Liver biopsy in the diagnosis of hepatocellular carcinoma

Deep Learning-Based Classification of Hepatocellular Nodular Lesions on Whole-Slide Histopathologic Images

BACKGROUND & AIMS

Hepatocellular nodular lesions (HNLs) constitute a heterogeneous group of disorders. Differential diagnosis among these lesions, especially high-grade dysplastic nodules (HGDNs) and well-differentiated hepatocellular carcinoma (WD-HCC), can be challenging, let alone biopsy specimens. We aimed to develop a deep learning system to solve these puzzles, improving the histopathologic diagnosis of HNLs (WD-HCC, HGDN, low-grade DN, focal nodular hyperplasia, hepatocellular adenoma), and background tissues (nodular cirrhosis, normal liver tissue).

METHODS

The samples consisting of surgical and biopsy specimens were collected from 6 hospitals. Each specimen was reviewed by 2 to 3 subspecialists. Four deep neural networks (ResNet50, InceptionV3, Xception, and the Ensemble) were used. Their performances were evaluated by confusion matrix, receiver operating characteristic curve, classification map, and heat map. The predictive efficiency of the optimal model was further verified by comparing with that of 9 pathologists.

RESULTS

We obtained 213,280 patches from 1115 whole-slide images of 738 patients. An optimal model was finally chosen based on F1 score and area under the curve value, named hepatocellular-nodular artificial intelligence model (HnAIM), with the overall 7-category area under the curve of 0.935 in the independent external validation cohort. For biopsy specimens, the agreement rate with subspecialists' majority opinion was higher for HnAIM than 9 pathologists on both patch level and whole-slide images level.

CONCLUSIONS

We first developed a deep learning diagnostic model for HNLs, which performed well and contributed to enhancing the diagnosis rate of early HCC and risk stratification of patients with HNLs. Furthermore, HnAIM had significant advantages in patch-level recognition, with important diagnostic implications for fragmentary or scarce biopsy specimens.

RESULTS OF IMMUNOHISTOCHEMISTRY IN THE DIFFERENTIAL DIAGNOSIS OF EARLY HEPATOCELLULAR CARCINOMA AND NODULES WITH HIGH-GRADE DYSPLASIA IN PATIENTS WITH CIRRHOSIS

BACKGROUND

Hepatocellular carcinoma (HCC) is the most frequent primary cancer of the liver and cirrhosis is considered a pre-malignant disease. In this context, the evolutionary sequence from low grade dysplastic nodule and high grade dysplastic nodule (HGDN) to early HCC and advanced HCC has been studied. The differential diagnosis between HGDN and early HCC is still a challenge, especially in needle biopsies.

OBJECTIVE

To evaluate an immunohistochemistry panel to differentiate dysplastic nodules and HCC.

METHODS

Patients with cirrhosis who underwent surgical resection or liver transplantation were included. The sensitivity, specificity and accuracy for the diagnosis of neoplasia were analyzed by evaluating five markers: heat shock protein 70, glypican 3, glutamine synthetase, clathrin heavy chain and beta-catenin. P≤0.05 was considered statistically significant.

RESULTS

One hundred and fifty-six nodules were included; of these, 57 were HCC, 14 HGDN, 18 low grade dysplastic nodules and 67 regenerative macronodules. Sensitivity of HCC diagnosis was 64.9% for glypican 3 and 77.2% for glutamine syntetase, while specificity was 96.0% and 96.0% respectively. When the panel of four markers was considered (excluding beta catenin), the specificity ranged from 87.9% for one positive marker to 100% for at least three markers. The best accuracy for HCC diagnosis was obtained with at least two positive markers, which was associated with a sensitivity of 82.5% and specificity of 99%.

CONCLUSION

Differential diagnosis of dysplastic nodules and HCC by morphological criteria can be challenging. Immunomarkers are useful and should be used for the differential diagnosis between HCC and HGDN.

Protein arginine methyl transferase 1- and Jumonji C domain-containing protein 6-dependent arginine methylation regulate hepatocyte nuclear factor 4 alpha expression and hepatocyte proliferation in mice

Alcohol is a well-established risk factor for hepatocellular carcinoma (HCC), but the mechanisms by which it promotes liver cancer are not well understood. Several studies have shown that cellular protein arginine methylation is inhibited by alcohol. Arginine methylation is controlled by the reciprocal activity of protein arginine methyltransferases, primarily protein arginine methyl transferase 1 (PRMT1), and a demethylase Jumonji C domain-containing protein 6 (JMJD6). The aim of this study was to explore the role of arginine methylation changes in alcohol pathogenesis. We found that PRMT1 activity is inhibited in livers of mice fed with alcohol compared to pair-fed mice. Using hepatocyte-specific PRMT1 knockout mice, we identified that loss of PRMT1 results in enhanced hepatocyte proliferation and a 33% increase in liver size. This increased hepatocyte proliferation was associated with reduced expression of hepatocyte nuclear factor 4 alpha (Hnf4α), an important regulator of liver tumorigenesis. We found that PRMT1 regulates Hnf4α expression directly through arginine methylation at the (Hnf4α) promoter. In the absence of PRMT1, JMJD6 can demethylate the Hnf4α promoter and suppress its expression. We were able to restore Hnf4α expression and abolish the increase in hepatocyte proliferation by knockdown of JMJD6 in PRMT1 knockout mice. Knockdown of JMJD6 in alcohol-fed mice similarly increased Hnf4α expression. We then examined whether loss of arginine methylation might play a role in alcohol-associated liver cancers. We examined 25 human HCC specimens and found a strong correlation (R = 0.8; P < 0.01) between arginine methylation levels and Hnf4α expression in these specimens, suggesting that the above mechanism is relevant in patients.

CONCLUSION

Taken together, these data suggest that PRMT1 inhibition, such as induced by alcohol, may result in epigenetic changes leading to loss of Hnf4α. This effect may contribute to alcohol's ability to promote liver tumors. (Hepatology 2018;67:1109-1126).

Anatomic pathology of hepatocellular carcinoma: histopathology using classic and new diagnostic tools

Hepatocellular carcinoma can be diagnosed on a needle biopsy of the liver; however, uncertainty may arise because of the inherent complexity of liver histology. This article aims to provide practicing pathologists with tools for the approach to mass-directed liver biopsies clinically concerning for hepatocellular carcinoma. The examination of routine hematoxylin-eosin stains and the use of ancillary histochemical and immunohistochemical stains are discussed. Sections reviewing liver carcinoma with biphenotypic differentiation and the challenge of dysplastic nodules are included.

Effect of liver regeneration on malignant hepatic tumors

Liver regeneration after major surgery may activate occult micrometastases and facilitate tumor growth, leading to liver tumor recurrence. Molecular changes during liver regeneration can provide a microenvironment that stimulates intrahepatic tumor propagation through alterations in cellular signaling pathways, where activation and proliferation of mature hepatocytes, hepatic progenitor cells, non-parenchymal liver cells might favor both liver regeneration and tumor growth. This review highlights recent advances of tumor growth and development in the regenerating liver, possible mechanisms and clinical implications.

Grading of small hepatocellular carcinomas (≤2 cm): correlation between histology, T2 and diffusion-weighted imaging

OBJECTIVE

To evaluate the capacity of diffusion-weighted imaging (DWI) to determine the histological grade of small-sized hepatocellular carcinomas (HCCs) in liver cirrhosis in comparison with T2 weighted imaging.

METHODS

51 cirrhotic patients with 63 histologically proven HCCs ≤2 cm underwent abdominal MRI, including DWI (b-values 50, 400 and 800 s mm(-2)) and T2 weighted sequences. HCCs were classified into well-differentiated HCCs (n = 37) and moderately differentiated HCCs (n = 26). Relative contrast ratios (RCRs) between the lesions and the surrounding liver were performed and compared between the two groups for T2 weighted images, each b-value and apparent diffusion coefficients (ADCs). A receiver operating characteristic (ROC) analysis was performed to compare RCRs in T2 and diffusion-weighted images.

RESULTS

We found significant differences in RCRs between well-differentiated vs moderately differentiated HCCs for b = 50, 400 and 800 s mm(-2) and T2 weighted images (1.35 ± 0.36 vs 1.86 ± 0.62; 1.35 ± 0.38 vs 1.82 ± 0.60; 1.27 ± 0.30 vs 1.74 ± 0.53; 1.14 ± 0.18 vs 1.43 ± 0.28, respectively; p < 0.001), whereas no significant differences were observed in ADC and ADC RCR (1.05 ± 0.19 vs 0.99 ± 0.15 and 1.1 ± 0.22 vs 1.09 ± 0.23; p = 0.16 and p = 0.82, respectively). No significant difference was found in the areas under the ROC curve for RCRs of T2 weighted images and every DWI b-value (p = 0.18).

CONCLUSION

The RCR measurement performed in DWI 50, 400 and 800 b-values and T2 demonstrated a significant difference between well-differentiated and moderately differentiated small-sized HCCs. Furthermore, no difference was shown by using either ADC or ADC RCR.

ADVANCES IN KNOWLEDGE

DWI with RCR measurement may be a valuable tool for non-invasively predicting the histological grade of small HCCs.

Impact of gadoxetic acid (Gd-EOB-DTPA)-enhanced magnetic resonance on the non-invasive diagnosis of small hepatocellular carcinoma: a prospective study

BACKGROUND

Gadoxetic acid (Gd-EOB-DTPA) is a 'hepatocyte-specific' contrast agent for magnetic resonance (MR) in both the vascular and the hepatobiliary phases.

AIM

To evaluate the contribution of the hepatobiliary phase of Gd-EOB-DTPA MR in the diagnosis of small hepatocellular carcinoma (HCC) in cirrhotic patients under surveillance.

METHODS

Between 2008 and 2011, 48 consecutive small (10-30 mm) liver nodules were detected in 33 patients, who prospectively underwent contrast-enhanced ultrasound (CEUS), Gd-EOB-DTPA-enhanced MR and helical-computed tomography (CT) in a blind study. The diagnosis of HCC was established according to AASLD 2005 criteria.

RESULTS

Of the 48 nodules, 38 (79%) were diagnosed as HCC, 24 (63%) of them based on AASLD non-invasive criteria, 11 diagnosed at histology and 3 during follow-up. The typical vascular pattern (arterial hypervascularisation and venous/late washout) was detected in 30 (79%) HCC nodules by MR, in 22 (58%) by CT and in 17 (45%) by CEUS. Hypointensity during the MR hepatobiliary phase was observed in all HCC nodules and in 3 nonmalignant nodules (sensitivity 100%, specificity 70%, positive predictive value 93%, negative predictive value 100%, positive likelihood ratio 3.33, negative likelihood ratio 0). Eight (21%) of the 38 HCC nodules, 7 of which lacked the typical vascular features at any of the imaging modalities, showed washout in the portal/venous phase and hypointensity in the hepatobiliary phase at MRI, while this pattern was not detected in any nonmalignant lesion.

CONCLUSIONS

Gadoxetic acid magnetic resonance may enhance the sensitivity of the non-invasive diagnosis of small hepatocellular carcinoma nodules in cirrhotic patients under surveillance. Double hypointensity in the portal/venous and hepatobiliary phases could be considered a new magnetic resonance pattern, highly suggestive of hypovascular hepatocellular carcinoma.

Keratin 19 epithelial patterns in cirrhotic stroma parallel hepatocarcinogenesis

Cirrhotic septa harbor vessels and inflammatory, fibrogenic, and ductular epithelial cells, collectively referred to as the ductular reaction (DR). Lack of the DR in the stromal compartment around hepatocellular carcinoma (HCC) has been documented; however, the relationship of epithelial keratin 19 (K19) structures to progression of intralesional carcinogenesis has not been explored. K19 immunoreactivity in the stromal compartment around 176 nodules in cirrhotic explants was examined. Quantitative differences (P < 0.0001) were manifested in three distinct histologically identifiable patterns: "complex" around cirrhotic nodules (CN), "attenuated" around dysplastic nodules (DN), and "absent" around HCC. Markers of necrosis or apoptosis could not explain the perinodular K19 epithelial loss; however, multicolor immunolabeling for K19, vimentin, E-Cadherin, SNAIL, and fibroblast-specific protein 1 (FSP-1) demonstrated discrepancies in immunophenotype and cytomorphologic features. Variability of cellular features was accompanied by an overall decrease in epithelial markers and significantly increased fractions of SNAIL- and FSP-1-positive cells in the DR around DN when compared with CN (P < 0.0001). Immunolabeling of transforming growth factor-β signaling components (TGFβR1, SMAD3, and pSMAD2/3) demonstrated increased percentages of pSMAD2/3 around DN when compared with CN (P < 0.0001). These findings collectively suggest marked alterations in cellular identity as an underlying mechanism for the reproducible extralesional K19 pattern that parallels progressive stages of intranodular hepatocarcinogenesis. Paracrine signaling is proposed as a link that emphasizes the importance of the epithelial-stromal compartment in malignant progression of HCC in cirrhosis.

Analysis of gadobenate dimeglumine-enhanced MR findings for characterizing small (1-2-cm) hepatic nodules in patients at high risk for hepatocellular carcinoma

PURPOSE

To retrospectively identify magnetic resonance (MR) imaging findings that are associated with hepatocellular carcinoma (HCC) in 1-2-cm nodules detected at surveillance ultrasonography (US) and to propose newer MR imaging diagnostic criteria.

MATERIALS AND METHODS

Institutional research ethics board approval was obtained, and informed patient consent was waived. Among 145 consecutive patients who had 1-2-cm nodules that were newly detected at surveillance US, 108 patients underwent gadobenate dimeglumine-enhanced MR imaging. After excluding hemangiomas and unconfirmed nodules, the study sample comprised 96 patients with 116 nodules, including 43 HCCs and 73 benign nodules. MR imaging findings were assessed for signal intensity at each sequence. On the basis of the results of univariate and multivariable logistic regression analyses, several diagnostic criteria were developed by using combinations of MR imaging findings, which were then compared with the American Association for the Study of Liver Diseases (AASLD) practice guideline.

RESULTS

Univariate analysis revealed four imaging findings associated with HCC, including arterial phase hyperintensity, portal or delayed phase hypointensity (washout), hyperintensity on T2-weighted images, and hepatobiliary phase hypointensity (P < .001 for each). In the multivariable analysis, arterial phase hyperintensity (adjusted odds ratio [OR], 17.1; P = .003) and washout (adjusted OR, 11.7; P = .007) were associated with HCC. Of the developed criteria, the criteria including nodules fitting the AASLD practice guideline (arterial phase hyperintensity and washout) or nodules having three or more findings were considered most reasonable, showing improved sensitivity (77% [33 of 43] versus 67% [29 of 43], P = .048) and comparable specificity (95% [69 of 73] versus 99% [72 of 73], P = .09), as compared with AASLD practice guideline.

CONCLUSION

Alternative MR imaging criteria for diagnosing HCC in 1-2-cm nodules detected at surveillance US that can improve sensitivity compared with the AASLD practice guideline were proposed. A larger study is needed to verify the preliminary criteria in this study.

SUPPLEMENTAL MATERIAL

http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.11101549/-/DC1.

Enhancement patterns and signal-intensity characteristics of small hepatocellular carcinoma in cirrhosis: pathologic basis and diagnostic challenges

Recent pathologic studies of hepatic resection and transplantation specimens have elucidated the morphologic features of the precancerous lesions and small hepatocellular carcinomas (HCCs) arising in cirrhotic livers. Small HCCs measuring less than 2 cm in diameter are of two types: vaguely nodular, well-differentiated tumors, also known as "early" HCCs, and distinctly nodular tumors, with histologic features of "classic" HCC. The precancerous lesions include dysplastic foci and dysplastic nodules. "Classic" small HCCs are supplied by nontriadal arteries, whereas early HCCs and dysplastic nodules may receive blood supply from both portal tracts and nontriadal arteries. The similarities in blood supply of these three types of nodular lesions result in significant overlap of findings on dynamic imaging. Nevertheless, small HCCs sometimes display characteristic radiologic features, such as "nodule-in-nodule" configuration and "corona enhancement" pattern. Moreover, various histologic features of these nodular lesions may also be related to a variety of signal intensities and attenuation coefficients, while the presence of cirrhosis is known to limit the sensitivity and specificity of any imaging modality, due to liver inhomogeneity. Because of these reasons, imaging findings of nodular lesions in cirrhotic livers are often inconclusive, emphasizing the need for a better understanding of these imaging features.

Sampling variability of computer-aided fractal-corrected measures of liver fibrosis in needle biopsy specimens

AIM: To assess the sampling variability of computer-aided, fractal-corrected measures of fibrosis in liver biopsies.

METHODS: Samples were derived from six to eight different parts of livers removed from 12 patients with clinically and histologically proven cirrhosis undergoing orthotopic liver transplantation. Sirius red-stained sections with a thickness of 2 μm were digitized using a computer-aided image analysis system that automatically measures the surface of fibrosis, as well as its outline perimeter, fractal surface and outline dimensions, wrinkledness, and Hurst coefficient.

RESULTS: We found a high degree of inter-sample variability in the measurements of the surface [coefficient of variation (CV) = 43% ± 13%] and wrinkledness (CV = 28% ± 9%) of fibrosis, but the inter-sample variability of Hurst’s exponent was low (CV = 14% ± 2%).

CONCLUSION: This study suggests that Hurst’s exponent might be used in clinical practice as the best histological estimate of fibrosis in the whole organ, and evidences the fact that biopsy sections, which are fundamental for the qualitative diagnosis of chronic hepatitis, play a key role in the quantitative estimate of architectural changes in liver tissue.