A Unified Framework for Chest CT Image Classification: Benchmarking Pre-Trained Architectures with Lightweight and Deep Models

Lung cancer is a common malignancy and a leading cause of cancer-related deaths. Accurate classification of lung cancer is essential for determining effective treatment options. Despite conventional methods for identifying lung cancer types, advancements in deep learning algorithms have enabled faster and more precise classification. While deep learning techniques have significantly advanced lung cancer classification, challenges such as explainability, computational complexity, and inconsistent model benchmarking remain in this fast-moving research area. We propose a novel approach to categorize forms of lung cancer by employing an architecture that integrates multiple deep learning models ResNet50, VGG16 and InceptionV3 into a unified framework for chest CT image classification. To reduce class imbalance and overfitting, our design makes use of class weighting and specialized data augmentation methods. To improve model transparency for clinical adoption, Grad-CAM was utilized for visual interpretability.