Detection, risk stratification and therapy monitoring of hepatocellular carcinoma – a deep learning approach based on iodine maps derived from dual energy computed tomography

点击翻译按钮获取中文摘要

Primary liver cancer is the sixth most common cancer type worldwide and the fourth major cause of cancer deaths, while hepatocellular carcinoma (HCC) represents its most common form. For diagnosis of HCC based on dynamic, contrast-enhanced computed tomography (CT) or magnetic resonance imaging, there are defined diagnostic criteria (LI-RADS) which, inter alia, are based on tumor perfusion. These perfusion patterns can be determined more accurately in dual-energy CT-derived iodine maps as compared to conventional CT as they allow for precise quantification of iodinated contrast media. Previous studies have shown that these iodine maps can be used to improve diagnosis of HCC, particularly in diagnostically challenging cases (e.g. very small HCC lesions). Moreover, it was shown that they can be beneficial for the response assessment of patients who underwent locoregional tumor therapy (e.g. microwave ablation, radiofrequency ablation, transarterial chemoembolization). On the other hand, advanced machine learning methods continuously gain importance in almost every field of radiology. One of these methods, the so-called deep learning, has been shown to be beneficial for imaging of HCC as well. In this project, quantitative iodine maps derived by several different dual-energy CT scanners will be transferred to deep learning models to assess whether the combination of these two emerging technologies is beneficial in terms of detection and differentiation of HCC and its risk stratification and therapy monitoring in patients who underwent locoregional tumor therapy.