This study aims to analyze the relative influence weights of key parameters—including the closest distance between the tumor and critical vessels, the ratio of initial tumor volume to liver volume, pathological tumor subtype, and alpha-fetoprotein (AFP) levels—on the selection of initial treatment strategies (surgical resection vs. neoadjuvant chemotherapy) for hepatoblastoma (HB). By systematically evaluating these core data, we seek to elucidate the rationale behind the choice of different treatment modalities, provide evidence-based guidance for clinical decision-making, and develop a quantitative clinical decision model.

Clinical data were collected from 55 pediatric patients diagnosed with hepatoblastoma and treated at the Affiliated Hospital of Qingdao University between 2015 and 2024. Using the Hisense Computer-Assisted Surgery (CAS) system, three-dimensional reconstructions of computed tomography (CT) images obtained prior to treatment intervention were performed to accurately measure tumor volume (in cm³), normal liver volume (in cm³), and the closest distance between the tumor and critical vessels (in mm). Additionally, clinicopathological subtypes (embryonal vs. fetal), serum AFP levels, and tumor lobar involvement (left lobe vs. right lobe) were documented. Univariate analysis, including Chi-square tests, Fisher's exact tests, independent samples t-tests, and Mann-Whitney U tests, was employed to evaluate the associations between these parameters and treatment selection outcomes. Furthermore, a multivariate binary logistic regression model was applied to assess the independent influence of these parameters on treatment strategy selection. Based on the statistically significant variables identified, an XGBoost machine learning model was constructed to develop a clinical decision-support tool for initial diagnosis and treatment planning.

All patients completed their treatment courses without major complications. Univariate analysis revealed that AFP levels, tumor-to-liver volume ratio, and the closest distance to critical vessels exhibited statistically significant differences (P < 0.05) in influencing the choice of initial treatment. Among these factors, the closest distance to critical vessels demonstrated the highest impact on treatment selection. In contrast, neither tumor lobar distribution nor histological subtype showed significant correlations (P > 0.05). Multivariate regression analysis further confirmed that AFP levels, the closest distance to critical vessels, and tumor-to-liver volume ratio were independent and significant factors affecting initial treatment selection (P < 0.05). Conversely, histological subtype and lobar involvement did not significantly influence treatment decisions (P > 0.05). This study validates that the choice of treatment strategy primarily depends on core indicators such as the closest distance to critical vessels, AFP levels, and tumor-to-liver volume ratio. The constructed XGBoost model effectively quantifies the selection of treatment modalities and provides an objective basis for personalized initial treatment planning.

This study demonstrates that AFP levels and the closest distance to critical vessels are pivotal factors influencing the selection of initial treatment strategies for children with hepatoblastoma. Additionally, the tumor-to-liver volume ratio is supported as an independent influencing factor and should be incorporated into treatment decision-making frameworks. The XGBoost model, developed based on these objective, readily accessible clinical and imaging parameters, exhibits strong predictive performance and calibration. By leveraging this model, clinicians can optimize individualized treatment approaches, thereby reducing the potential harm associated with suboptimal treatment choices, such as surgical risks or chemotherapy-related side effects, and ultimately improving clinical outcomes for pediatric patients.