Visualization and segmentation of liver tumors using dynamic contrast MRI

Hepatocellular carcinoma (liver tumor) is one of the most common malignancies causing an estimated one million deaths annually, and the fastest growing form of cancer in the United States. Dynamic contrast enhanced MRI (DCE-MRI) is a useful way to characterize tumor response to contrast agent uptake, but the method still lacks maturity in terms of quantifying tumor burden and viability. We propose a semi-supervised technique for visualizing and measuring liver tumor burden and viability from DCE-MRI examinations. In order to solve the challenging segmentation problem, we exploit prior information about the spatio-temporal characteristics of DCE-MRI data, and perform k-means clustering in a hybrid intensity-spatial feature space.