Object Contour in Medical Images Based on Saliency Map Combined with Active Contour Model

Abstract

Medical imaging is useful in the diagnosis and treatment of diseases. A wide range of pathologies have been discovered by identifying abnormalities of the object’s boundary in the medical image. Almost tokens of illness are shown in salient regions of objects, such as the intensity of color, connection together, etc. Therefore, the contour is a pressing-issue in the cutting-edge diagnose. This paper proposed a method for detecting contour of the object in medical images based on the saliency map combined with the active contour model. The proposed method detected salient regions of the medical image and activated the boundary of objects by energy reducing. The proposed method includes four steps: firstly, the pre-processing is applied for input images by the median filter. Secondly, the threshold of intensity is divided by super-pixels. Thirdly, the outstanding map is made by clustering the above super-pixels. Finally, the active contour model is built in salient regions of concerned about. We test results from DICOM images dataset, which includes high-quality and low-quality medical images. The accuracy of the proposed method is 97.52%, more than the results of the other methods, 5.0%.