Classification of Cancerous Lung Images by using Transfer Learning

Lung cancer detection is an interesting research domain in the field of bio-medical applications. In this work we have implemented three systems to detect and classify type of lung cancer using VGG16, VGG19 and ResNet50 models. This work classifies different type of lung cancer namely large cell undifferentiated, adenocarcinoma, and squamous cell carcinoma. This work consists of passing the input lung Computed tomography images through pre-processing stage which performs several image operations and rescaling. Secondly, pre-defined transfer learning models are used to train the given data and thus to design a system to detect lung cancer. Finally, the system is tested if the image is cancerous or noncancerous. The performance of the transfer learning techniques presented in this paper are compared in terms of precision, accuracy, Area under curve, and recall. Also, we have compared the performance of these transfer learning techniques for different epochs. The performance comparison shows that Resnet 50 model is giving more accurate and precise result among the considered transfer learning techniques.