Brain tumor segmentation in multispectral MR images

This paper introduces a Multispectral approach for segmentation of tumor core using three MRI imagery viz. T1-contrast, T2 & Flair. In all countries the number of people diagnosed with brain tumor is increasing rapidly. This puts people's life in serious danger. A large amount of MRI scans showing brain tumor are currently being generated in clinics. The manual process for tumor segmentation by Radiologists is highly time consuming and prone to human errors and biases. Hence automatic and reliable systems to do the task are required. The proposed algorithm for segmentation of brain tumor is independent of input parameters and estimates. So it is entirely autonomous. The algorithm works in five steps: Image Acquisition, Image Enhancement followed by Segmentation (Fuzzy C Means algorithm) then Extraction of tumor core using Area and Circularity as criteria and finally evaluating the result with Ground Truth. The mean value of dice coefficient for segmenting tumor core is 0.88. Hence segmentation of tumor core using proposed method yields efficient results.