A novel methodology for brain tumor detection based on two stage segmentation of MRI images

Medical imaging modalities like Magnetic Resonance Imaging, Computed Tomography, Positron emission tomography are complex modalities. It contains different objects like soft tissues and bony structures along with the noise included during the process of image acquisition. Segmentation of these images is a challenging task, as we need to separate the different objects and noise to get better diagnostic results. This paper presents a novel methodology of two stage segmentation. In the first stage, we apply Gabor filter banks generated using different frequencies and orientations. In the filtered image Region Of Interest (ROI) is identified and iterative contour detection method is applied to the filtered image, which detects different objects in the image. Segmentation is achieved using discrete gray level sets. Tumor region is extracted using threshold segmentation and it is analyzed for shape and size. The simulation results obtained are very promising for the detection and analysis of the brain tumor.