer cases and roughly 20% of cancer-related deaths. This study evaluates the relative efficacy of two advanced deep learning models, Vision Transformer (ViT) and Inception V3, in the classification of lung cancer using a publicly accessible dataset of MRI scans. The ViT model achieved a high training accuracy of 98% but struggled with generalization, stabilizing at around 77% validation accuracy due to potential overfitting. In contrast, the Inception V3 model, employing a transfer learning approach with frozen base layers, achieved a consistent enhancement in validation accuracy of 85.8% with a diminished validation loss of 0.3457. Metrics such as accuracy, precision, recall, F1-score, and the area under the ROC curve are used to provide a comprehensive view of a model's effectiveness, thus better reflecting its performance across all classes. Overall, the Inception V3 model shows greater reliability in processing new input, underscoring the importance of model selection in developing effective AI-driven diagnostic tools for lung cancer.