Some Studies on Automatic Liver Segmentation

Abstract

According to liver disease burden worldwide, around 2 million cases are recorded newlineper year, including 1 million due to viral hepatitis and hepatocellular carcinoma newline(HCC), which is the primary liver cancer and 1 million due to cirrhosis. The status newlinereport on the Global Burden of Cancer worldwide (GLOBOCAN) 2018 estimates cancer newlineincidences and mortality across the 20 different geographic regions of the world. newlineLiver cancer growth and mortality rate are rapidly increasing worldwide. As a result, newlineit became the sixth most widespread cancer and the second prominent cause of cancer newlinedeaths worldwide. Radio imaging is a non-invasive approach to visualizing the newlineinternal body structure and abnormalities. It has become a tool for medical experts to newlinediagnose hepatic anomalies and complications. The various medical imaging modalities newlineutilized include X-ray, Ultrasound, MRI (Magnetic Resonance Imaging) and newlineCT (Computerized Tomography) to diagnose liver diseases and their complications. newlineThe CT images are the preferred imaging modality of medical experts for diagnosing newlineliver diseases and their complications. However, medical experts manually delineate newlinethe liver from the CT images in routine clinical practices, which involves many newlineconstraints and the possibility of introducing errors due to human limitations. Further, newlinemedical imaging volume increases due to technological advancements, and it newlinebecomes challenging to segment the images manually by experts. Therefore, many newlineresearchers attempted to automate the process of liver and tumor segmentation in newlineCT images. However, automatic liver and tumor segmentation in CT images is still newlinechallenging because of the liver anatomical characteristics and CT image acquiring newlineprotocol. newlineRecently, deep learning gained significant attention due to its powerful automatic newlinefeature extraction capability using different filters at various stages in the network.