Development of novel deep multimodal representation learning-based model for the differentiation of liver tumors on B-mode ultrasound images.

BACKGROUND AND AIM: Recently, multimodal representation learning for images and other information such as numbers or language has gained much attention. The aim of the current study was to analyze the diagnostic performance of deep multimodal representation model-based integration of tumor image, patient background, and blood biomarkers for the differentiation of liver tumors observed using B-mode ultrasonography (US). METHOD: First, we applied supervised learning with a convolutional neural network (CNN) to 972 liver nodules in the training and development sets to develop a predictive model using segmented B-mode tumor images. Additionally, we also applied a deep multimodal representation model to integrate information about patient background or blood biomarkers to B-mode images. We then investigated the performance of the models in an independent test set of 108 liver nodules. RESULTS: Using only the segmented B-mode images, the diagnostic accuracy and area under the curve (AUC) values were 68.52% and 0.721, respectively. As the information about patient background and blood biomarkers was integrated, the diagnostic performance increased in a stepwise manner. The diagnostic accuracy and AUC value of the multimodal DL model (which integrated B-mode tumor image, patient age, sex, aspartate aminotransferase, alanine aminotransferase, platelet count, and albumin data) reached 96.30% and 0.994, respectively. CONCLUSION: Integration of patient background and blood biomarkers in addition to US image using multimodal representation learning outperformed the CNN model using US images. We expect that the deep multimodal representation model could be a feasible and acceptable tool for the definitive diagnosis of liver tumors using B-mode US.

A Novel Non-invasive Method for Predicting Liver Fibrosis by Quantifying the Hepatic Vein Waveform.

The hepatic vein (HV) waveform by Doppler ultrasound reflects the severity of liver fibrosis. We conducted a proof-of-concept study of a new method for quantifying the HV waveform. We calculated the coefficient of variation (CV) of the HV flow velocity and created a new index "q-HV" (quantified HV) and analyzed its performance for predicting histologic liver fibrosis in 114 patients with chronic liver disease. The CV of the HV flow velocity was well associated with flattening of the waveform and the q-HV significantly increased with the progression of liver fibrosis. The areas under the curve for the prediction of fibrosis stage were 0.732 for F2, 0.772 for F3 and 0.805 for F4. Combined q-HV and FIB-4 index (widely used liver fibrosis score) increased the diagnostic accuracy for liver fibrosis. The q-HV showed good accuracy for predicting liver fibrosis; thus, q-HV is feasible and acceptable as a non-invasive tool for predicting liver fibrosis.

Identification of liver fibrosis using the hepatic vein waveform in patients with Fontan circulation.

AIM: Liver fibrosis caused by congestive hepatopathy has emerged as an important complication after Fontan procedure. We evaluated the utility of the hepatic vein (HV) waveform using Doppler ultrasound for identification of liver fibrosis in Fontan patients. METHODS: We investigated the HV waveforms in 41 Fontan patients and assessed correlations with clinical parameters, liver fibrosis markers, and hemodynamic data. RESULTS: Based on our preliminary analysis of 64 adult patients with chronic liver disease who underwent liver biopsy, we classified HV waveforms into five types with reference to the degree of flattening (from type 1, normal triphasic waveform; to type 5, a monophasic waveform indicating cirrhosis), and confirmed a significant correlation between waveform pattern and fibrosis stage. Notably, we detected HV waveforms in all of the Fontan patients and classified them into five types. The HV waveform pattern positively correlated with γ-glutamyl transferase and hyaluronic acid levels, and negatively correlated with albumin level and platelet count, but did not correlate with central venous pressure or brain natriuretic peptide level, suggesting that HV waveform could reflect pathophysiological changes in the liver without being affected by hepatic congestion. The highest area under the receiver operating characteristic curve of the HV waveform for detecting advanced liver fibrosis, as defined by ultrasonic findings and clinical features, was 0.829 (81.8% sensitivity, 73.3% specificity), which was higher than that of other non-invasive fibrosis markers. CONCLUSIONS: Hepatic vein waveforms change in accordance with liver fibrosis progression in Fontan patients, and can be a useful indicator of liver fibrosis after the Fontan procedure.

Eradication of hepatitis C virus is associated with the attenuation of steatosis as evaluated using a controlled attenuation parameter.

Chronic hepatitis C virus (HCV) infection was shown to cause hepatic steatosis or suppression of serum lipid levels. However, little is known about the changes in hepatic steatosis following HCV eradication. We aimed to evaluate this issue using the controlled attenuation parameter (CAP), which was recently shown to provide a standardized non-invasive measure of hepatic steatosis. We enrolled 70 patients with chronic HCV infections and steatosis (CAP of over 248 dB/m) who had achieved a sustained viral response at 12 weeks after discontinuation of antiviral treatment using direct-acting antivirals (DAA). We then evaluated the state of hepatic steatosis before and after HCV eradication. We also investigated the changes in serum parameters such as cholesterol and glucose levels. The median value of CAP level decreased significantly after HCV eradication from 273 dB/m to 265 dB/m (P = 0.034). Also, LDL and HDL cholesterol levels increased significantly after HCV eradication (P = 0.002 and P = 0.027, respectively). In conclusion, a decrease in hepatic steatosis after HCV eradication with DAA was revealed in chronic hepatitis C patients with significant steatosis. Cancellation of the viral effect is a possible underlying cause of hepatic steatosis improvement and increase in HDL and LDL cholesterol levels.

Disappearance of perihepatic lymph node enlargement after hepatitis C viral eradication with direct-acting antivirals.

Perihepatic lymph node enlargement (PLNE) which has been shown to be negatively associated with hepatocellular carcinoma (HCC) occurrence is frequently observed in chronic liver disease; however, changes in the state of perihepatic lymph nodes after eradication of hepatitis C virus (HCV) have not been investigated yet. We aimed to evaluate this issue. We enrolled 472 patients with chronic HCV infection who achieved viral eradication with direct-acting antivirals (DAA). We investigated whether the status of perihepatic lymph nodes changed before and after HCV eradication (primary endpoint). We also evaluated the association between PLNE and clinical findings such as liver fibrosis or hepatocellular injury before HCV eradication (secondary endpoint). Perihepatic lymph node enlargement was detected in 164 of 472 (34.7%) patients before DAA treatment. Surprisingly, disappearance of PLNE was observed in 23.8% (39 patients) of all PLNE-positive patients after eradication of HCV. Disappearance of PLNE was not associated with baseline clinical parameters or changing rates of clinical findings before and after DAA treatment. At baseline, presence of PLNE was significantly associated with a lower serum HCV-RNA level (P = .03), a higher serum AST level (P = .004) and a higher ALT level (P < .001) after adjustment for sex and age. In conclusion, PLNEs became undetectable after DAA treatment in 23.8% of PLNE-positive patients. Further study with a longer follow-up period is needed to clarify the clinical importance of this phenomenon especially in relationship with the risk of HCC development.

Ultrasound evaluation of liver stiffness: accuracy of ultrasound imaging for the prediction of liver cirrhosis as evaluated using a liver stiffness measurement.

BACKGROUND AND AIMS: Because of the low penetration rate of transient elastography (TE) or its limitations in patients with obesity, narrow intercostal spaces, or ascites, the physical appearance of the liver as visualized using ultrasonography (US) is still thought to provide important information for the prediction of liver fibrosis. We examined the accuracy of various US signs when assessing the presence of liver cirrhosis, compared with TE. METHODS: We enrolled 189 patients who had undergone both conventional US and TE examinations. We then assessed the associations between US parameters of the liver (surface, edge, and parenchymal texture) or the US score (sum of each parameter score), and the presence of liver cirrhosis as determined based on a liver stiffness measurement (LSM) of >15. RESULTS: A significant increase in the LSM was observed according to the liver surface score (P < 0.001), liver edge score (P < 0.001), parenchymal texture score (P < 0.001), and US score (P < 0.001). The areas under the curves (AUROC) for the prediction of an LSM >15 for the liver surface, liver edge, parenchymal texture, and the US score were 0.859, 0.768, 0.837, and 0.902, respectively. The AUROC of the US score was higher than that of the APRI score (0.823) or the FIB-4 index (0.804). Using an optimal cut-off value of 3.5, the sensitivity and specificity of the US score were 0.815 and 0.858, respectively. CONCLUSIONS: The US score was clinically useful for the diagnosis of an LSM >15. The US score can be used as a substitute for TE data in patients with obesity, narrow intercostal spaces, or ascites or in hospitals where TE is unavailable.