Predictive value of a combined clinical, serum biomarker, and radiomic model for poor short-term prognosis in acute pulmonary embolism.

This study evaluated the predictive value of combining clinical characteristics, serum biomarkers, and radiomic features for poor prognosis in patients with acute pulmonary embolism (APE). Clinical data, serum biomarker data (e.g., inflammatory and coagulation biomarkers), and computed tomography pulmonary angiography (CTPA) image from patients with APE were retrospectively collected from the First Affiliated Hospital of Hebei Northern University, First Hospital of Zhangjiakou, and Second Affiliated Hospital of Hebei Northern University. Patients were divided into good and poor prognosis groups. Data from the first two hospitals were randomly split into a training cohort (166) and internal validation cohort (72) using a 7:3 ratio; data from the third hospital formed the external validation cohort (37). Using 3D Slicer software, thrombus regions were outlined to extract radiomic features were constructed using R software. Model performance was assessed via receiver operating characteristic (ROC) curve, calibration, and decision curves analysis, and the Delong test. Six statistically significant radiomic features and independent clinical risk factors, including white blood cell count, neutrophile percentage, lymphocyte percentage, respiratory rate, pulse, syncope, smoking history, and RV / LV ratio, were identified (P < 0.05). The combined model outperformed individual models in the training (AUC = 0.85) (P < 0.05), internal validation (AUC = 0.86), and external validation cohorts (AUC = 0.88) (P < 0.05), demonstrating high clinical utility. The combined model effectively predicts early poor prognosis in APE, offering a robust tool for clinical evaluation and intervention planning.