Automated Lung Cancer Detection Using Convolutional Neural Networks from Chest CT Image Datasets

Abstract

Lung cancer is one of the top diseases causing the death worldwide due to very late-stage diagnosis and its high mortality. Early detection is critical for improving patient outcomes, as the survival rate significantly decreases when the cancer spreads beyond the lungs. The aim of this research is to develop an automated and efficient lung cancer system using deep learning algorithms applied to medical imaging data, including Xrays and CT scans. This study utilized the Lung Imaging Database Consortium-Image Database Resource Initiative (LIDC-IDRI) dataset, consisting of 900 images, divided into 57% for training, 35% for testing, and 8% for validation. The proposed system employs the ResNet 50 model, using the Stochastic Gradient Descent with Momentum (SGDM) optimization technique. The experimental results demonstrated that ResNet 50 consistently outperformed Inception mv4 across all metrics. ResNet 50 achieved perfect diagnostic accuracy, with minimal losses, whereas Inception mv4 showed lower accuracy and higher losses. These findings indicate that the ResNet 50 model offers superior diagnostic accuracy and efficiency for early lung cancer detection, highlighting its potential for clinical application to improve patient outcomes and reduce mortality rates.