The following is a summary of “Ensemble Machine Learning Model Incorporating Radiomics and Body Composition for Predicting Intraoperative HDI in PPGL,” published in the February 2024 issue of Endocrinology by Fu, et al.

During surgery for pheochromocytoma/paraganglioma (PPGL), intraoperative hemodynamic instability (HDI) poses risks of cardiovascular and cerebrovascular complications. For a study, researchers sought to assess the risk of intraoperative HDI in patients with PPGL to enhance surgical outcomes.

A retrospective analysis included 199 consecutive patients with confirmed PPGL. The cohort was divided into HDI (n = 101) and hemodynamic stability (HDS) groups (n = 98). Additionally, subcohorts were established for predictive modeling: a training cohort (n = 140) and a validation cohort (n = 59). Prediction models were developed using ensemble machine learning (EL) and multivariate logistic regression, incorporating computed tomography-based body composition parameters, tumor radiomics, and clinical data. Model performance was evaluated for discrimination, calibration, and clinical utility.

The EL model demonstrated excellent discrimination, with an area under the curve (AUC) of 96.2% (95% CI, 93.5%-99.0%) in the training cohort and 93.7% (95% CI, 88.0%-99.4%) in the validation cohort. Comparatively, the logistic regression model had lower AUC values: 74.4% (95% CI, 66.1%-82.6%) in the training cohort and 74.2% (95% CI, 61.1%-87.3%) in the validation cohort. The EL model exhibited favorable calibration and clinical applicability.

The EL model, integrating preoperative computed tomography-based body composition, tumor radiomics, and clinical data, offered promise in predicting intraoperative HDI in patients with PPGL.

Reference: academic.oup.com/jcem/article-abstract/109/2/351/7274065?redirectedFrom=fulltext