Deep Learning Model for Brain Tumor Segmentation & Analysis

Gliomas are the most hostile among brain tumors and other known cranial ailments. Patient with high-grade Gliomas often results in diminished life span. In an attempt to provide a solution, we propose a brain segmentation model for tumor detection which could be considered an appropriate tool for medical image processing. Due to the presence of an enormous quantity of magnetic resonance scanned images, performing cancer diagnosis by lab scanned images becomes quite laborious and hand-operated segmentation for brain tumor becomes complicated and slow-moving chore. One reason for its complexity could be explained by the fact that, these tumors have high probabilities of materializing in any part of the human brain without having defined size or shape. The aforementioned problems fueled our research to built a novel approach using deep learning which utilizes both global as well as local features of medical imaging for precise segmentation. In our research, we have implemented and tested Convolutional Neural Network (CNN) based models for detecting brain tumors by utilizing the concept of deep neural networks. Our proposed deep learning model combines Two-Pathway and Cascade architectures to analyse and implement brain segmentation. These models are used for segmenting Gliomas of both types. Our results are evaluated over Input Cascade and outcomes showed better performance than existing MFC Cascade.