Postpartum ovarian vein thrombosis: plain CT with clinical indicators-diagnostic performance and clinical application.
Authors
Affiliations (4)
Affiliations (4)
- Department of Radiology, Guangdong Provincial Key Laboratory of Major Obstetric Diseases, Guangdong Provincial Clinical Research Center for Obstetrics and Gynecology, The Third Affiliated Hospital, Guangzhou Medical University, Guangzhou, China.
- Lecong Hospital of Shunde, 45 Lecong Avenue, Shunde District, Foshan, Foshan, Guangdong, China.
- Department of Radiology, Guangdong Provincial Key Laboratory of Major Obstetric Diseases, Guangdong Provincial Clinical Research Center for Obstetrics and Gynecology, The Third Affiliated Hospital, Guangzhou Medical University, Guangzhou, China. [email protected].
- Department of Radiology, Guangdong Provincial Key Laboratory of Major Obstetric Diseases, Guangdong Provincial Clinical Research Center for Obstetrics and Gynecology, The Third Affiliated Hospital, Guangzhou Medical University, Guangzhou, China. [email protected].
Abstract
Postpartum ovarian vein thrombosis (POVT) is a rare but serious condition, often presenting with nonspecific symptoms and potentially leading to life-threatening complications such as pulmonary embolism. The utility of plain computed tomography (CT) in POVT remains unexplored. We retrospectively analyzed 25 POVT patients and 27 postpartum controls with fever or abdominal pain. CT parameters, including plain and enhanced thrombus CT values and their ratios to the inferior vena cava (IVC), along with clinical markers, were measured and compared. Random forest models were constructed for discrimination and identifying concurrent thrombosis. Compared to controls, POVT patients showed significantly higher plain thrombus CT value (P-CT) and P-CT ratio, and lower enhanced thrombus CT value (E-CT) and E-CT ratio (all P < 0.001). ROC analysis demonstrated good discrimination performance for plain CT parameters (AUC: 0.773 for P-CT). For identifying concurrent thrombosis in other organs, D-dimer (AUC: 0.813) and P-CT (AUC: 0.807) were effective predictors. In repeated cross‑validation, the random forest models achieved mean AUCs of 0.94 (95% CI: 0.54-1.00) for classifying POVT and 0.82 (95% CI: 0.00-1.00) for identifying concurrent thrombosis. Plain CT can serve as a valuable initial screening tool for POVT. The integration of imaging and clinical parameters through machine learning models showed encouraging but preliminary discriminative ability (mean AUC 0.94 and 0.82, respectively), though with notable instability reflecting the modest sample size. These models may aid in early identification but require external validation.