Automatic Segmentation of Region of Interests in MR Images Using Saliency Information and Active Contours

Magnetic resonance imaging (MRI) is the most clinically used and gifted modality to identify brain abnormalities in individuals who might be at risk for brain cancer. To date, automated brain tumor segmentation from MRI modalities remains a sensitive, computationally expensive, and a demanding task. This paper presents an automated and robust segmentation method to enable investigators to make successful diagnosis and planning of radiosurgery by reducing the risk factor and study duration. The proposed system consists of following steps: (1) remove the non-brain part from MRI, (2) estimate saliency map of MRI, (3) use the salient region (tumor) as an identification marker and segment the salient object by finding the “optimal” closed contour around the tumor. The system has been tested on real patient images with excellent results. The qualitative and quantitative evaluations by comparing with ground truths and with other existing approaches demonstrate the effectiveness of the proposed method.