Deep learning-based methods, in particular, convolutional neural networks and fully convolutional networks are now widely used in the medical image analysis domain. The scope of this review focuses on the analysis using deep learning of focal liver lesions, with a special interest in hepatocellular carcinoma and metastatic cancer; and structures like the parenchyma or the vascular system. Here, we address several neural network architectures used for analyzing the anatomical structures and lesions in the liver from various imaging modalities such as computed tomography, magnetic resonance imaging and ultrasound. Image analysis tasks like segmentation, object detection and classification for the liver, liver vessels and liver lesions are discussed. Based on the qualitative search, 91 papers were filtered out for the survey, including journal publications and conference proceedings. The papers reviewed in this work are grouped into eight categories based on the methodologies used. By comparing the evaluation metrics, hybrid models performed better for both the liver and the lesion segmentation tasks, ensemble classifiers performed better for the vessel segmentation tasks and combined approach performed better for both the lesion classification and detection tasks. The performance was measured based on the Dice score for the segmentation, and accuracy for the classification and detection tasks, which are the most commonly used metrics.

BACKGROUND: Hepatic steatosis (HS) can be comprehensively assessed by visually comparing the hepatic and vessel attenuation on unenhanced computed tomography (CT). We aimed to evaluate the reliability and reproducibility of a CT-based visual grading system (VGS) for comprehensive assessment of HS.
METHODS: In this retrospective study, a four-point VGS based on the visual comparison of liver and hepatic vessels was validated by six reviewers with diverse clinical experience using the unenhanced CT images of 717 potential liver donors. The diagnostic performance of VGS and quantitative indices (difference and ratio of the hepatic and splenic attenuation) to diagnose HS were evaluated using multi-reader multi-case receiver operating characteristics (ROC) analysis (reference: pathology). The interobserver agreement was assessed using Fleiss κ statistics.
RESULTS: Using the VGS, all six reviewers showed areas under the ROC curves (AUROCs) higher than 0.9 for diagnosing total steatosis (TS) ≥ 30%, macrovesicular steatosis (MaS) ≥ 30%, and MaS ≥ 10%. No difference was noted between the AUROCs of the VGS and quantitative indices (p ≥ 0.1). The reviewers showed substantial agreement (Fleiss κ, 0.61). Most discrepancies occurred between the two lowest grades of VGS (81.5%; 233/283), in which most subjects (97.0%; 226/233) had a MaS < 10%. The average-reader sensitivity and specificity of the VGS were 0.80 and 0.94 to detect TS ≥ 30% and 0.93 and 0.81 to detect MaS ≥ 10%.
CONCLUSIONS: VGS was reliable and reproducible in assessing HS. It may be useful as a non-invasive and simple tool for comprehensive HS assessment.

Liver fibrosis is a dynamic condition caused by wound-healing in which scar tissue replaces the liver parenchyma following repetitive injuries. It is hypothesized that α-mangostin (AM), the major constituent of the xanthone fraction in extracts of Garcinia mangostana L., may protect the hepatic microvascular bed from thioacetamide (TAA)-induced fibrosis. In the present study, rats were divided into 4 groups: Control rats received no treatment; TAA-treated rats received 150 mg/kg TAA 3 times per week intraperitoneally; AM-treated rats received 75 mg/kg AM twice per week intraperitoneally; and TAA+AM-treated rats received both TAA and AM as described above. Rat livers were processed either for light microscopy or for vascular corrosion casting after 30 and 60 days of treatment. Vascular parameters were measured by 3D morphometry analysis of scanning electron micrographs. AM attenuated hepatocellular injuries and delayed both periportal and pericentral fibrosis in the TAA-treated rats. The comparison of findings at day 30 and 60 showed that TAA-induced fibrotic changes were progressive in time, and that the beneficial effects of AM only became apparent after prolonged treatment. The livers of rats treated with both TAA and AM had less space surrounding the portal vessels, improved preservation of the hepatic microvascular pattern, and minimally altered sinusoidal patterns with few signs of terminal portal venule remodeling. AM therefore partially protected the liver against hepatotoxin-induced fibrosis and the associated microvascular changes. The mechanism of the protective effect of AM on the liver remains to be investigated.

Background Radiologic findings after electrochemotherapy of large hepatic blood vessels and healthy hepatic parenchyma have not yet been described. Materials and methods We performed a prospective animal model study with regulatory approval, including nine grower pigs. In each animal, four ultrasound-guided electroporated regions were created; in three regions, electrodes were inserted into the lumen of large hepatic vessels. Two types of electrodes were tested; variable linear- and fixed hexagonal-geometry electrodes. Ultrasonographic examinations were performed immediately and up to 20 minutes after the procedure. Dynamic computed tomography was performed before and at 60 to 90 minutes and one week after the procedure. Results Radiologic examinations of the treated areas showed intact vessel walls and patency; no hemorrhage or thrombi were noted. Ultrasonographic findings were dynamic and evolved from hyperechogenic microbubbles along electrode tracks to hypoechogenicity of treated parenchyma, diffusion of hyperechogenic microbubbles, and hypoechogenicity fading. Contrast-enhanced ultrasound showed decreased perfusion of the treated area. Dynamic computed tomography at 60 to 90 minutes after the procedure showed hypoenhancing areas. The total hypoenhancing area was smaller after treatment with fixed hexagonal electrodes than after treatment with variable linear geometry electrodes. Conclusions Radiologic findings of porcine liver after electrochemotherapy with bleomycin did not show clinically significant damage to the liver, even if a hazardous treatment strategy, such as large vessel intraluminal electrode insertion, was employed, and thus further support safety and clinical use of electrochemotherapy for treatment of hepatic neoplasia.