Unenhanced CT quantitative vascular morphometry for detecting central pulmonary embolism: a multicenter study

Acute central pulmonary embolism (ACPE) is a life-threatening condition. Its early detection is critical for patient survival. ACPE presents with nonspecific symptoms such as dyspnea, chest pain, and syncope. Physicians frequently rely on unenhanced chest computed tomography (CT) as the initial imaging modality, where measurable vascular alterations may provide valuable diagnostic information for ACPE. This study aimed to develop a nomogram using unenhanced CT features for the early diagnosis of ACPE.

In total, 374 patients were enrolled in this multicenter study. All the patients underwent unenhanced CT and computed tomography pulmonary angiography (CTPA). Of the patients, 302 from Center A were randomly allocated to a training set (70%, n=212) and a validation set (30%, n=90), while 72 patients from two other centers were allocated to an independent test set for external validation. Vascular parameters, including the main pulmonary artery (mPA), inferior vena cava (IVC), azygos vein (AzV) arch, and superior vena cava (SVC) diameters, main pulmonary artery-to-aorta (mPA:AO) ratio, and longitudinal diameter of the main pulmonary artery (L-mPA) diameter from multiplanar reconstruction (MPR), were measured. Computed tomography attenuation at the origin of the pulmonary artery (CT-OPA) was also assessed. A nomogram was constructed using a stepwise logistic regression model.

The L-mPA and IVC diameters were independent risk factors for ACPE diagnosis. The nomogram achieved area under the curve (AUC) values of 0.821 [95% confidence interval (CI): 0.726-0.894] in the validation set and 0.827 (95% CI: 0.720-0.906) in the test set, with high specificity ranging from 89.2% (95% CI: 83.0-93.7%) to 96.0% (95% CI: 86.3-99.5%) but moderate sensitivity ranging from 54.6% (95% CI: 32.2-75.6%) to 55.6% (95% CI: 35.3-74.5%) across the two sets. The calibration curve demonstrated excellent goodness-of-fit, and the decision curve analysis (DCA) revealed a favorable clinical net benefit.

The L-mPA and IVC diameters on unenhanced CT were identified as independent risk factors for ACPE. The nomogram integrating these predictors performed well in confirming the diagnosis of ACPE.