Personalized Lung Cancer Risk Assessment : A Data – Driven Predictive Approach

Lung cancer remains a leading global health challenge, with early detection being crucial for improving patient outcomes. Artificial intelligence (AI) has shown promise in cancer risk assessment, existing models lack multimodal data integration and clinical interpretability, limiting real-world applicability. This study presents a novel hybrid AI framework that combines 3D deep learning for CT radiomics with clinical risk factors (e.g., smoke history, etc) to enable more accurate and personalized lung cancer prediction using XGBoost for clinical data and 3D ResNet-50 for imaging feature extraction, followed by explainability techniques (SHAP, Grad-CAM) to ensure transparency in model decisions. Validated on the LIDC-IDRI dataset (1,018 CT scans), the model achieves an AUC of 0.92, outperforming unimodal approaches. Key findings demonstrate that integrating radiomic and clinical data significantly enhances predictive performance while providing interpretable insights for clinicians.