From prediction to clinical decision-making: Explainable machine learning in acute suppurative cholecystitis

Abstract

Acute suppurative cholecystitis is a severe and potentially life-threatening form of gallbladder inflammation, in which delayed diagnosis and intervention are associated with increased morbidity and perioperative risk. Early differentiation from uncomplicated disease remains challenging, as clinical presentation and conventional imaging often fail to accurately reflect disease severity at initial assessment. Recent advances in machine learning, particularly the integration of clinical data with computed tomography-derived radiomic features, have improved preoperative risk stratification by capturing complex, high-dimensional patterns beyond traditional diagnostic frameworks. However, clinical adoption has been limited by concerns regarding interpretability and trust. Explainable machine learning addresses this limitation by providing transparent insights into model predictions. Techniques such as SHapley Additive exPlanations enable quantification of feature contributions at both population and individual levels, supporting clinical understanding and facilitating integration into decision-making. In this Opinion Review, we argue that the value of explainable machine learning lies not only in prediction, but in structuring clinical decision-making. Interpretable models may enhance early risk recognition, guide surgical timing, and improve multidisciplinary coordination. Future research should focus on prospective validation, workflow integration, and evaluation of real-world clinical impact.

Keywords: Acute suppurative cholecystitis; Explainable machine learning; Clinical decision-making; Radiomics; Surgical prioritization; Perioperative care

Core Tip: Early and accurate identification of acute suppurative cholecystitis is essential for timely surgical intervention and improved outcomes. Explainable machine learning models that integrate clinical and radiomic data provide transparent risk stratification, supporting surgical decision-making and perioperative prioritization. When embedded into multidisciplinary care pathways, such tools may enhance patient safety and real-world management of acute biliary disease.