Author : Akash A. Nitnaware and Prof. Sonal Gore

A Neurologist uses many test to locate or diagnose a brain tumor. Several tests may be conducted to identify if tumor has spread to any other part of body from where it started. This is rare in case of brain tumor and known as metastasis otherwise. One of the difficulty arises for a Radiologist or Neurologist while analyzing imaging tests for MRI scans i.e to differentiate healthy tissue from tumorous tissues. The dataset used is BRATS2015 containing both 3D and 2D images but only 2D images are considered. The proposed method is to combine SSD neural network along with GrabCut for proper segmentation of tumor from healthy tissues. Due to unpredictable shape or place of appearance, segmenting brain tumors from imaging data is one of the most challenging tasks in medical image analysis.

Keywords: Medical Resonance Imaging, Brain Tumor, GrabCut, Single Shot Multibox Detector, Segmentation.