Published Paper Details:

Lung cancer, a leading cause of cancer mortality worldwide, demands early detection for improved patient survival rates, yet accurate diagnosis remains challenging due to subtle differences in imaging data between malignant and benign tissues. Traditional diagnostic methods rely on radiologists' manual interpretations of CT scans or X-rays, which can be subjective and error-prone, especially for early-stage tumors. While machine learning models like Convolutional Neural Networks (CNNs) have been applied to assist in diagnosis, they often face limitations due to data scarcity, imbalance, and difficulty in capturing complex spatial relationships within the lungs. This study proposes a hybrid deep learning approach for lung cancer detection, leveraging Generative Adversarial Networks (GANs) for generating synthetic images to expand and balance the dataset, Capsule Networks (CapsNet) to enhance spatial feature extraction, and Rotational Networks (RotNet) to incorporate rotation invariance for improved accuracy. By combining these components with ensemble learning, the system achieves more robust and accurate predictions. This multifaceted approach not only addresses existing limitations but also provides a reliable and effective tool for early lung cancer detection, supporting radiologists in timely diagnosis and improving potential patient outcomes.