Current updates in machine learning in the prediction of therapeutic outcome of hepatocellular carcinoma: what should we know?

A Machine Learning Algorithm Facilitates Prognosis Prediction and Treatment Selection for Barcelona Clinic Liver Cancer Stage C Hepatocellular Carcinoma

PURPOSE

Given its heterogeneity and diverse clinical outcomes, precise subclassification of Barcelona Clinic Liver Cancer stage C (BCLC-C) hepatocellular carcinoma (HCC) is required for appropriately determining patient prognosis and selecting treatment.

EXPERIMENTAL DESIGN

We recruited 2,626 patients with BCLC-C HCC from multiple centers, comprising training/test (n = 1,693) and validation cohorts (n = 933). The XGBoost model was chosen for maximum performance among the machine learning (ML) models. Patients were categorized into low-, intermediate-, high-, and very high-risk subgroups based on the estimated prognosis, and this subclassification was named the CLAssification via Machine learning of BCLC-C (CLAM-C).

RESULTS

The areas under the receiver operating characteristic curve of the CLAM-C for predicting the 6-, 12-, and 24-month survival of patients with BCLC-C were 0.800, 0.831, and 0.715, respectively-significantly higher than those of the conventional models, which were consistent in the validation cohort. The four subgroups had significantly different median overall survivals, and this difference was maintained among various patient subgroups and treatment modalities. Immune-checkpoint inhibitors and transarterial therapies were associated with significantly better survival than tyrosine kinase inhibitors (TKI) in the low- and intermediate-risk subgroups. In cases with first-line systemic therapy, the CLAM-C identified atezolizumab-bevacizumab as the best therapy, particularly in the high-risk group. In cases with later-line systemic therapy, nivolumab had better survival than TKIs in the low-to-intermediate-risk subgroup, whereas TKIs had better survival in the high- to very high-risk subgroup.

CONCLUSIONS

ML modeling effectively subclassified patients with BCLC-C HCC, potentially aiding treatment allocation. Our study underscores the potential utilization of ML modeling in terms of prognostication and treatment allocation in patients with BCLC-C HCC.

Radiomics and machine learning based on preoperative MRI for predicting extrahepatic metastasis in hepatocellular carcinoma patients treated with transarterial chemoembolization

Purpose

To develop and validate a radiomics machine learning (Rad-ML) model based on preoperative MRI to predict extrahepatic metastasis (EHM) in hepatocellular carcinoma (HCC) patients receiving transarterial chemoembolization (TACE) treatment.

Methods

A total of 355 HCC patients who received multiple TACE procedures were split at random into a training set and a test set at a 7:3 ratio. Radiomic features were calculated from tumor and peritumor in arterial phase and portal venous phase, and were identified using intraclass correlation coefficient, maximal relevance and minimum redundancy, and least absolute shrinkage and selection operator techniques. Cox regression analysis was employed to determine the clinical model. The best-performing algorithm among eight machine learning methods was used to construct the Rad-ML model. A nomogram combining clinical and Rad-ML parameters was used to develop a combined model. Model performance was evaluated using C-index, decision curve analysis, calibration plot, and survival analysis.

Results

In clinical model, elevated neutrophil to lymphocyte ratio and alpha-fetoprotein were associated with faster EHM. The XGBoost-based Rad-ML model demonstrated the best predictive performance for EHM. When compared to the clinical model, both the Rad-ML model and the combination model performed better (C-indexes of 0.61, 0.85, and 0.86 in the training set, and 0.62, 0.82, and 0.83 in the test set, respectively). However, the combined model's and the Rad-ML model's prediction performance did not differ significantly. The most influential feature was peritumoral waveletHLL_firstorder_Minimum in AP, which exhibited an inverse relationship with EHM risk.

Conclusions

Our study suggests that the preoperative MRI-based Rad-ML model is a valuable tool to predict EHM in HCC patients treated with TACE.

Application of interpretable machine learning algorithms to predict distant metastasis in ovarian clear cell carcinoma

BACKGROUND

Ovarian clear cell carcinoma (OCCC) represents a subtype of ovarian epithelial carcinoma (OEC) known for its limited responsiveness to chemotherapy, and the onset of distant metastasis significantly impacts patient prognoses. This study aimed to identify potential risk factors contributing to the occurrence of distant metastasis in OCCC.

METHODS

Utilizing the Surveillance, Epidemiology, and End Results (SEER) database, we identified patients diagnosed with OCCC between 2004 and 2015. The most influential factors were selected through the application of Gaussian Naive Bayes (GNB) and Adaboost machine learning algorithms, employing a Venn test for further refinement. Subsequently, six machine learning (ML) techniques, namely XGBoost, LightGBM, Random Forest (RF), Adaptive Boosting (Adaboost), Support Vector Machine (SVM), and Multilayer Perceptron (MLP), were employed to construct predictive models for distant metastasis. Shapley Additive Interpretation (SHAP) analysis facilitated a visual interpretation for individual patient. Model validity was assessed using accuracy, sensitivity, specificity, positive predictive value, negative predictive value, F1 score, and the area under the receiver operating characteristic curve (AUC).

RESULTS

In the realm of predicting distant metastasis, the Random Forest (RF) model outperformed the other five machine learning algorithms. The RF model demonstrated accuracy, sensitivity, specificity, positive predictive value, negative predictive value, F1 score, and AUC (95% CI) values of 0.792 (0.762-0.823), 0.904 (0.835-0.973), 0.759 (0.731-0.787), 0.221 (0.186-0.256), 0.974 (0.967-0.982), 0.353 (0.306-0.399), and 0.834 (0.696-0.967), respectively, surpassing the performance of other models. Additionally, the calibration curve's Brier Score (95%) for the RF model reached the minimum value of 0.06256 (0.05753-0.06759). SHAP analysis provided independent explanations, reaffirming the critical clinical factors associated with the risk of metastasis in OCCC patients.

CONCLUSIONS

This study successfully established a precise predictive model for OCCC patient metastasis using machine learning techniques, offering valuable support to clinicians in making informed clinical decisions.

From prediction to prevention: Machine learning revolutionizes hepatocellular carcinoma recurrence monitoring

In this editorial, we comment on the article by Zhang et al entitled Development of a machine learning-based model for predicting the risk of early postoperative recurrence of hepatocellular carcinoma. Hepatocellular carcinoma (HCC), which is characterized by high incidence and mortality rates, remains a major global health challenge primarily due to the critical issue of postoperative recurrence. Early recurrence, defined as recurrence that occurs within 2 years posttreatment, is linked to the hidden spread of the primary tumor and significantly impacts patient survival. Traditional predictive factors, including both patient- and treatment-related factors, have limited predictive ability with respect to HCC recurrence. The integration of machine learning algorithms is fueled by the exponential growth of computational power and has revolutionized HCC research. The study by Zhang et al demonstrated the use of a groundbreaking preoperative prediction model for early postoperative HCC recurrence. Chall-enges persist, including sample size constraints, issues with handling data, and the need for further validation and interpretability. This study emphasizes the need for collaborative efforts, multicenter studies and comparative analyses to validate and refine the model. Overcoming these challenges and exploring innovative approaches, such as multi-omics integration, will enhance personalized oncology care. This study marks a significant stride toward precise, effi-cient, and personalized oncology practices, thus offering hope for improved patient outcomes in the field of HCC treatment.

A machine learning model for predicting hepatocellular carcinoma risk in patients with chronic hepatitis B

BACKGROUND

Machine learning (ML) algorithms can be used to overcome the prognostic performance limitations of conventional hepatocellular carcinoma (HCC) risk models. We established and validated an ML-based HCC predictive model optimized for patients with chronic hepatitis B (CHB) infections receiving antiviral therapy (AVT).

METHODS

Treatment-naïve CHB patients who were started entecavir (ETV) or tenofovir disoproxil fumarate (TDF) were enrolled. We used a training cohort (n = 960) to develop a novel ML model that predicted HCC development within 5 years and validated the model using an independent external cohort (n = 1937). ML algorithms consider all potential interactions and do not use predefined hypotheses.

RESULTS

The mean age of the patients in the training cohort was 48 years, and most patients (68.9%) were men. During the median 59.3 (interquartile range 45.8-72.3) months of follow-up, 69 (7.2%) patients developed HCC. Our ML-based HCC risk prediction model had an area under the receiver-operating characteristic curve (AUC) of 0.900, which was better than the AUCs of CAMD (0.778) and REAL B (0.772) (both p < .05). The better performance of our model was maintained (AUC = 0.872 vs. 0.788 for CAMD and 0.801 for REAL B) in the validation cohort. Using cut-off probabilities of 0.3 and 0.5, the cumulative incidence of HCC development differed significantly among the three risk groups (p < .001).

CONCLUSIONS

Our new ML model performed better than models in terms of predicting the risk of HCC development in CHB patients receiving AVT.

From Reductionistic Approach to Systems Immunology Approach for the Understanding of Tumor Microenvironment

The tumor microenvironment (TME) is a complex and dynamic ecosystem that includes a variety of immune cells mutually interacting with tumor cells, structural/stromal cells, and each other. The immune cells in the TME can have dual functions as pro-tumorigenic and anti-tumorigenic. To understand such paradoxical functions, the reductionistic approach classifies the immune cells into pro- and anti-tumor cells and suggests the therapeutic blockade of the pro-tumor and induction of the anti-tumor immune cells. This strategy has proven to be partially effective in prolonging patients' survival only in a fraction of patients without offering a cancer cure. Recent advances in multi-omics allow taking systems immunology approach. This essay discusses how a systems immunology approach could revolutionize our understanding of the TME by suggesting that internetwork interactions of the immune cell types create distinct collective functions independent of the function of each cellular constituent. Such collective function can be understood by the discovery of the immunological patterns in the TME and may be modulated as a therapeutic means for immunotherapy of cancer.

Deep learning nomogram based on Gd-EOB-DTPA MRI for predicting early recurrence in hepatocellular carcinoma after hepatectomy

OBJECTIVES

The accurate prediction of post-hepatectomy early recurrence in patients with hepatocellular carcinoma (HCC) is crucial for decision-making regarding postoperative adjuvant treatment and monitoring. We aimed to explore the feasibility of deep learning (DL) features derived from gadoxetate disodium (Gd-EOB-DTPA) MRI, qualitative features, and clinical variables for predicting early recurrence.

METHODS

In this bicentric study, 285 patients with HCC who underwent Gd-EOB-DTPA MRI before resection were divided into training (n = 195) and validation (n = 90) sets. DL features were extracted from contrast-enhanced MRI images using VGGNet-19. Three feature selection methods and five classification methods were combined for DL signature construction. Subsequently, an mp-MR DL signature fused with multiphase DL signatures of contrast-enhanced images was constructed. Univariate and multivariate logistic regression analyses were used to identify early recurrence risk factors including mp-MR DL signature, microvascular invasion (MVI), and tumor number. A DL nomogram was built by incorporating deep features and significant clinical variables to achieve early recurrence prediction.

RESULTS

MVI (p = 0.039), tumor number (p = 0.001), and mp-MR DL signature (p < 0.001) were independent risk factors for early recurrence. The DL nomogram outperformed the clinical nomogram in the training set (AUC: 0.949 vs. 0.751; p < 0.001) and validation set (AUC: 0.909 vs. 0.715; p = 0.002). Excellent DL nomogram calibration was achieved in both training and validation sets. Decision curve analysis confirmed the clinical usefulness of DL nomogram.

CONCLUSION

The proposed DL nomogram was superior to the clinical nomogram in predicting early recurrence for HCC patients after hepatectomy.

KEY POINTS

• Deep learning signature based on Gd-EOB-DTPA MRI was the predominant independent predictor of early recurrence for hepatocellular carcinoma (HCC) after hepatectomy. • Deep learning nomogram based on clinical factors and Gd-EOB-DTPA MRI features is promising for predicting early recurrence of HCC. • Deep learning nomogram outperformed the conventional clinical nomogram in predicting early recurrence.

Machine Learning to Predict the Adsorption Capacity of Microplastics

Nowadays, there is an extensive production and use of plastic materials for different industrial activities. These plastics, either from their primary production sources or through their own degradation processes, can contaminate ecosystems with micro- and nanoplastics. Once in the aquatic environment, these microplastics can be the basis for the adsorption of chemical pollutants, favoring that these chemical pollutants disperse more quickly in the environment and can affect living beings. Due to the lack of information on adsorption, three machine learning models (random forest, support vector machine, and artificial neural network) were developed to predict different microplastic/water partition coefficients (log K_d) using two different approximations (based on the number of input variables). The best-selected machine learning models present, in general, correlation coefficients above 0.92 in the query phase, which indicates that these types of models could be used for the rapid estimation of the absorption of organic contaminants on microplastics.

Stemness Analysis Uncovers That The Peroxisome Proliferator-Activated Receptor Signaling Pathway Can Mediate Fatty Acid Homeostasis In Sorafenib-Resistant Hepatocellular Carcinoma Cells

Hepatocellular carcinoma (HCC) stem cells are regarded as an important part of individualized HCC treatment and sorafenib resistance. However, there is lacking systematic assessment of stem-like indices and associations with a response of sorafenib in HCC. Our study thus aimed to evaluate the status of tumor dedifferentiation for HCC and further identify the regulatory mechanisms under the condition of resistance to sorafenib. Datasets of HCC, including messenger RNAs (mRNAs) expression, somatic mutation, and clinical information were collected. The mRNA expression-based stemness index (mRNAsi), which can represent degrees of dedifferentiation of HCC samples, was calculated to predict drug response of sorafenib therapy and prognosis. Next, unsupervised cluster analysis was conducted to distinguish mRNAsi-based subgroups, and gene/geneset functional enrichment analysis was employed to identify key sorafenib resistance-related pathways. In addition, we analyzed and confirmed the regulation of key genes discovered in this study by combining other omics data. Finally, Luciferase reporter assays were performed to validate their regulation. Our study demonstrated that the stemness index obtained from transcriptomic is a promising biomarker to predict the response of sorafenib therapy and the prognosis in HCC. We revealed the peroxisome proliferator-activated receptor signaling pathway (the PPAR signaling pathway), related to fatty acid biosynthesis, that was a potential sorafenib resistance pathway that had not been reported before. By analyzing the core regulatory genes of the PPAR signaling pathway, we identified four candidate target genes, retinoid X receptor beta (RXRB), nuclear receptor subfamily 1 group H member 3 (NR1H3), cytochrome P450 family 8 subfamily B member 1 (CYP8B1) and stearoyl-CoA desaturase (SCD), as a signature to distinguish the response of sorafenib. We proposed and validated that the RXRB and NR1H3 could directly regulate NR1H3 and SCD, respectively. Our results suggest that the combined use of SCD inhibitors and sorafenib may be a promising therapeutic approach.

A Machine Learning Model Based on Health Records for Predicting Recurrence After Microwave Ablation of Hepatocellular Carcinoma

Background and Aim

Early recurrence (ER) presents a challenge for the survival prognosis of patients with hepatocellular carcinoma (HCC). The aim of this study was to investigate machine learning (ML) models using clinical data for predicting ER after microwave ablation (MWA).

Methods

Between August 2005 and December 2019, 1574 patients with early-stage HCC underwent MWA at four hospitals were reviewed. Then, 36 clinical data points per patient were collected, and the patients were assigned to the training, internal, and external validation set. Apart from traditional logistic regression (LR), three ML models—random forest, support vector machine, and eXtreme Gradient Boosting (XGBoost)—were built and validated for their predictive ability with the area under ROC curve (AUC). Algorithms such as SHapley Additive exPlanations (SHAP) and local interpretable model-agnostic explanations (LIME) were used to realize their interpretability.

Results

The three ML models all outperformed LR (P < 0.001 for all) in predictive ability. When nine variables (tumor number, platelet, α-fetoprotein, comorbidity score, white blood cell, cholinesterase, prothrombin time, neutrophils, and etiology) were extracted simultaneously using recursive feature elimination with cross-validation, the XGBoost model achieved the best discrimination among all models, with an AUC value 0.75 (95% CI [confidence interval]: 0.72–0.78) in the training set, 0.74 (95% CI: 0.69–0.80) in the internal validation set, and 0.76 (95% CI: 0.70–0.82) in the external validation set, and it was interpreted depending on the visualization of risk factors by the SHAP and LIME algorithms. The predictive system of post-ablation recurrence risk stratification was provided on online (http://114.251.235.51:8001/) based on XGboost analysis.

Conclusion

The XGBoost model based on clinical data can effectively predict ER risk after MWA, which can contribute to surveillance, prevention, and treatment strategies for HCC.

Combined transarterial iodized oil injection and computed tomography-guided thermal ablation for hepatocellular carcinoma: utility of the iodized oil retention pattern

Purpose

To investigate whether the iodized oil (Lipiodol, Guerbet Group, Villepinte, France) retention pattern influences the treatment efficacy of combined transarterial Lipiodol injection (TLI) and thermal ablation in patients with hepatocellular carcinoma (HCC).

Methods

Data of 198 patients (280 HCC lesions), who underwent TLI plus computed tomography (CT)-guided thermal ablation at three separate medical institutions between June 2014 and September 2020, were reviewed and analyzed. The Lipiodol retention pattern was classified as complete or incomplete based on non-enhanced CT at the time of ablation. The primary outcome was local recurrence-free survival (LRFS) for lesions; the secondary outcome was overall survival (OS) for patients. Propensity score matching (PSM) was performed using a caliper width of 0.1 between the two groups. Differences in LRFS and OS between the two groups were compared using the log-rank test.

Results

A total of 133 lesions exhibited a complete Lipiodol retention pattern, while 147 exhibited an incomplete pattern. After PSM analysis of baseline characteristics of the lesions, 121 pairs of lesions were matched. LRFS was significantly longer for lesions exhibiting complete retention than for those exhibiting incomplete retention (P = 0.030). After PSM analysis of patient baseline characteristics, 74 pairs of patients were matched. There was no significant difference in OS between the two groups (P = 0.456).

Conclusion

Lipiodol retention patterns may influence the treatment efficacy of combined TLI and thermal ablation for HCC lesions. However, a survival benefit for the Lipiodol retention pattern among HCC patients was not observed and needs further confirmation.

Tumor enhancement ratio with unenhanced imaging is an independent prognostic factor for patients with hepatocellular carcinoma after transarterial chemoembolization

Objective

To investigative whether the odds tumor enhancement ratio (OTER) on cross-sectional imaging is a prognostic factor for hepatocellular carcinoma after transarterial chemoembolization (TACE).

Methods

This study involved 126 patients who underwent TACE from May 2015 to March 2019. The signal intensity/Hounsfield units (HU) was measured by placing regions of interest on the tumor and surrounding liver in unenhanced and arterial-phase contrast-enhanced cross-sectional images. The OTER was calculated as follows: OTER = (HU_TUMORart − HU_TUMORun)/ (HU_LIVERart − HU_LIVERun). Univariate analysis was performed to determine the factors associated with overall survival (OS). Variables with a P value of <0.10 were included in the multivariate Cox regression analysis.

Results

The median OS was 757 days. Tumors with a peripheral location, small size, and low OTER had better OS than those with a central location, large size, and high OTER. OS did not differ according to the extent of tumor involvement or tumor enhancement pattern. The OTER, tumor location, and size were included in the multivariate Cox regression analysis. A low OTER was the predictor of better OS.

Conclusion

A high OTER is a risk factor for poor OS in patients undergoing TACE. This should be taken into consideration before the procedure.

Predicting early refractoriness of transarterial chemoembolization in patients with hepatocellular carcinoma using a random forest algorithm: A pilot study

Purpose: To develop and validate a random forest (RF) based predictive model of early refractoriness to transarterial chemoembolization (TACE) in patients with unresectable hepatocellular carcinoma (HCC).

Methods: A total of 227 patients with unresectable HCC who initially treated with TACE from three independent institutions were retrospectively included. Following a random split, 158 patients (70%) were assigned to a training cohort and the remaining 69 patients (30%) were assigned to a validation cohort. The process of variables selection was based on the importance variable scores generated by RF algorithm. A RF predictive model incorporating the selected variables was developed, and five-fold cross-validation was performed. The discrimination and calibration of the RF model were measured by a receiver operating characteristic (ROC) curve and the Hosmer-Lemeshow test.

Results: The potential variables selected by RF algorithm for developing predictive model of early TACE refractoriness included patients' age, number of tumors, tumor distribution, platelet count (PLT), and neutrophil-to-lymphocyte ratio (NLR). The results showed that the RF predictive model had good discrimination ability, with an area under curve (AUC) of 0.863 in the training cohort and 0.767 in the validation cohort, respectively. In Hosmer-Lemeshow test, the RF model had a satisfactory calibration with P values of 0.538 and 0.068 in training cohort and validation cohort, respectively.

Conclusion: The RF algorithm-based model has a good predictive performance in the prediction of early TACE refractoriness, which may easily be deployed in clinical routine and help to determine the optimal patient of care.

Experimental Models of Hepatocellular Carcinoma-A Preclinical Perspective

Hepatocellular carcinoma (HCC), the most frequent form of primary liver carcinoma, is a heterogenous and complex tumor type with increased incidence, poor prognosis, and high mortality. The actual therapeutic arsenal is narrow and poorly effective, rendering this disease a global health concern. Although considerable progress has been made in terms of understanding the pathogenesis, molecular mechanisms, genetics, and therapeutical approaches, several facets of human HCC remain undiscovered. A valuable and prompt approach to acquire further knowledge about the unrevealed aspects of HCC and novel therapeutic candidates is represented by the application of experimental models. Experimental models (in vivo and in vitro 2D and 3D models) are considered reliable tools to gather data for clinical usability. This review offers an overview of the currently available preclinical models frequently applied for the study of hepatocellular carcinoma in terms of initiation, development, and progression, as well as for the discovery of efficient treatments, highlighting the advantages and the limitations of each model. Furthermore, we also focus on the role played by computational studies (in silico models and artificial intelligence-based prediction models) as promising novel tools in liver cancer research.

US-Based Deep Learning Model for Differentiating Hepatocellular Carcinoma (HCC) From Other Malignancy in Cirrhotic Patients

The aim was to build a predictive model based on ultrasonography (US)-based deep learning model (US-DLM) and clinical features (Clin) for differentiating hepatocellular carcinoma (HCC) from other malignancy (OM) in cirrhotic patients. 112 patients with 120 HCCs and 60 patients with 61 OMs were included. They were randomly divided into training and test cohorts with a 4:1 ratio for developing and evaluating US-DLM model, respectively. Significant Clin predictors of OM in the training cohort were combined with US-DLM to build a nomogram predictive model (US-DLM+Clin). The diagnostic performance of US-DLM and US-DLM+Clin were compared with that of contrast enhanced magnetic resonance imaging (MRI) liver imaging and reporting system category M (MRI LR-M). US-DLM was the best independent predictor for evaluating OMs, followed by clinical information, including high cancer antigen 199 (CA199) level and female. The US-DLM achieved an AUC of 0.74 in the test cohort, which was comparable with that of MRI LR-M (AUC=0.84, p=0.232). The US-DLM+Clin for predicting OMs also had similar AUC value (0.81) compared with that of LR-M+Clin (0.83, p>0.05). US-DLM+Clin obtained a higher specificity, but a lower sensitivity, compared to that of LR-M +Clin (Specificity: 82.6% vs. 73.9%, p=0.007; Sensitivity: 78.6% vs. 92.9%, p=0.006) for evaluating OMs in the test set. The US-DLM+Clin model is valuable in differentiating HCC from OM in the setting of cirrhosis.