Hepatocellular carcinoma (HCC) is characterized by a high mortality rate. The Liver Imaging Reporting and Data System (LI-RADS) results in a considerable number of indeterminate observations, rendering an accurate diagnosis difficult.

We developed four deep learning models for diagnosing HCC on computed tomography (CT) via a training-validation-testing approach. Thin-slice triphasic CT liver images and relevant clinical information were collected and processed for deep learning. HCC was diagnosed and verified via a 12-month clinical composite reference standard. CT observations among at-risk patients were annotated using LI-RADS. Diagnostic performance was assessed by internal validation and independent external testing. We conducted sensitivity analyses of different subgroups, deep learning explainability evaluation, and misclassification analysis.

From 2,832 patients and 4,305 CT observations, the best-performing model was Spatio-Temporal 3D Convolution Network (ST3DCN), achieving area under receiver-operating-characteristic curves (AUCs) of 0.919 (95% CI, 0.903-0.935) and 0.901 (95% CI, 0.879-0.924) at the observation (n = 1,077) and patient (n = 685) levels, respectively during internal validation, compared with 0.839 (95% CI, 0.814-0.864) and 0.822 (95% CI, 0.790-0.853), respectively for standard of care radiological interpretation. The negative predictive values of ST3DCN were 0.966 (95% CI, 0.954-0.979) and 0.951 (95% CI, 0.931-0.971), respectively. The observation-level AUCs among at-risk patients, 2-5-cm observations, and singular portovenous phase analysis of ST3DCN were 0.899 (95% CI, 0.874-0.924), 0.872 (95% CI, 0.838-0.909) and 0.912 (95% CI, 0.895-0.929), respectively. In external testing (551/717 patients/observations), the AUC of ST3DCN was 0.901 (95% CI, 0.877-0.924), which was non-inferior to radiological interpretation (AUC 0.900; 95% CI, 0.877--923).

ST3DCN achieved strong, robust performance for accurate HCC diagnosis on CT. Thus, deep learning can expedite and improve the process of diagnosing HCC.

The clinical applicability of deep learning in HCC diagnosis is potentially huge, especially considering the expected increase in the incidence and mortality of HCC worldwide. Early diagnosis through deep learning can lead to earlier definitive management, particularly for at-risk patients. The model can be broadly deployed for patients undergoing a triphasic contrast CT scan of the liver to reduce the currently high mortality rate of HCC.

Hepatocellular carcinoma (HCC) is one of the most common malignancies worldwide. The majority of HCCs develop in cirrhotic livers, and the early detection and characterization of this entity is very important. Pathologically, human HCC develops in a multistep fashion in the following sequence: from low-grade dysplastic nodule (LGDN), to high-grade dysplastic nodule (HGDN), early HCC, well-differentiated HCC, nodule-in-nodule HCC, and, finally, to moderately differentiated HCC. Differentiation between early HCC and DN is the most important issue in the clinical setting. CT during hepatic angiography (CTHA) and CT during arterial portography (CTAP) are the most sensitive tools in the differentiation of premalignant/borderline lesions (LGDN and HGDN) and early HCC. Recent progress in imaging modality, especially Sonazoidenhanced US and Gd-EOB-DTPA MRI, is starting to play a very important role in the imaging of multistep hepatocarcinogenesis, resulting in changing the therapeutic strategy of these nodular lesions associated with liver cirrhosis.

Ultrasound (US) is the imaging method most frequently used for the detection and diagnosis of hepatocellular carcinoma (HCC). US image quality has greatly improved in recent years, with advances in transducer, beam-former technology and sophisticated image processing. In addition, the recent availability of US contrast agents including first- and second-generation contrast agents and the remarkable advances in US technology have led to the rapid development of new imaging methods such as low and high mechanical index (MI) imaging technology for the vascular phase and parenchymal phase of the liver. In terms of contrast agent, most reported studies in Asian countries have used a very fragile, first-generation microbubble agent. Nowadays, more stable,second-generation contrast agents are developing and used in clinical application to the liver. In the future, better results can be obtained in the diagnosis of HCC, decision of treatment method and assessment of therapeutic response if more durable contrast agents are used with advanced US techniques. In this article, the principles, methods, imaging findings and potential roles of new diagnostic US techniques with contrast agents in the diagnosis of HCC are discussed.

Various types of hepatocellular nodules are seen in cirrhotic livers. In these nodules, two types of human hepatocarcinogenesis are now considered. One is de novo hepatocarcinogenesis and the other is the stepwise development from high-grade dysplastic nodule (DN), high-grade DN with well-differentiated HCC foci, and overt HCC. According to our analysis by CT during arterial portography (CTAP) and CT during hepatic arteriography (CTHA) and histological study, in accordance with the elevation of the grade of malignancy of the nodules, the portal tract including normal portal vein (intranodular portal supply) and hepatic artery (intranodular arterial supply through normal hepatic arteries) are decreased. On the other hand, abnormal artery (intranodular arterial supply through newly formed abnormal arteries) gradually increases. Therefore, we can estimate the grade of malignacy of the nodules from intranodular blood supply. To know this blood supply pattern is important for the early detection, characterization and treatment of early stage HCCs. We also revealed that there was a close correlation between the prognosis of the nodules and the blood supply patterns.

Hepatocellular carcinoma (HCC) is a multistage disease whose occurrence is linked to environmental, dietary and life-style factors. Risk factors for HCC have been grouped into two clusters, physiologic and pathologic factors. Much evidence indicates that these risk factors may interact with each other either as a group or as a single factor to modulate development of liver cancer.

To compare the diagnostic accuracies of Lipiodol computed tomography (CT) and helical biphasic CT as preoperative imaging modalities for hepatocellular carcinoma (HCC).

Lipiodol CT after digital subtraction angiography has long been used as a highly sensitive imaging modality for HCC. The recent advent of helical CT has allowed scanning the entire liver during both the arterial and portal venous phase of contrast enhancement.

The authors analyzed data from 164 patients who underwent hepatic resection for HCC to calculate the sensitivity and specificity of these modalities. Findings of intraoperative ultrasonography followed by histologic confirmation were set as the gold standard.

Although sensitivity decreased with both modalities as tumors became small and well differentiated, helical CT showed a higher sensitivity than Lipiodol CT in detecting well-differentiated HCC nodules smaller than 2 cm. In contrast, Lipiodol CT was superior to helical CT for the detection of small but moderately to poorly differentiated nodules. The overall sensitivity of helical CT was higher than that of Lipiodol CT. These findings suggest that helical CT is superior in delineating early HCC, whereas Lipiodol CT is specific to the detection of intrahepatic metastases. In terms of specificity, helical CT was superior to Lipiodol CT.

Helical CT and Lipiodol CT are complementary modalities. At present, helical biphasic CT does not obviate the need for invasive techniques such as angiography and Lipiodol CT as preoperative examinations for HCC.

Early diagnosis of HCC is possible because certain risk factors for this tumor are known and because sensitive and relatively inexpensive diagnostic tools are available. Early diagnosis of HCC is also possible because of the long phase of asymptomatic tumor growth and the tumor's tendency to grow as a solitary mass in many patients. In two consensus development conferences held in Anchorage, Alaska and in Milan, Italy, chronic carriers of HBsAg, patients with cirrhosis, patients with rare metabolic liver diseases, and individuals with family histories of HCC were identified as patients at high risk for HCC and therefore as candidates for periodic screening. At the Anchorage conference, it was recommended that healthy carriers have at least yearly determinations of serum AFP and that carriers with additional risk factors (e.g., cirrhosis) be screened every 6 months by abdominal US scans and determination of serum AFP levels. No specific recommendations were released for HBsAg-negative patients with chronic liver disease. At the Milan conference, it was recommend that patients with cirrhosis or with certain congenital metabolic conditions known to be at risk for HCC should be screened by AFP determination and US scan twice a year. It was also recommended that HBsAg carriers older than 35 years or with family histories of HCC should be screened for HCC by determinations of serum AFP levels and aminotransferase levels once a year.

Surgical resection has been the treatment of choice for hepatocellular carcinoma (HCC), but the resection rate remains low in cirrhotic patients and recurrence is common. Unfavorable results compared with benign disease and the shortage of organ donors have led to a restricted indication for orthotopic liver transplantation (OLT) for HCC.

The aim of this study was to analyze the results of our surgical approach to HCC in patients with cirrhosis. The first treatment strategy indicated in these patients was OLT. From January 1990 to May 1999, 85 patients underwent OLT and the remaining 35 had surgical resection.

One-, 3-, and 5-year survival rates were 84%, 74%, and 60% versus 83%, 57%, and 51%, respectively, in the OLT and resection groups (p = 0.34). Hepatic tumor recurrence was much less frequent in the OLT group than in the resection group. The 1-, 3-, and 5-year disease-free survival rates were 83%, 72%, and 60% versus 70%, 44%, and 31%, respectively (p = 0.027). In the multivariate Cox regression analysis, macroscopic vascular invasion was the only factor independently associated with death or recurrence after OLT (p = 0.006). After partial liver resection, the tumors significantly associated with mortality and recurrence in the multivariate analysis were solitary or multiple tumors greater than 2cm with microscopic vascular invasion (pathologic pT3) (p = 0.01).

Our results confirm that in cirrhotic patients, OLT may provide better outcomes than liver resection in carefully selected HCC and that longterm survival is similar to the results of OLT in cirrhotic patients without tumors.

Hepatocellular carcinoma (HCC) is increasing in many countries as a result of an increase in hepatitis C virus (HCV) infection since World War II. The epidemiology of HCC varies with the global region. There have been conflicting observations from different parts of the world concerning the frequency of HCC in patients who in the distant past had post-transfusion non-A, non-B hepatitis. The genetic basis of hepatocarcinogenesis is still poorly understood. In hepatitis B virus (HVB) associated HCC, codon 249 mutation in the p 53 gene seems more related to exposure to aflatoxin B1 than to hepatocarcinogenesis itself. HCC that occurs in children in high HBV endemic regions could be associated with germ-line mutations, but little information is available; not much is known about chemical hepatocarcinogens in the environment other than aflatoxins. The X gene of HBV seems to play an important role in HBV-associated hepatocarcinogenesis. There are preliminary observations on the molecular mechanism of HCV-associated HCC, such as HCV core protein inducing HCC in transgenic mice and the NS3 genome transforming NIH 3T3 cells. Pathological distinction between preneoplastic and very early transformed lesions still depends on classical morphology, and a more genetically oriented differential diagnosis is required. Clinical diagnosis based on modern imaging has improved greatly, but is still unsatisfactory in the differential diagnosis of preneoplastic and early transformed nodules, because the vasculature changes that occur within the nodule are not accurately discerned with the current imaging. Use of sensitive des-gamma-carboxy prothrombin (PIVKA II) assay, and lectin affinity chromatography separating HCC specific subspecies of AFP molecules with a more practical biochemical technique will further improve diagnosis. Early diagnosis and transplantation are the best treatment at the moment, but transplantation is not widely available because of the donor shortage. Despite successful resection, the remnant cirrhotic liver frequently develops new HCC lesions, seriously curtailing long-term survival. All-out efforts should be directed to the prevention of HCC, through prevention of viral hepatitis, prevention of acute hepatitis from becoming chronic, prevention of chronic hepatitis from progressing to cirrhosis, and prevention of the cirrhotic liver from developing HCC (chemoprevention). At the moment, very few such studies exist.

Hepatocellular carcinomas (HCC) often contain subpopulations of cells showing heterogeneous differentiation within each tumour. The majority of HCCs first appear as well-differentiated lesions and proliferate with gradual dedifferentiation. The present study was designed to investigate the clonal diversity which is seen with progression in neoplasms. The degree of genomic heterogeneity of HCC nodules was assessed using the arbitrarily primed-polymerase chain reaction technique. Two or more sectors of 31 HCC nodules were needle-microdissected and amplified with two different arbitrary primers in appropriate conditions. In every HCC less than 6 mm in diameter (n=18, range 3-6 mm, mean diameter 4.7 mm), all sectors of each of these lesions had the same DNA fingerprint. All HCC nodules greater than 6 mm diameter (n=13, range 7-30 mm, mean diameter 15.4 mm) showed distinct DNA fingerprints in each sector sampled (p< 0. 05, compared with size less than 6 mm in diameter). When synchronous HCCs were present, no two tumour nodules had the same DNA fingerprint. These results suggest that a process of clonal evolution occurs in expanding HCC, with neoplasms more than 6 mm in diameter developing as multiple clones. The advent of laser capture microdissection technology makes such analysis much more rapid and easily applied. Studies of clonality in HCCs, including borderline cases, are made possible by the combination of these novel techniques.

Lipiodol has been shown to concentrate in most hepatocellular carcinomas as well as in some liver metastases, including those of neuroendocrine origin. Our aim was to determine the proportion of neuroendocrine liver metastases that take up lipiodol and to identify tumour characteristics that predict avidity.

Avidity was assessed in 12 patients with neuroendocrine liver metastases by performing an abdominal CT scan immediately after selective hepatic arterial injection of 5 ml of unlabelled lipiodol and this was correlated with number and size of lesions as well as angiographic and plain CT scan features.

In seven patients the tumours displayed lipiodol avidity (four solitary, three multiple); five patients had non-avid lesions (all multiple). A large dominant liver tumour was the only predictor of avidity (mean diameter of largest lesion 9 cm vs. 3 cm for patients with non-avid tumours: P=0.01). Avidity was not related to vascularity or CT density of lesions.

Although this is a small study, it would appear that approximately 50% of neuroendocrine liver metastases selectively concentrate lipiodol, which could have implications for targeted cancer therapy.