Early Stage Lung Cancer Detection Using Deep Learning

This research presents an optimized approach for early-stage lung cancer detection, utilizing the VGG-16 convolutional neural network (CNN). Achieving a final accuracy of 99%, the study focuses on refining and optimizing VGG-16, a deep learning model, to distinguish between malignant, benign, and normal conditions in CT- scan images. The application of various optimization techniques, such as Gaussian blur, SMOTE (Synthetic Minority Over-sampling Technique), transfer learning, and early callback, significantly enhances the model's performance. These techniques address issues like overfitting and class imbalance, contributing to the overall robustness of the VGG-16 model. The implications of early-stage lung cancer detection extend to healthcare facilities, offering the potential for timely interventions and improved patient outcomes. The paper details the methodology, implementation insights, and performance analysis, highlighting the relevance of optimized VGG-16 in advancing medical imaging capabilities. The study categorizes CT-scan images cancerous (malignant), non- cancerous (benign), or normal across diverse datasets like IQ- OTH/NCCD-Lung Cancer Dataset, showcasing the model's adaptability to varied patient demographics. In essence, the goal is to enhance VGG-16's proficiency in detecting subtle anomalies, contributing significantly to early lung cancer detection and impacting healthcare outcomes.