A Hybrid Explainable AI Framework for Early Lung Cancer Detection Using CTGAN-Augmented Clinical Data, Gene Biomarkers, and Transformer-CNN Networks

Abstract

Due to delayed diagnosis and restricted access to early screening, lung cancer continues to be a major cause of cancer-related death. In order to improve early lung cancer detection, this study suggests a hybrid AI-driven diagnostic system that integrates transformer-CNN-based deep learning, synthetic data generation using CTGAN, and gene expression profiling. The Kruskal-Wallis statistical approach is utilized to identify important gene biomarkers, while CTGAN is employed to address class imbalance and enrich the dataset. A new explainable AI architecture is created to accurately classify patient outcomes by combining a bespoke CNN with a Pyramid Vision Transformer (PVT). The suggested model achieves 98.93% accuracy with full explainability via GradCAM, outperforming conventional classifiers. The findings show that there is a great deal of promise for better clinical oncology diagnosis and individualized care.