Automatic liver tumor diagnosis with Dynamic-Contrast Enhanced MRI

Dynamic-contrast enhanced MRI is currently used as a complementary diagnosis tool to assess the malignancy of the liver tumor, called hepatoma or adult primary liver cancer. This paper proposes a set of features and computation methods to extract them in order to design a classifier for automatic diagnosis of the hepatoma. A dynamic discrete linear pharmacokinetic model is used to estimate the perfusion curves from the noisy observations, based on the multi-compartment paradigm. The arterial response to the contrast agent bolus injection, called arterial input function, is also estimated since no arteries are available in the neighborhood of the tumor. The compensation of involuntary movements of the patient, such as the respiratory activity, during the acquisition process is performed using a mutual information based registration algorithm with non-rigid transformations. The tumor is isolated in a small region of interest in order to speed up the analysis and the tumor itself is segmented using an active contour algorithm. The classification of the tumor can be based on the mean and variance values of the Maximum, Washln and WashOut rates of the perfusion curves inside the tumor as it is shown that these rates are adequate discriminative features to automatically classify the tumors with respect to its malignancy. Tests using real data are used to illustrate the ability of using these features to classify the tumor with respect to its malignancy.