Radiomics and clinicoradiological factors as a promising approach for predicting microvascular invasion in hepatitis B-related hepatocellular carcinoma

Abstract

Microvascular invasion (MVI) is a critical factor in hepatocellular carcinoma (HCC) prognosis, particularly in hepatitis B virus (HBV)-related cases. This editorial examines a recent study by Xu et al who developed models to predict MVI and high-risk (M2) status in HBV-related HCC using contrast-enhanced computed tomography (CECT) radiomics and clinicoradiological factors. The study analyzed 270 patients, creating models that achieved an area under the curve values of 0.841 and 0.768 for MVI prediction, and 0.865 and 0.798 for M2 status prediction in training and validation datasets, respectively. These results are comparable to previous radiomics-based approaches, which reinforces the potential of this method in MVI prediction. The strengths of the study include its focus on HBV-related HCC and the use of widely accessible CECT imaging. However, limitations, such as retrospective design and manual segmentation, highlight areas for improvement. The editorial discusses the implications of the study including the need for standardized radiomics approaches and the potential impact on personalized treatment strategies. It also suggests future research directions, such as exploring mechanistic links between radiomics features and MVI, as well as integrating additional biomarkers or imaging modalities. Overall, this study contributes significantly to HCC management, paving the way for more accurate, personalized treatment approaches in the era of precision oncology.

Key Words: Hepatocellular carcinoma; Hepatitis-B; Microvascular invasion; Radiomics; Predicting factors

Core Tip: This editorial examines a recent study that predicts microvascular invasion (MVI) in hepatitis B-related hepatocellular carcinoma (HCC) using contrast-enhanced computed tomography (CT) radiomics and clinicoradiological factors. The study developed models that achieve high predictive accuracy for MVI and high-risk (M2) status. These findings align with previous radiomics-based approaches, reinforcing their potential in MVI prediction. The strengths of the study include its focus on hepatitis B virus-related HCC and the use of widely accessible CT imaging. However, limitations such as retrospective design highlight areas for improvement. This research contributes significantly to HCC management, paving the way for more accurate, personalized treatment approaches in precision oncology.

INTRODUCTION

Hepatocellular carcinoma (HCC), or liver cancer, remains a significant global health challenge, with hepatitis B virus (HBV) infection being a primary risk factor, especially in China[1-3]. HCC often affects individuals with preexisting liver problems. The main causes are long-term hepatitis B or C infections, excessive alcohol use, and non-alcoholic fatty liver disease. HCC symptoms include abdominal pain often localized to the upper right quadrant, weight loss and food aversion over a short period, yellowing of the skin and eyes due to bile accumulation, as well as persistent tiredness that does not improve with rest. The diagnosis of HCC usually involves imaging studies such as computed tomography (CT) or magnetic resonance imaging (MRI), blood tests for tumor markers such as α-fetoprotein (AFP), and sometimes liver biopsy for confirmation[4-6]. Despite advances in surgical techniques, the five-year recurrence rate following liver resection remains alarmingly high, reaching up to 70%[6,7]. Studies have shown that microvascular invasion (MVI) is a significant indicator of liver cancer recurrence after surgery. MVI occurs in 15% to 57.1% of patients with HCC[8,9]. It is vital to accurately forecast the presence of MVI prior to surgery for HCC to create individualized treatment plans and improve patient results.

In light of these circumstances, the manuscript by Xu et al[10] entitled "Evaluating MVI in hepatitis B virus-related HCC based on contrast-enhanced computed tomography radiomics and clinicoradiological factors" arouses great interest. This editorial aims to recapitulate the above-mentioned paper and highlight the importance of integrating symptomatology into clinical discussions about HCC and to emphasize the need for continued research into predictive models for MVI in HBV-related HCC.

STUDY OVERVIEW AND COMPARISON TO OTHER RESEARCH

The authors aimed to develop and validate models for predicting MVI and identifying high-risk (M2) status in patients with HBV-related HCC using a combination of contrast-enhanced computed tomography (CECT) radiomics and clinicoradiological factors. The study included 270 patients who underwent surgical resection, and the obtained information was divided into training and validation datasets. The methodology involved extracting radiomics features from tumor and peritumoral regions, selecting the significant features, and integrating them with clinicoradiological factors to construct predictive models.

The study's feature selection methodology raises important technical considerations. The use of an intra-class correlation coefficient followed by Pearson or Spearman correlation analysis, while statistically sound, may overlook significant non-linear relationships between radiomics features. The marginal change in performance between the integrated model (AUC 0.841) and standalone approaches (AUC 0.804 for clinicoradiological; 0.870 for radiomics) questions the value of complex feature integration strategies. For predicting M2 status, the model integrating AFP level, enhancing capsule, aspartate aminotransferase level, and radscore from the 5 mm peritumoral area in the arterial phase achieved AUCs of 0.865 and 0.798 in the training and validation datasets, respectively.

These results are promising and align with previous studies that have explored radiomics-based approaches for predicting MVI in HCC. For instance, Zhang et al[11] developed an MVI classifier integrating radiomics signature with clinical factors, achieving AUCs of 0.806, 0.803, and 0.796 in the training, test, and independent validation cohorts, respectively. Furthermore, Meng et al[12] achieved AUCs of 0.801 and 0.804 for their radiomics signatures, while Xu et al[10] reported AUCs of 0.841 and 0.768 for their combined radiologic-radiomics (RR) model. The selection of delayed-phase CECT images for the final model lacks detailed biological rationale, despite showing superior performance. Additionally, the choice of a 5 mm peritumoral region appears empirical rather than physiologically justified. The study's reliance on manual segmentation introduces potential reproducibility concerns that future automated approaches should address. By using different methodologies, simpler approaches might achieve similar results with greater clinical practicality. The performance of the models presented by Xu et al[10] is comparable to that achieved in the existing studies, which further validates the potential of radiomics-based approaches in predicting MVI.

STRENGTHS AND LIMITATIONS

The article's main strength is the combination of image analysis and clinical information, which allows for a more comprehensive understanding of micrometastasis prediction. Another strength is the focus on HBV-related HCC, which is particularly relevant in China where HBV is the predominant risk factor for HCC. The authors' decision to use CECT instead of MRI is pragmatic, considering the wider accessibility and lower costs of CT in clinical settings. However, the manual process of identifying specific areas of interest in the images could introduce bias, which emphasizes the importance of developing more automated and consistent methods for analyzing medical images.

It is noteworthy that the study found no significant difference in MVI predictive ability among the RR, radiomics-only, and clinicoradiological models, implying there is no significant difference in prediction capacity between RR models, radiomics-based models, and clinicoradiological models. This observation raises questions about the added value of radiomics features over traditional clinical and radiological factors in this context. Future research should aim to clarify the specific populations that could benefit most from radiomics-based models and explore ways to enhance their predictive power.

A common limitation across recent studies, including works by Bai et al[13] and the current study by Xu et al[10], is their retrospective, single-center design. While individual studies show promising results, the field collectively faces challenges in establishing multicenter validation. Some studies, including those by Liu et al[14] and Zhong et al[15], emphasize the need for prospective, multicenter trials to validate these findings.

The ethical framework of this study, while appropriately approved by the Institutional Ethics Committee (approval number: 2021-RE-043), could be further strengthened. Future studies should explicitly address data privacy protection measures and compliance with international healthcare data regulations. This becomes particularly important as predictive models move towards multicenter validation and implementation.

CLINICAL SIGNIFICANCE

Regarding clinical implementation, several practical considerations need to be addressed before widespread adoption. The current requirement for manual tumor segmentation introduces time constraints and potential observer variability, highlighting the need for automated delineation algorithms. Implementation of these predictive models in clinical practice would require the development of automated tools, standardization of imaging protocols across centers, and seamless integration with existing clinical workflows. The clinical significance of these AUC differences requires careful consideration. The modest improvement in predictive accuracy (approximately 4 additional correct predictions per 100 patients) must be weighed against the increased complexity and resource requirements of radiomics analysis. Additionally, the performance drop in the validation dataset (AUC 0.768) compared to the training set (0.841) suggests potential overfitting issues that could affect clinical reliability. Additionally, cost-effectiveness analyses comparing conventional approaches with radiomics-based methods would be beneficial for healthcare systems considering the adoption of these techniques.

FUTURE DIRECTIONS AND PERSPECTIVES

Looking prospectively, several avenues for future research emerge from this study. The first one, focused on exploring the mechanistic links between radiomics features and MVI, could provide valuable insights into tumor biology and potentially reveal new therapeutic targets[16]. Moreover, investigating the impact of these predictive models on clinical decision-making and patient outcomes would be an adequate consecutive step[17,18]. The question is how surgical strategies or adjuvant therapies could be modified based on these predictions, and what effect this would have on recurrence rates and overall survival. Furthermore, the integration of other emerging biomarkers or imaging modalities with these radiomics-based models could potentially enhance their predictive accuracy. For instance, liquid biopsy techniques or advanced functional imaging might provide complementary information to further refine MVI prediction[19].

While delayed-phase CECT demonstrated superior performance, the physiological basis for this advantage requires further exploration. Similarly, the M2 prediction model, integrating AFP, enhancing capsule, aspartate aminotransferase, and arterial phase radiomics features shows promise, but a stronger biological rationale for this combination is needed. Future studies should address the standardization of feature extraction across different imaging protocols and equipment to enhance reproducibility and clinical applicability.

To validate these findings and assess real-world impact, prospective, multicenter studies are crucial across diverse populations. Future research should prioritize automated analysis, biomarker integration, and standardized clinical implementation protocols. While CECT is effective, MRI and positron emission tomography (PET)-CT warrant further investigation due to their potential for additional information. However, CECT remains accessible and cost-effective. As demonstrated by Meng et al[20], CT and MRI show comparable MVI prediction in solitary HCC. Further studies, especially in HBV-related HCC, are needed to optimize imaging protocols and establish standardized guidelines.

Prospective research should look into the molecular relationships between radiomic characteristics and tumor biology, which may lead to novel therapeutic targets. Furthermore, more research into how these predictive models affect clinical decision-making and patient outcomes is required. For example, how can surgical methods or adjuvant medicines be altered in response to these projections, and what effect would this have on recurrence rates and overall survival? Key concerns that require more investigation include the precise clinical value of radiomics and its potential benefits across different patient populations.

CONCLUSION

Radiomics models offer a promising approach to personalize HCC treatment. The research conducted by Xu et al[10] is a valuable addition to the field of liver cancer treatment as it offers a new method for predicting the presence of MVI and high-risk (M2) status in HBV-related liver cancer before surgery. Although more research is needed to confirm and improve this method, it does provide a strong anchorage point for developing tailored treatment plans for liver cancer patients. Successful clinical implementation necessitates standardizing methods, conducting robust clinical trials, and addressing practical limitations such as the need to automate manual tumor segmentation processes, expand beyond single-center retrospective designs, mitigate model overfitting indicated by validation performance drops, establish a biological rationale for imaging parameter selection, and validate chosen peritumoral regions based on physiological evidence rather than empirical choices. Continued advancements in imaging and artificial intelligence will likely lead to further refinement and integration of these models into routine HCC care. Radiomics-based prediction models may become essential tools for making informed clinical decisions and improving patient outcomes.