DEEP LEARNING-BASED LUNG CANCER DETECTION WITH VISION TRANSFORMER-BASE

Abstract

In order to lower death rates and increase patient survival, early identification of lung cancer is essential. In this study, a Vision Transformer (ViT) model for CT scan image analysis is used to propose an automated lung cancer detection system. The suggested method classifies lung CT scans into several categories using deep learning techniques, allowing for precise and effective diagnosis help. Four classes of labeled CT scan data were used to train and assess the model. To improve model performance and generalization, image preprocessing methods like scaling, normalization, and augmentation were used. Compared to conventional convolution-based methods, the Vision Transformer design improves classification accuracy by capturing global contextual information from medical images.A web application built on Streamlit is used to deploy the system, enabling users to register, upload CT scan images, and receive real-time forecasts and confidence scores. Metrics like accuracy, precision, recall, F1-score, and confusion matrix analysis were used to evaluate performance. The experimental findings show that the suggested ViT-based model performs reliably in predictions and achieves good classification accuracy. This research offers a user-friendly AI-powered platform to provide clinical decision support and early lung cancer diagnosis, while also demonstrating the efficacy of transformer-based architectures in medical image processing.