Identifying Morphological Markers for Differential Diagnosis of Liver Lesion

1. Introduction Liver lesions stem from a variety of causes ranging from benign to malignant. The ability to efficiently and accurately differentiate these lesions as malignant or benign using non-invasive medical imaging is important to patient treatment and outcome. Contrast-enhanced computed tomography (CT) imaging is the dominant technology used for liver lesion diagnosis [1]. This modality takes advantage of the fact that the liver receives blood from two main sources, the portal vein and the hepatic artery. About 80% of blood in the liver comes from portal vein and the other 20% comes from the hepatic artery, whereas liver cancer receive all their blood from the hepatic artery [2, 3]. Injecting contrast enhancing agents into the portal vein and performing a liver CT scan shortly thereafter will cause the liver to appear brighter than the cancer because it receives the entire dosage of contrast agent. This technique is called Portal Venous Phase (PVP) imaging. Unfortunately the specificity of this method is a function of the size of the tumor and prone to a high false positive rate because other types of benign liver lesions have similar manifestation on CT images [4]. Human variability has been shown to be an issue regarding lesion differential diagnosis, and automated methods are being investigated to facilitate the diagnosis of these lesions [5]. Recent research has investigated computer-aided methods to facilitate diagnosis by providing a database of annotated images that can be retrieved by similarity [6]. It suggests that using semantic annotations and computational imaging features can lead to highly accurate diagnosis. Semantic annotations are a