Brain Tumor Detection and Classification Using Cycle Generative Adversarial Networks

Brain cancer ranks tenth on the list of leading causes of death in both men and women. Biopsy is one of the most used methods for diagnosing cancer. However, the biopsy process is quite dangerous and take a long time to reach a decision. Furthermore, as the tumor size is rising quickly, non-invasive, automatic diagnostic equipment is required which can automatically detect the tumor and its stage precisely in a few seconds. In recent years, techniques based on Machine Learning and Deep Learning (DL) for detecting and classifying cancers has gained remarkable success in recent years. This paper suggested an ensemble method for detecting and classifying brain tumor and its stages using brain Magnetic Resonance Imaging (MRI). A modified InceptionResNetV2 pre-trained model is used for tumor detection from MRI image. After tumor detection, a combination of InceptionResNetV2 and Random Forest Tree (RFT) is used to determine the cancer stage, which includes glioma, meningioma, and pituitary cancer. The size of the dataset is small, so C-GAN (Cyclic Generative Adversarial Networks) is used to increase the dataset size. The experiment results demonstrate that the suggested tumor detection and tumor classification models achieve the accuracy of 99% and 98%, respectively.