ASO Author Reflections: Clinical-Radiomic Machine Learning Model Predicts Pheochromocytomas and Paragangliomas Surgical Difficulty: A Retrospective Study.

This study developed a machine learning (ML) model combining clinical and radiomic features to predict surgical difficulty in pheochromocytomas and paragangliomas (PPGLs), aiming to optimize preoperative planning and reduce perioperative complications. Retrospective clinical and imaging data from PPGLs patients were analyzed to construct two sets of models: clinical parameter models and clinical-radiomic models. Seven ML algorithms were tested, with the SVM-based clinical-radiomic model achieving the highest performance (training area under the curve [AUC]: 0.96, validation AUC: 0.85), significantly surpassing the clinical parameter model. SHAP analysis highlighted radiomic signature (Rad-score) as the strongest predictor, followed by body mass index, age, tumor size, and preoperative heart rate. The model enables objective stratification of surgical difficulty, aiding tailored preoperative strategies. Future directions include integrating multi-omics data, refining surgical difficulty criteria through multicenter studies, developing real-time intraoperative predictive tools, and automating radiomic workflows via deep learning. This research advances personalized surgical management for PPGLs, demonstrating significant clinical translation potential.