Deep Learning Classification and Quantification of Pejorative and Nonpejorative Architectures in Resected Hepatocellular Carcinoma from Digital Histopathologic Images.

Liver resection is one of the best treatments for small hepatocellular carcinoma (HCC), but post-resection recurrence is frequent. Biotherapies have emerged as an efficient adjuvant treatment, making the identification of patients at high risk of recurrence critical. Microvascular invasion (mVI), poor differentiation, pejorative macrotrabecular architectures, and vessels encapsulating tumor clusters architectures are the most accurate histologic predictors of recurrence, but their evaluation is time-consuming and imperfect. A supervised deep learning-based approach with ResNet34 on 680 whole slide images (WSIs) from 107 liver resection specimens allowed us to build an algorithm for the identification and quantification of these pejorative architectures. This model achieved an accuracy of 0.864 at patch level and 0.823 at WSI level. To assess its robustness, it was validated on an external cohort of 29 HCCs from another hospital, with an accuracy of 0.787 at WSI level, affirming its generalization capabilities. Moreover, the largest connected areas of the pejorative architectures extracted from the model were positively correlated to the presence of mVI and the number of tumor emboli. These results suggest that the identification of pejorative architectures could be an efficient surrogate of mVI and have a strong predictive value for the risk of recurrence. This study is the first step in the construction of a composite predictive algorithm for early post-resection recurrence of HCC, including artificial intelligence-based features.

Integration of new technologies in the multidisciplinary approach to primary liver tumours: The next-generation tumour board.

Primary liver tumours, including benign liver tumours, hepatocellular carcinoma and cholangiocarcinoma, present a multifaceted challenge, necessitating a collaborative approach, as evidenced by the role of the multidisciplinary tumour board (MDTB). The approach to managing primary liver tumours involves specialised teams, including surgeons, radiologists, oncologists, pathologists, hepatologists, and radiation oncologists, coming together to propose individualised treatment plans. The evolving landscape of primary liver cancer treatment introduces complexities, particularly with the expanding array of systemic and locoregional therapies, alongside the potential integration of molecular biology and artificial intelligence (AI) into MDTBs in the future. Precision medicine demands collaboration across disciplines, challenging traditional frameworks. In the next decade, we anticipate the convergence of AI, molecular biology, pathology, and advanced imaging, requiring adaptability in MDTB structure to incorporate these cutting-edge technologies. Navigating this evolution also requires a focus on enhancing basic, translational, and clinical research, as well as boosting clinical trials through an upgraded use of MDTBs as hubs for scientific collaboration and raising literacy about AI and new technologies. In this review, we will delineate the current unmet needs in the clinical management of primary liver cancers, discuss our perspective on the future role of MDTBs in primary liver cancers ("next generation" MDTBs), and unravel the potential power and limitations of novel technologies that may shape the multidisciplinary care landscape for primary liver cancers in the coming decade.

Association of LI-RADS and Histopathologic Features with Survival in Patients with Solitary Resected Hepatocellular Carcinoma.

Background Both Liver Imaging Reporting and Data System (LI-RADS) and histopathologic features provide prognostic information in patients with hepatocellular carcinoma (HCC), but whether LI-RADS is independently associated with survival is uncertain. Purpose To assess the association of LI-RADS categories and features with survival outcomes in patients with solitary resected HCC. Materials and Methods This retrospective study included patients with solitary resected HCC from three institutions examined with preoperative contrast-enhanced CT and/or MRI between January 2008 and December 2019. Three independent readers evaluated the LI-RADS version 2018 categories and features. Histopathologic features including World Health Organization tumor grade, microvascular and macrovascular invasion, satellite nodules, and tumor capsule were recorded. Overall survival and disease-free survival were assessed with Cox regression models. Marginal effects of nontargetoid features on survival were estimated using propensity score matching. Results A total of 360 patients (median age, 64 years [IQR, 56-70 years]; 280 male patients) were included. At CT and MRI, the LI-RADS LR-M category was associated with increased risk of recurrence (CT: hazard ratio [HR] = 1.83 [95% CI: 1.26, 2.66], P = .001; MRI: HR = 2.22 [95% CI: 1.56, 3.16], P < .001) and death (CT: HR = 2.47 [95% CI: 1.72, 3.55], P < .001; MRI: HR = 1.80 [95% CI: 1.32, 2.46], P < .001) independently of histopathologic features. The presence of at least one nontargetoid feature was associated with an increased risk of recurrence (CT: HR = 1.80 [95% CI: 1.36, 2.38], P < .001; MRI: HR = 1.93 [95% CI: 1.81, 2.06], P < .001) and death (CT: HR = 1.51 [95% CI: 1.10, 2.07], P < .010) independently of histopathologic features. In matched samples, recurrence was associated with the presence of at least one nontargetoid feature at CT (HR = 2.06 [95% CI: 1.15, 3.66]; P = .02) or MRI (HR = 1.79 [95% CI: 1.01, 3.20]; P = .048). Conclusion In patients with solitary resected HCC, LR-M category and nontargetoid features were negatively associated with survival independently of histopathologic characteristics. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Kartalis and Grigoriadis in this issue.

Artificial intelligence in liver cancer - new tools for research and patient management.

Liver cancer has high incidence and mortality globally. Artificial intelligence (AI) has advanced rapidly, influencing cancer care. AI systems are already approved for clinical use in some tumour types (for example, colorectal cancer screening). Crucially, research demonstrates that AI can analyse histopathology, radiology and natural language in liver cancer, and can replace manual tasks and access hidden information in routinely available clinical data. However, for liver cancer, few of these applications have translated into large-scale clinical trials or clinically approved products. Here, we advocate for the incorporation of AI in all stages of liver cancer management. We present a taxonomy of AI approaches in liver cancer, highlighting areas with academic and commercial potential, and outline a policy for AI-based liver cancer management, including interdisciplinary training of researchers, clinicians and patients. The potential of AI in liver cancer is immense, but effort is required to ensure that AI can fulfil expectations.

Combining a deep learning model with clinical data better predicts hepatocellular carcinoma behavior following surgery.

Hepatocellular carcinoma (HCC) is among the most common cancers worldwide, and tumor recurrence following liver resection or transplantation is one of the highest contributors to mortality in HCC patients after surgery. Using artificial intelligence (AI), we developed an interdisciplinary model to predict HCC recurrence and patient survival following surgery. We collected whole-slide H&E images, clinical variables, and follow-up data from 300 patients with HCC who underwent transplant and 169 patients who underwent resection at the Cleveland Clinic. A deep learning model was trained to predict recurrence-free survival (RFS) and disease-specific survival (DSS) from the H&E-stained slides. Repeated cross-validation splits were used to compute robust C-index estimates, and the results were compared to those obtained by fitting a Cox proportional hazard model using only clinical variables. While the deep learning model alone was predictive of recurrence and survival among patients in both cohorts, integrating the clinical and histologic models significantly increased the C-index in each cohort. In every subgroup analyzed, we found that a combined clinical and deep learning model better predicted post-surgical outcome in HCC patients compared to either approach independently.

Macroscopic Characterization of Hepatocellular Carcinoma: An Underexploited Source of Prognostic Factors.

The macroscopic appearance of a tumor such as hepatocellular carcinoma (HCC) may be defined as its phenotype which is de facto dictated by its genotype. Therefore, macroscopic characteristics of HCC are unlikely random but rather reflect genomic traits of cancer, presumably acting as a valuable source of information that can be retrieved and exploited to infer prognosis. This review aims to provide a comprehensive overview of the available data on the prognostic value of macroscopic characterization in HCC. A total of 57 studies meeting eligible criteria were identified, including patients undergoing liver resection (LR; 47 studies, 83%) or liver transplant (LT; 9 studies, 16%). The following macroscopic variables were investigated: tumor size (n = 42 studies), number of nodules (n = 28), vascular invasion (n = 24), bile duct invasion (n = 6), growth pattern (n = 15), resection margin (n = 11), tumor location (n = 6), capsule (n = 2) and satellite (n = 1). Although the selected studies provided insightful data with notable prognostic performances, a lack of standardization and substantial gaps were noted in the report and the analysis of gross findings. This topic remains incompletely covered. While the available studies underscored the value of macroscopic variables in HCC prognostication, important lacks were also observed. Macroscopic characterization of HCC is likely an underexploited source of prognostic factors that must be actively explored by future multidisciplinary research.

Should Hypervascular Incidentalomas Detected on Per-Interventional Cone Beam Computed Tomography during Intra-Arterial Therapies for Hepatocellular Carcinoma Impact the Treatment Plan in Patients Waiting for Liver Transplantation?

BACKGROUND: Current guidelines do not indicate any comprehensive management of hepatic hypervascular incidentalomas (HVIs) discovered in hepatocellular carcinoma (HCC) patients during intra-arterial therapies (IATs). This study aims to evaluate the prognostic value of HVIs detected on per-interventional cone beam computed tomography (CBCT) during IAT for HCC in patients waiting for liver transplantation (LT). MATERIAL AND METHODS: In this retrospective single-institutional study, all liver-transplanted HCC patients between January 2014 and December 2018 who received transarterial chemoembolization (TACE) or radioembolization (TARE) before LT were included. The number of ≥10 mm HCCs diagnosed on contrast-enhanced pre-interventional imaging (PII) was compared with that detected on per-interventional CBCT with a nonparametric Wilcoxon test. The correlation between the presence of an HVI and histopathological criteria associated with poor prognosis (HPP) on liver explants was investigated using the chi-square test. Tumor recurrence (TR) and TR-related mortality were investigated using the chi-square test. Recurrence-free survival (RFS), TR-related survival (TRRS), and overall survival (OS) were assessed according to the presence of HVI using Kaplan-Meier analysis. RESULTS: Among 63 included patients (average age: 59 ± 7 years, H/F = 50/13), 36 presented HVIs on per-interventional CBCT. The overall nodule detection rate of per-interventional CBCT was superior to that of PII (median at 3 [Q1:2, Q3:5] vs. 2 [Q1:1, Q3:3], respectively, p < 0.001). No significant correlation was shown between the presence of HVI and HPP (p = 0.34), TR (p = 0.095), and TR-related mortality (0.22). Kaplan-Meier analysis did not show a significant impact of the presence of HVI on RFS (p = 0.07), TRRS (0.48), or OS (p = 0.14). CONCLUSIONS: These results may indicate that the treatment plan during IAT should not be impacted or modified in response to HVI detection.

Transcriptomic profiling of intermediate cell carcinoma of the liver.

BACKGROUND: Intermediate cell carcinoma (Int-CA) is a rare and enigmatic primary liver cancer characterized by uniform tumor cells exhibiting mixed features of both HCC and intrahepatic cholangiocarcinoma. Despite the unique pathological features of int-CA, its molecular characteristics remain unclear yet. METHODS: RNA sequencing and whole genome sequencing profiling were performed on int-CA tumors and compared with those of HCC and intrahepatic cholangiocarcinoma. RESULTS: Int-CAs unveiled a distinct and intermediate transcriptomic feature that is strikingly different from both HCC and intrahepatic cholangiocarcinoma. The marked abundance of splicing events leading to intron retention emerged as a signature feature of int-CA, along with a prominent expression of Notch signaling. Further exploration revealed that METTL16 was suppressed within int-CA, showing a DNA copy number-dependent transcriptional deregulation. Notably, experimental investigations confirmed that METTL16 suppression facilitated invasive tumor characteristics through the activation of the Notch signaling cascade. CONCLUSIONS: Our results provide a molecular landscape of int-CA featured by METTL16 suppression and frequent intron retention events, which may play pivotal roles in the acquisition of the aggressive phenotype of Int-CA.

Weakly Supervised Deep Learning Predicts Immunotherapy Response in Solid Tumors Based on PD-L1 Expression.

Programmed death-ligand 1 (PD-L1) IHC is the most commonly used biomarker for immunotherapy response. However, quantification of PD-L1 status in pathology slides is challenging. Neither manual quantification nor a computer-based mimicking of manual readouts is perfectly reproducible, and the predictive performance of both approaches regarding immunotherapy response is limited. In this study, we developed a deep learning (DL) method to predict PD-L1 status directly from raw IHC image data, without explicit intermediary steps such as cell detection or pigment quantification. We trained the weakly supervised model on PD-L1-stained slides from the non-small cell lung cancer (NSCLC)-Memorial Sloan Kettering (MSK) cohort (N = 233) and validated it on the pan-cancer-Vall d'Hebron Institute of Oncology (VHIO) cohort (N = 108). We also investigated the performance of the model to predict response to immune checkpoint inhibitors (ICI) in terms of progression-free survival. In the pan-cancer-VHIO cohort, the performance was compared with tumor proportion score (TPS) and combined positive score (CPS). The DL model showed good performance in predicting PD-L1 expression (TPS ≥ 1%) in both NSCLC-MSK and pan-cancer-VHIO cohort (AUC 0.88 ± 0.06 and 0.80 ± 0.03, respectively). The predicted PD-L1 status showed an improved association with response to ICIs [HR: 1.5 (95% confidence interval: 1-2.3), P = 0.049] compared with TPS [HR: 1.4 (0.96-2.2), P = 0.082] and CPS [HR: 1.2 (0.79-1.9), P = 0.386]. Notably, our explainability analysis showed that the model does not just look at the amount of brown pigment in the IHC slides, but also considers morphologic factors such as lymphocyte conglomerates. Overall, end-to-end weakly supervised DL shows potential for improving patient stratification for cancer immunotherapy by analyzing PD-L1 IHC, holistically integrating morphology and PD-L1 staining intensity. SIGNIFICANCE: The weakly supervised DL model to predict PD-L1 status from raw IHC data, integrating tumor staining intensity and morphology, enables enhanced patient stratification in cancer immunotherapy compared with traditional pathologist assessment.

Regression-based Deep-Learning predicts molecular biomarkers from pathology slides.

Deep Learning (DL) can predict biomarkers from cancer histopathology. Several clinically approved applications use this technology. Most approaches, however, predict categorical labels, whereas biomarkers are often continuous measurements. We hypothesize that regression-based DL outperforms classification-based DL. Therefore, we develop and evaluate a self-supervised attention-based weakly supervised regression method that predicts continuous biomarkers directly from 11,671 images of patients across nine cancer types. We test our method for multiple clinically and biologically relevant biomarkers: homologous recombination deficiency score, a clinically used pan-cancer biomarker, as well as markers of key biological processes in the tumor microenvironment. Using regression significantly enhances the accuracy of biomarker prediction, while also improving the predictions' correspondence to regions of known clinical relevance over classification. In a large cohort of colorectal cancer patients, regression-based prediction scores provide a higher prognostic value than classification-based scores. Our open-source regression approach offers a promising alternative for continuous biomarker analysis in computational pathology.

Selective internal radiation therapy for unresectable HCC: The SIRT downstaging study.

BACKGROUND: Selective internal radiation therapy (SIRT) is recommended as a downstaging (DS) strategy for solitary unresectable HCC <8 cm. The aim of this study was to report the results of acquired experience in a tertiary center for all unresectable HCCs. METHODS: We conducted a retrospective, observational study using data collected from consecutive patients undergoing SIRT between October 2013 and June 2020. DS was considered achieved when a curative treatment could be proposed 6 months after SIRT. RESULTS: One hundred twenty-seven patients were included (male = 90%, 64 ± 11 y), of whom 112 (n = 88%) had cirrhosis. HCC was classified as BCLC stage C in 64 patients (50%), with a median diameter of 61 mm, an infiltrative pattern in 51 patients (40%), and portal vein invasion in 62 (49%) patients. Fifty patients (39%) achieved DS 6 months following SIRT, with 29 of them (23%) undergoing curative treatment in a median time of 4.3 months: 17 (13%) were transplanted, 11 (85%) had liver resection, and 1 patient had a radiofrequency ablation. The median overall survival of patients with or without DS was 51 versus 10 months, respectively (p < 0.001). In patients who achieved DS, progression-free survival was higher in patients who underwent surgery: 47 versus 11 months (p < 0.001). Four variables were independently associated with DS: age (OR: 0.96, 95% CI: [0.92, 0.99]; p = 0.032), baseline α-fetoprotein (OR: 1.00, 95% CI: [1.00, 1.00]; p = 0.034), HCC distribution (OR: 0.3, 95% CI: [0.11, 0.75]; p = 0.012), and ALBI grade (OR: 0.34. 95% CI: [0.14, 0.80]; p = 0.014). CONCLUSIONS: These results suggest that SIRT in patients with unresectable HCC could be an effective treatment: DS was achieved for around 39% of the patients and more than half of these then underwent curative treatment.

Deep learning for dual detection of microsatellite instability and POLE mutations in colorectal cancer histopathology.

In the spectrum of colorectal tumors, microsatellite-stable (MSS) tumors with DNA polymerase ε (POLE) mutations exhibit a hypermutated profile, holding the potential to respond to immunotherapy similarly to their microsatellite-instable (MSI) counterparts. Yet, due to their rarity and the associated testing costs, systematic screening for these mutations is not commonly pursued. Notably, the histopathological phenotype resulting from POLE mutations is theorized to resemble that of MSI. This resemblance not only could facilitate their detection by a transformer-based Deep Learning (DL) system trained on MSI pathology slides, but also indicates the possibility for MSS patients with POLE mutations to access enhanced treatment options, which might otherwise be overlooked. To harness this potential, we trained a Deep Learning classifier on a large dataset with the ground truth for microsatellite status and subsequently validated its capabilities for MSI and POLE detection across three external cohorts. Our model accurately identified MSI status in both the internal and external resection cohorts using pathology images alone. Notably, with a classification threshold of 0.5, over 75% of POLE driver mutant patients in the external resection cohorts were flagged as "positive" by a DL system trained on MSI status. In a clinical setting, deploying this DL model as a preliminary screening tool could facilitate the efficient identification of clinically relevant MSI and POLE mutations in colorectal tumors, in one go.