Reconstruction, identification and classification of brain tumor using gan and faster regional-CNN

Brain tumor defines rapid development of abnormal cells in the brain. These tumors are capable of cause problems in the working of the brain as the space inside the skull is small and extra growth in a tiny area can lead to severe problems. There are two types of brain tumors: cancerous (malignant) and noncancerous (non-cancerous) (benign). Since malignant tumours may spread to other areas of the body, they are more harmful than benign tumours. These malignant tumors can be controlled from spreading to other parts of the body and possibly cured if its detection is done at an early stage. In the system we propose, we first identify, then find its exact location and finally classify the type of tumor inexpensively and quickly. Our proposed system uses DCGAN as a preprocessing technique in which generator creates fake images to fool the discriminator as it a real image to give a large data set less expensively. Hence the accuracy of preprocessed data in more compare to other preprocessing technique. These preprocessed data is trained using Faster R-CNN and classified into three types namely meningioma, glioma, pituitary and a plain type. In Faster R-CNN the region proposal time is 10ms which is very much lesser compared to other R-CNN technique.