Lung cancer detection and classification by CNN - Sell Project

Among other prevalent types of cancer, lung cancer is considered to be more severe. One theory is that it is not generally identified until it has spread and is very difficult to treat. Early diagnosis of lung cancer can significantly increase the chances of survival of a cancer patient. As a result, an effective nodule detection device will make it more likely to be successfully treated in the early detection of lung cancer. Current detection solutions are largely observation-based, where clinical experts rely on low dose CT scans & MRI images to diagnose the disease, while our project aims to improve the detection process of potentially cancerous lung nodules in these scans. The concept discussed in this study is the use of Convolutional neural networks (Alexnet, VGG16 & VGG19) to effectively identify the image data set as cancer and tumour and also classify it as one of the non-small cell lung cancer. We download lung cancer data from Kaggle in the form of CT SCAN & MRI images and process it through different CNN nets to detect and classify lung cancer on Matlab. The types of lung cancers classified are Adenocarcinoma, Squamous cell carcinoma and large cell carcinoma in our selected dataset by the help of our application. Our application determined the accuracy, precision, errors, F1 score, FPR, specificity and sensitivity of each CNN net.