This project aims at automatically detecting and segmenting focal liver lesions in CT images. CT images are commonly used for diagnosis, treatment planning and follow-up examinations of liver tumors. The images are acquired after intravenous application of a contrast agent, showing up to 6 different phases of enhancement of the tissue. The benefit of this procedure for tumor diagnostics is twofold: The visibility of the target lesions is improved and different types of lesions can be differentiated thanks to their characteristic enhancement behavior.

So far, we focus on segmenting hypodense lesions in images with venous contrast enhancement, as these images offer the best contrast between lesions and parenchyma for most types of lesions.

To detect and segment the target lesions simultaneously, a voxel classification approach is pursued, which is based on the recently proposed Probabilistic Boosting Tree [1]. During preprocessing, the search space is confined by segmenting the liver first. Also, image intensities inside the liver are standardized using the method of Jäger [2] in order to compensate for variation in acquisition timing and individual perfusion. Next, liver voxels are classified as lesion or background using an iterative scheme. From the resulting probability maps, a lesion mask is generated.