Investigating Transfer Learning Models for Lung Cancer Detection from CT Scans: A Comparative Evaluation

Lung cancer remains a significant global health challenge despite advancements in medical science and technology. In 2020, it became the leading cause of cancer-related deaths, prompting a critical need for improved detection and prevention strategies. Computed Tomography (CT) imaging has shown promise in early lung cancer diagnosis due to its high resolution and 3D visualization capabilities. Leveraging the advancements in deep learning and transfer learning, this research paper conducts an extensive comparative evaluation of four prominent transfer learning models—Xception, MobileNetV2, DenseNet121, and InceptionResNetV2. A meticulously curated dataset comprising diverse CT scans with varying pathological conditions and imaging qualities forms the foundation of this study. A comprehensive performance assessment employing sensitivity, specificity, and accuracy metrics gauges the effectiveness of the models accurately. InceptionResNetV2 emerges as the frontrunner, demonstrating superior performance. MobileNetV2, Xception, and DenseNet121 display commendable performance, albeit with slight variations in their strengths across the performance metrics.