A Hybrid Deep Learning Framework with Explainable AI for Lung Cancer Classification with DenseNet169 and SVM

AI can match human-level accuracy in medical imaging: Both the DenseNet169 and SVM models achieved 98% accuracy in classifying lung cancer from CT scans, demonstrating that deep learning systems can perform reliably for this critical diagnostic task.

Explainability is essential for clinical adoption: The researchers integrated two explainability tools—Grad-CAM (which visualizes which regions of CT scans the AI focuses on) and SHAP (which explains which features drive predictions)—recognizing that doctors need to understand why the AI makes its recommendations before trusting it in patient care.

Addressing real-world data challenges improves fairness: The study used Focal Loss to handle class imbalance in the dataset, meaning the system was specifically designed to perform fairly across different diagnostic categories (Normal, Benign, and Malignant) rather than favoring the most common cases.

Hybrid approaches can enhance both accuracy and interpretability: By combining multiple techniques (attention mechanisms, multi-scale feature analysis, and SVM models), the researchers demonstrated that responsible AI doesn't require choosing between high performance and explainability—both can be achieved together.

Public datasets enable reproducible, responsible AI development: Using the publicly available IQOTHNCCD dataset allows other researchers to verify these findings and build upon this work, supporting transparency and accountability in medical AI development.