Myeloma, Melanoma, Lung, Breast, Colon and Brain Cancer Detection Using Deep Learning in Web Based Application

Abstract

Cancer is a terminal disease that is caused by the assemblage of abnormal cells in the human body. Various pathological changes, and at times inherited problems cause such malignancies-the frequent being melanoma, myeloma, colon, lung, and breast cancers. There are several ways for identifying these cancers, however the process is very expensive and require experts. This project aims to create a web-based tool to detect these multiple malignancies. The project was split into two sections. In the first section, a detection model was created for each of the cancers. Then employed pre-trained convolutional neural network models to identify cancers using picture input. In the second part, these models were deployed on a website for detection. This website will help patients and doctors worldwide to diagnose these diseases without any advanced technology, which is very costly. We have merged all the cancer dataset and made a hybrid dataset for detection of all the cancer with only one model. The trained model gave 93% accuracy (3 classes with below 80% accuracy) on the test dataset without K-Fold and 93% accuracy (no individual class accuracy less than 80%) with K-Fold Cross Validation.