Preoperative prediction of hepatocellular carcinoma tumour grade and micro-vascular invasion by means of artificial neural network: a pilot study

Independent factors and predictive score for extrahepatic metastasis of hepatocellular carcinoma following curative hepatectomy

BACKGROUND

Postoperative extrahepatic metastasis (EHM) contributes to a poor prognosis in patients with hepatocellular carcinoma (HCC) after hepatectomy. This study was aimed to develop a practical method that can be used to predict postoperative EHM.

METHODS

In total, 578 patients were enrolled. We analyzed the clinicopathological features of the tumors and did a long-term follow-up to observe HCC recurrence. Postoperative EHM was detected in 136 patients, and multivariate analysis was used to confirm independent risk factors for postoperative EHM. After the factors were identified, a predictive scoring system was constructed as a weighted sum of these factors. The cutoff value that determines a high risk for EHM was defined by maximizing the Youden's index of the receiver operating characteristic curve.

RESULTS

Microvascular invasion, incomplete capsule, and larger tumor diameter were the three independent factors predictive for a high risk for EHM. The scoring system was derived with an area under the curve (AUC) of 0.81 for postoperative 10-year EHM prediction. A cutoff value of 43 was derived and validated with a sensitivity >90% and specificity >60% to predict the development of EHM. This system was further verified in a subgroup of Barcelona Clinic Liver Cancer stage 0-A patients with an AUC of 0.82. When the cutoff value was set at 43, the sensitivity and specificity were 90.38% and 64.88%, respectively.

CONCLUSIONS

Our predictive scoring system may be used to identify HCC patients who have a high risk for EHM following curative hepatectomy.

Comparison of artificial neural network and logistic regression models for predicting in-hospital mortality after primary liver cancer surgery

BACKGROUND

Since most published articles comparing the performance of artificial neural network (ANN) models and logistic regression (LR) models for predicting hepatocellular carcinoma (HCC) outcomes used only a single dataset, the essential issue of internal validity (reproducibility) of the models has not been addressed. The study purposes to validate the use of ANN model for predicting in-hospital mortality in HCC surgery patients in Taiwan and to compare the predictive accuracy of ANN with that of LR model.

METHODOLOGY/PRINCIPAL FINDINGS

Patients who underwent a HCC surgery during the period from 1998 to 2009 were included in the study. This study retrospectively compared 1,000 pairs of LR and ANN models based on initial clinical data for 22,926 HCC surgery patients. For each pair of ANN and LR models, the area under the receiver operating characteristic (AUROC) curves, Hosmer-Lemeshow (H-L) statistics and accuracy rate were calculated and compared using paired T-tests. A global sensitivity analysis was also performed to assess the relative significance of input parameters in the system model and the relative importance of variables. Compared to the LR models, the ANN models had a better accuracy rate in 97.28% of cases, a better H-L statistic in 41.18% of cases, and a better AUROC curve in 84.67% of cases. Surgeon volume was the most influential (sensitive) parameter affecting in-hospital mortality followed by age and lengths of stay.

CONCLUSIONS/SIGNIFICANCE

In comparison with the conventional LR model, the ANN model in the study was more accurate in predicting in-hospital mortality and had higher overall performance indices. Further studies of this model may consider the effect of a more detailed database that includes complications and clinical examination findings as well as more detailed outcome data.

Artificial neural network modeling of biomarkers to infer characteristics of contaminant exposure in Clarias gariepinus

This study examined the potential of artificial neural network (ANN) modeling to infer timing, route and dose of contaminant exposure from biomarkers in a freshwater fish. Hepatic glutathione S-transferase (GST) activity and biliary concentrations of BaP, 1-OH BaP, 3-OH BaP and 7,8D BaP were quantified in juvenile Clarias gariepinus injected intramuscularly or intraperitoneally with 10-50 mg/kg benzo[a]pyrene (BaP) 1-3 d earlier. A feedforward multilayer perceptron (MLP) ANN resulted in more accurate prediction of timing, route and exposure dose than a linear neural network or a radial basis function (RBF) ANN. MLP sensitivity analyses revealed contribution of all five biomarkers to predicting route of exposure but no contribution of hepatic GST activity or one of the two hydroxylated BaP metabolites to predicting time of exposure and dose of exposure. We conclude that information content of biomarkers collected from fish can be extended by judicious use of ANNs.

http://www.D-MELD.com, the Italian survival calculator to optimize donor to recipient matching and to identify the unsustainable matches in liver transplantation

Optimization of donor-recipient match is one of the exciting challenges in liver transplantation. Using algorithms obtained by the Italian D-MELD study (5256 liver transplants, 21 Centers, 2002-2009 period), a web-based survival calculator was developed. The calculator is available online at the URL http://www.D-MELD.com. The access is free. Registration and authentication are required. The website was developed using PHP scripting language on HTML platform and it is hosted by the web provider Aruba.it. For a given donor (expressed by donor age) and for three potential recipients (expressed by values of bilirubin, creatinine, INR, and by recipient age, HCV, HBV, portal thrombosis, re-transplant status), the website calculates the patient survival at 90days, 1year, 3years, and allows the identification of possible unsustainable matches (i.e. donor-recipient matches with predicted patient survival less than 50% at 5 years). This innovative approach allows the selection of the best recipient for each referred donor, avoiding the allocation of a high-risk graft to a high-risk recipient. The use of the D-MELD.com website can help transplant surgeons, hepatologists, and transplant coordinators in everyday practice of matching donors and recipients, by selecting the more appropriate recipient among various candidates with different prognostic factors.

Disease-free survival after hepatic resection in hepatocellular carcinoma patients: a prediction approach using artificial neural network

Background

A database for hepatocellular carcinoma (HCC) patients who had received hepatic resection was used to develop prediction models for 1-, 3- and 5-year disease-free survival based on a set of clinical parameters for this patient group.

Methods

The three prediction models included an artificial neural network (ANN) model, a logistic regression (LR) model, and a decision tree (DT) model. Data for 427, 354 and 297 HCC patients with histories of 1-, 3- and 5-year disease-free survival after hepatic resection, respectively, were extracted from the HCC patient database. From each of the three groups, 80% of the cases (342, 283 and 238 cases of 1-, 3- and 5-year disease-free survival, respectively) were selected to provide training data for the prediction models. The remaining 20% of cases in each group (85, 71 and 59 cases in the three respective groups) were assigned to validation groups for performance comparisons of the three models. Area under receiver operating characteristics curve (AUROC) was used as the performance index for evaluating the three models.

Conclusions

The ANN model outperformed the LR and DT models in terms of prediction accuracy. This study demonstrated the feasibility of using ANNs in medical decision support systems for predicting disease-free survival based on clinical databases in HCC patients who have received hepatic resection.

Tumoral and angiogenesis factors in hepatocellular carcinoma after locoregional therapy

Locoregional therapy (LRT) is used as a bridge to orthotopic liver transplant (OLT) for hepatocellular carcinoma (HCC) patients. Liver explants in OLT patients with HCC were studied regarding both tumor stage, histology, and immunohistochemical staining for cytokeratin (CK)7, CK19, P53, Ki-67, and vascular endothelial growth factor (VEGF). Patients receiving no LRT (control) (n=30) were compared with LRT treatment groups with conventional transarterial chemoembolization (cTACE) (n=25) or drug-eluting bead transarterial chemoembolization (DEB TACE) (n=17). Tumor stage and histology were similar between treatment and control groups. The mean percent necrosis was significantly higher for treatment groups versus the control group (p<0.0001 for both groups versus control) and was significantly higher in the cTACE group versus the DEB TACE group. Only the DEB TACE group showed peritumoral CK19 positivity, and tumors were all CK19-negative. Using a threshold of 50% of tumoral cells, tumoral VEGF was significantly different between groups, with the control group having the highest degree of positivity; however, peritumoral VEGF was not significantly different between the groups. The Ki-67 proliferation fraction was higher in the treated groups with a statistically significant difference between the DEB-treated group and those without treatment (p=0.02). There were no statistically significant differences in tumoral or peritumoral CK7 or p53. Percent necrosis and percent Ki-67 positivity were higher with LRT, with a significant difference between groups for percent necrosis, confirming that LRT causes necrosis and suggesting that treatment leads to increased proliferation and decreased tumoral VEGF.

Prediction of microvascular invasion of hepatocellular carcinoma: usefulness of peritumoral hypointensity seen on gadoxetate disodium-enhanced hepatobiliary phase images

PURPOSE

To determine whether peritumoral hypointensity seen on hepatobiliary phase images of preoperative gadoxetate disodium-enhanced magnetic resonance imaging (EOB-MRI) is useful for predicting microvascular invasion of hepatocellular carcinoma (HCC).

MATERIALS AND METHODS

This study was approved by the Institutional Review Board. In all, 104 HCC masses in 104 patients who had undergone EOB-MRI and liver surgery within 1 month after EOB-MRI were evaluated. Two radiologists independently recorded the presence of a peritumoral hypointensity on hepatobiliary phase. Interobserver agreement was assessed and consensus records were used. Tumor size was measured. A chi-square test and independent t-test were used for univariate analysis. Multiple logistic regression was performed to determine factors for predicting microvascular invasion. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of peritumoral hypointensity were calculated.

RESULTS

Sixty HCCs had microvascular invasion and 44 did not. Interobserver agreement in determining peritumoral hypointensity was excellent (κ = 0.83). By univariate analysis, peritumoral hypointensity and tumor size were significant for predicting microvascular invasion of HCC. On multiple logistic regression analysis, only peritumoral hypointensity was significant in predicting microvascular invasion of HCC (P = 0.013). The sensitivity, specificity, PPV, and NPV of peritumoral hypointensity were 38.3%, 93.2%, 88.5%, and 52.6%, respectively.

CONCLUSION

Peritumoral hypointensity on the hepatobiliary phase of EOB-MRI is not sensitive but has high specificity for predicting microvascular invasion of HCC.

Molecular classification and clonal differentiation of hepatocellular carcinoma: the step forward for patient selection for liver transplantation

Liver transplantation is a successful treatment for hepatocellular carcinoma (HCC). However, advanced stages are not selected for transplant, based on the United Network for Organ Sharing selection criteria's. Nowadays, molecular biology-based techniques constitute an excellent option to better understand HCC origin differentiation and biological behavior. Moreover, microarray technology is a powerful tool to address a variety of tumor tissues at molecular level and is actively used for the discovery of biomarkers in cancer research. This article will discuss published data in the field of HCC origin differentiation and its potential impact on outcomes following liver transplantation. Although preliminary results are presented, these findings encourage the use of gene-expression profiling microarrays for studying HCC biology and behavior and ultimately optimizing treatment success.

Can the dropout risk of candidates with hepatocellular carcinoma predict survival after liver transplantation?

In the last US national conference on liver transplantation for hepatocellular carcinoma (HCC), a continuous priority score, that incorporates model for end-stage liver disease (MELD), alpha-fetoprotein and tumor size, was recommended to ensure a more equitable liver allocation. However, prioritizing highest alpha-fetoprotein levels or largest tumors may select lesions at a higher risk for recurrence; similarly, patients with higher degree of liver failure could have lower postoperative survival. Data from 300 adult HCC recipients were reviewed and the proposed HCC-MELD equation was applied to verify if it can predict post-transplantation survival. The 5-year survival and recurrence rates after transplantation were 72.8 and 13.5%, respectively. Cox regression analysis confirmed HCC-MELD as predictive of both postoperative survival and recurrence (p < 0.001). The 5-year predicted survival and recurrence rates were plotted against the HCC-MELD-based dropout probability: the higher the dropout probability while on waiting list, the lower the predicted survival after transplantation, that is worsened by hepatitis C positivity; similarly, the higher the predicted HCC recurrence rate after transplantation. The HCC priority score could predict the postoperative survival of HCC recipients and could be useful in selecting patients with greater possibilities of survival, resulting in higher post-transplantation survival rates of HCC populations.

Role of alpha-fetoprotein in selection of patients with hepatocellular carcinoma waiting for liver transplantation: must we reconsider it?

BACKGROUND

Milan criteria (MC) represent the most commonly adopted criteria for the selection of patients with hepatocellular carcinoma (HCC) waiting for liver transplantation (LT). However, MC are exclusively based on morphological aspects. The aim of the present study was to evaluate pre-LT-detectable biological parameters, to compare them with morphological ones in terms of tumor recurrence prediction and patient survival.

METHODS

A cohort of 153 consecutive adult patients who underwent LT for HCC on cirrhosis from January 1999 to March 2009 was retrospectively analyzed.

RESULTS

HCC recurrence was observed in 12 patients (7.8%). At multivariate logistic regression analysis, serum alpha-fetoprotein (AFP) was the unique independent negative risk factor for the development of HCC recurrence (odds ratio 2.0, p=0.03). Adopting a cutoff value of 210 ng/mL, patients who presented serum AFP =210 ng/mL showed a 5-year survival rate of 23.3% versus 76.2% observed in patients with pre-LT serum AFP <210 ng/mL (log-rank test: <0.0001).

CONCLUSIONS

In our experience, AFP was the strongest predictor of HCC recurrence, stronger than tumor morphology. AFP could ameliorate the selection of LT candidates. Further studies to evaluate the combination of morphological and biological criteria are needed.

Diagnosis of hepatocellular carcinoma in cirrhosis by dynamic contrast imaging: the importance of tumor cell differentiation

Dynamic contrast imaging techniques are considered the standard of care for the radiological diagnosis of hepatocellular carcinoma (HCC) in cirrhosis. However, the accuracy of radiological diagnosis depends largely on the degree of arterial hypervascularization, which increases with tumor size. Owing to the interplay and prognostic relevance of tumor vascularization and cell differentation, we asked ourselves whether tumor grade also affects the outcome of radiological diagnosis. Sixty-two HCCs (47 of which measured 1-2 cm) were consecutively detected in 59 patients with compensated cirrhosis under surveillance with ultrasound and confirmed by way of echo-guided biopsy and concurrent investigations with contrast-enhanced ultrasound (CE-US), computed tomography (CT), and gadolinium magnetic resonance imaging (MRI). Tumor cell differentiation was evaluated using Edmondson-Steiner criteria in liver cores of 0.9-5.0 cm (median 1.6 cm). Eighteen (29%) HCCs were grade I (1.5 cm), 28 (45%) were grade II (1.5 cm), 16 (26%) were grade III (1.8 cm), and none were grade IV. Contrast wash-in and wash-out were concurrently demonstrated in 21 (34%) tumors by way of CE-US, including three (16%) grade I and 18 (41%) grade II-III (P = 0.08); in 32 (52%) tumors by way of CT, including three (16%) grade I and 29 (66%) grade II-III (P = 0.0006); and 28 (47%) tumors by way of MRI, including three grade I (16%) and 25 (57%) grade II-III (P = 0.01). Among 1- to 2-cm tumors, the radiological diagnosis was achieved in two of 16 grade I and 17 of 31 grade II-III tumors (P = 0.006).

CONCLUSION

Tumor grade, a relevant predictor of disease severity, influences the accuracy of dynamic contrast techniques in the diagnosis of HCC.