Brain Tumor Classification Using Deep Hybrid Techniques

Brain tumour pose a substantial health risk, potentially resulting in serious neurological issues and even fatality. Magnetic Resonance Imaging (MRI) is a frequently employed diagnostic instrument for detecting and categorizing brain tumour; however, manually examining MRI images can be labour-intensive, subject to bias, and susceptible to human mistakes. Consequently, there is an increasing demand for creating automated and dependable techniques for classifying brain tumour using deep learning strategies. In this study, it was attempted to identify the disease’s stage or whether it was a brain tumour using scans of the brain. To identify the illness, CNN architectures are employed. Additionally, a hybrid approach that we created has been suggested. Using the architectures currently in use, it is divided into 4 kinds based on the rate of disease progression. Alexnet, Resnet18, Googlenet, and the Hybrid technique we devised each produce the findings independently. The Resnet18 approach is the cornerstone of the proposed hybrid model. With the created hybrid model, an accuracy of 91% has been attained. Therefore, it is determined that the hybrid model created to diagnose brain tumours has attained the success achieved by previous CNN architectures and even delivers better outcomes after looking into other scientific papers in the literature.