Fully automated brain tumor segmentation and volume estimation based on symmetry analysis in MR images

Abstract

Abnormal and uncontrolled cell divisions cause brain tumors. Fast and accurate detection of tumors in early phase is important for succesfull diagnosis and treatment. Expert physicians use image slices obtained from advanced imaging techniques such as Magnetic Resonance Imaging (MRI), Computed Tomopghraphy (CT) to define existing of a tumor. This process has a difficulty as it requires a high concentration on many image slices. On the other hand, image processing techniques can successfully be used to detect a tumor and its sizes in order to assist to expert physicians. In this work, brain tumor detection and volume estimation by using FLAIR, T1 Pre Gadolinium and T1 Post Gadolinium (T1C) MRI protocols is presented. Method used in this study is fully automatic and applicable to different types of tumors. The work has been tested on 500 visual DICOM format axial brain MR slices of ten patients. Tumor detection is realized by using left-right symmetry analysis assuming that brain consists of two symmetric cerebral hemispheres. Also, thresholding, skull stripping and fuzzy c mean clustering techniques are applied to detect abnormal brain regions. Tumor volume is calculated by the help of detected tumor area of each MRI slice and MRI slice thickness information obtained from DICOM header.