Brain Tumor Detection With Tumor Region Analysis Using Adaptive Thresholding And Morphological Operation

Abstract

Tumor in brain is life threatening but proper detection of tumor at early stage may save many lives. In our research, we have proposed an adaptive threshold value selection technique with morphological operation for the detection of brain tumor which is very much promising. Our proposed method can adapt with different kinds of intensity values of the pixels of MRI FLAIR image and can detect tumor efficiently. Initially we have detected the highest intensity value of the brightest region assuming as tumorous cell and also specify an intensity value which is covering maximum pixels assuming as healthy cells and their difference is also being calculated. If the difference between the maximum intensity value of the brightest region and the intensity value of any random pixel is in the range of the previous difference, then the pixel is detected as a member of tumor cell otherwise not. In this research, we have used BRATS 2013 and 2015 datasets with accuracy 95% and 89.78% respectively. As our datasets have ground truth value, we have examined our detected images with the ground truth images through the parameters centroid and area.