Deep Convolutional Neural Networks and Transfer Learning Based Approach for Lung Cancer Detection from CT Scan Images

A recent WHO survey ranked lung cancer as the second most frequent reason for death worldwide. As a result, early detection of lung cancer has become essential, as early detection can save many lives. Modern approaches to deep learning and image processing have made the detection and treatment of lung cancer much more straightforward. The ability of convolutional neural networks (CNNs) and transfer learning-based models to identify Cancer from CT scan pictures are investigated in this research. Deep CNN models already trained, like ResNet50, MobileNetV2, and VGG19, were employed to extract the deep features. Two datasets were collected and merged to increase training and validation datasets. MobileNetV2 had a 97% accuracy rate. VGG19 had an almost 100% accuracy rate, but the Model suffered from overfitting. ResNet50 had a 78% accuracy rate, and CNN had 99% accuracy with overfitting issues. Among these networks, MobileNetV2 provided the highest level of accuracy with less overfitting compared to other models. These accuracies assist in detecting malignancies early before they cause physical side effects such as paralysis and other complications.