Two Stage Adaptive Histogram Valley based Thresholding for Tumor Extraction in Brain MRI Images

Abstract

Detection of brain tumors is a delicate aspect as far as research is concerned in the field of medical engineering. The prerequisite of removing tumors from human brain is to accurately identify the affected area of interest. Being the central processing unit of the human body any micro harm to any portion of the brain can either put the subject to miserable conditions or may be a reason for the cause of death. The work in this paper is concerned in accurately detecting the non-functional cells called tumor in human brain. A six-stage process based on two level wavelet decomposition and local and global adaptive thresholding based on histogram decomposition is used. A global threshold is then calculated using the average value of all six threshold values and the reconstructed image is once again applied for global thresholding. 77 normal and 70 affected images were considered for this work and the results were validated with the ground truth. The performance was accurate for all the images considered for this work. Inhomogeneous and negative illumination to hard and soft tissues in other images not considered in this work results in over and under segmentation.