A Segmentation Algorithm for Contrast-Enhanced Images

Abstract

Medical imaging often involves the injection of con-trast agents and the subsequent analysis of tissue en-hancement patterns. Many important types of tissue have characteristic enhancement patterns; for exam-ple, in magnetic resonance (MR) mammography, ma-lignancies exhibit a characteristic “wash out ” temporal pattern, while in MR angiography, arteries, veins and parenchyma each have their own distinctive temporal signature. In such image sequences, there are substan-tial changes in intensities; however, this change is due primarily to the contrast agent rather than the motion of scene elements. As a result, the task of segmenting contrast-enhanced images poses interesting new chal-lenges for computer vision. In this paper, we propose a new image segmentation algorithm for image sequences with contrast enhance-ment, using a model-based time series analysis of in-dividual pixels. We use energy minimization via graph cuts to efficiently ensure spatial coherence. The en-ergy is minimized in an expectation-maximization fash-ion that alternates between segmenting the image into a number of non-overlapping regions and finding the temporal profile parameters which best describe the be-havior of each region. Preliminary experiments on MR mammography and MR angiography studies show the algorithm’s ability to find an accurate segmentation. 1. Contrast-Enhanced Image Se-quences Many medical imaging studies involve the injection of contrast agents in order to visualize structures that cannot otherwise be distinguished. The choice of con-trast medium depends upon the imaging modality; in CT the contrast agents are iodine-based, while MR contrast agents use gadolinium [13]. In the result-ing contrast-enhanced image sequences the patient is generally stationary, and the substantial changes in intensities that occur are due primarily to the con-trast agent. Moreover, different types of tissues of-ten have characteristic enhancement patterns, which