A Deep Learning based Image Processing Technique for Early Lung Cancer Detection

The world’s most common cause of cancer-related death is still lung cancer, which is mostly caused by rising tobacco use and industrialization in developing countries. Early detection combined with state-of-the-art medical facilities is essential since early diagnosis can drastically lower the death rate. In this study, we used 1,190 CT scan images from the Kaggle IQ-OTH/NCCD lung cancer dataset to test deep learning (DL) techniques. In order to identify high-risk patients, diagnose lung cancer, and predict malignancy, we developed enhanced images representing normal, malignant, and benign instances after completing thorough image preprocessing methods. This allowed for early intervention to avert long-term repercussions. We performed a thorough performance comparison within the framework of conventional Convolutional Neural Networks (CNN). Rigorous image preprocessing techniques such as affine transformation and Gaussian noise were utilized. Our approach achieved a validation accuracy of 99%, demonstrating the efficacy of our preprocessing methodology in reducing model complexity. The comparison highlighted that our suggested preprocessing method resulted in a higher F1-score of 99%, validating the robustness and accuracy of our approach.