Imaging-based prediction of durable benefit in conversion therapy for hepatocellular carcinoma: implications for surgical decision-making

In recent years, the integration of immune checkpoint inhibitors (ICIs) with targeted agents or locoregional therapies has substantially expanded the therapeutic landscape for patients with unresectable or potentially convertible hepatocellular carcinoma, enabling a proportion of patients to ultimately undergo curative-intent resection. Nevertheless, a persistent and unresolved clinical challenge remains: how to identify, at an early stage of treatment, those patients who are most likely to achieve durable clinical benefit, thereby avoiding ineffective therapy that may result in loss of time, deterioration of liver function, and missed opportunities for surgery. Within this context, the study by Lin et al[1] addresses a clinically relevant question by proposing an imaging-centered predictive model to estimate durable clinical benefit following ICI-based conversion therapy. According to the reported results, the integrated model demonstrated favorable discriminative performance in predicting sustained objective responses and outperformed models based solely on clinical variables. These findings suggest that quantitative information embedded within routine contrast-enhanced computed tomography may capture aspects of tumor biology and treatment sensitivity that are not readily apparent through conventional clinical assessment. From a surgical perspective, this observation is particularly noteworthy. Imaging plays a central role in outpatient evaluation and multidisciplinary discussions; however, its use has traditionally focused on qualitative assessments of tumor burden, vascular involvement, and technical resectability. By contrast, the present study attempts to translate imaging features into a quantitative tool that may provide an additional dimension for treatment pathway selection. From a methodological standpoint, one of the major strengths of this study lies in its choice of outcome measure. Rather than focusing on short-term radiologic response, the authors emphasize durable clinical benefit, a concept that more closely aligns with the fundamental objective of conversion therapy – namely, to achieve sustained disease control and thereby create a safe and feasible window for surgical intervention. Furthermore, imaging feature assessment was performed by two radiologists in a blinded fashion, and deep learning–derived features were integrated with clinical variables, which collectively enhances the robustness and clinical interpretability of the model. The incorporation of radiogenomic analyses to explore associations between model predictions and immune-related pathways further supports the biological plausibility of the imaging-based approach, a feature that remains relatively uncommon in comparable studies. Despite these strengths, several aspects of the study warrant cautious interpretation from the standpoint of hepatobiliary surgical decision-making. First, the retrospective, single-center design and the predominance of patients with specific etiologic and disease characteristics limit the generalizability of the findings to broader and more heterogeneous populations. Second, although an association was observed between model-derived scores and subsequent surgical resection or pathological response, a substantial translational gap remains between imaging-based prediction and the complex clinical decisions regarding whether and when to proceed with surgery. In routine practice, surgical candidacy is determined by a comprehensive assessment of liver functional reserve, portal hypertension, future liver remnant volume, treatment-related adverse events, and overall patient condition. As such, a single predictive model is unlikely to independently support these multifaceted decisions. In addition, the immune-related features identified through radiogenomic analyses remain correlative in nature, and their mechanistic relevance requires further experimental validation. Nonetheless, this study provides meaningful insights that may inform future research and clinical practice. From a hepatobiliary surgical perspective, the most promising direction lies not merely in further improving predictive performance, but in integrating such tools into prospective, multicenter conversion treatment workflows. Rigorous validation is needed to determine whether imaging-based prediction models can meaningfully optimize treatment sequencing, reduce ineffective therapy, and ultimately improve resection rates and long-term oncologic outcomes. If combined with standardized imaging protocols, objective assessment of resectability, and dynamic evaluation of liver function, imaging-driven predictive tools may eventually serve as a valuable adjunct within multidisciplinary decision-making frameworks. In summary, rather than offering definitive answers, this study proposes a thoughtful and clinically oriented approach to early assessment of treatment response in the setting of conversion therapy. From the perspective of surgeons, its value lies not in replacing clinical judgment, but in prompting a reconsideration of the role of imaging – not only as a means of response evaluation, but also as a potential window into tumor biology and therapeutic susceptibility. With the accumulation of higher-quality evidence, this line of investigation merits continued attention and careful exploration[2].