An Optimized Neural Network Model to Classify Lung Nodules from CT-Scan Images

The early identification of lung nodules in chest X-rays is vital for human life and can prevent health emergencies. Manual prediction of lung nodules is consistent, and at early stages of lung cancer, they cannot be predicted, so an artificial intelligence system is required to identify lung nodules at the early stage. So many researchers have worked on lung nodule prediction and classification by machine learning and deep learning, but the models implemented could be more robust and consistent. So, we have proposed a novel approach to detect lung nodules early using customized CNN. It can easily segment the small nodules in classification. And we used a kernel regularizer to avoid overfitting. This model was implemented on the LIDC-IDRI dataset from Kaggle with 25,000 samples. Finally, we got an accuracy of 0.951, with calculated precision, recall, and F1-score. With this, we can confirm that our model is consistently performing.