Deep learning for liver evaluation: A comprehensive review and implications for ulcerative colitis detection

Abstract

This review explores the applications of deep learning based computer-aided diagnosis (DL-CAD) systems when evaluating liver images derived from Computed Tomography (CT) scans. It highlights the ability of contemporary state of the art deep learning frameworks such as Convolutional Neural Networks (CNNs) and UNets, to automate the liver lesions segmentation and classification with great accuracy. The analysis further expands on the relationship that existed between some systemic illnesses such as ulcerative colitis (UC) and specific liver related conditions such as Primary Sclerosing Cholangitis, fatty liver and autoimmune hepatitis. The above conditions which are frequently present in UC patients once again underpin the importance of imaging techniques in the provision of appropriate and timely treatment. Our research shows that the DL-CAD system may be modified appropriately in order to identify liver changes caused by UC which has advantages in diagnosis without overburdening radiologists. Furthermore, the inclusion of wearable devices for periodic liver evaluation further supports the concept of personalized patient management. Hence, this study includes notable improvements in the analysis of liver lesions and their complications in UC patients with respect to the clinical practice and treatment results.