Using Saliency Features for Graphcut Segmentation of Perfusion Kidney Images

Abstract

In this paper we propose a method that uses visual saliency information to segment dynamic kidney perfusion data. Segmenting dynamic data requires the use of temporal information available due to the flow of contrast agent. It is found that although the intensity of the kidney changes due to flow of contrast agent, its saliency profile remains nearly constant. We exploit this characteristic to account for regional information, which we use in a graph cut formulation for segmentation. Our tests on real patient datasets show that our algorithm compares well with other approaches for segmenting dynamic data.