Brain Tumor Segmentation using U-Net based on Inception

Abstract

In the field of medicine, Segmentation of a Brain Tumor is a compelling and critical aspect. The main purpose of the brain tumor segmentation is to detect the scope and locale of a brain tumor which benefits to effective clinical treatment. There are numerous techniques introduced for Brain Tumor Segmentation in deep learning like CNNs, U-Net, ResNet, and DenseNet, etc. This paper intends to demonstrate the algorithm for brain tumor segmentation using a modified U-Net using inception as Backbone. Due to variable shape or locale of occurrence, segmenting brain tumors from MRIs is one of the most arduous activity in image analysis in the field of medicine. This paper provides a modification in U-Net that provides boost in the F1-Score and IoU score in segmenting brain tumor.