Leveraging Convolutional Neural Networks and Multimodal Imaging Data for Accurate and Early Lung Cancer Screening

Lung cancer is best treated when detected at an early stage. This work proposes a novel approach for early detection, incorporating additional images of non-cancerous lungs to the dataset on cancerous lungs. The investigated dataset, which is composed of chest X-ray and computed tomography scans, contains images of healthy lungs as well as lungs across multiple stages of cancer. In this way, the Convolutional neural network for early detection is trained to differentiate between normal tissue and cancer. The CNN architecture consists of feature extraction stages and an image classifier. Balanced cancerous/non-cancerous images are used for training to ensure that the model is not biased. Substantial experiments have been conducted, and the results prove that the model performs well for lung cancer detection. With a sensitivity of 83.3%, the suggested CNN-based method provides cutting-edge results and a specificity of 91.7% in identifying lung cancer from chest X-rays and CT scans. By leveraging the complementary information from cancerous and non-cancerous images, the model effectively captures the subtle patterns and anomalies associated with early-stage lung cancer, enabling timely diagnosis and treatment.