Evaluation of 18 CT signs in diagnosing cecal volvulus: a multi-reader retrospective study

Purpose: To assess the diagnostic performance and reliability of 18 CT signs to diagnose cecal volvulus, a surgical emergency, versus a group of non-volvulus mimickers. Materials and methods: Four radiologists retrospectively and independently assessed 18 CT signs in 191 patients with cecal volvulus (n = 63) or a non-volvulus control group ((n = 128), including cecal bascule (n = 19), mobile cecum (n = 95), and colonic pseudo-obstruction (n = 14)) at a single institution from 2013 to 2021. Fleiss’ kappa coefficient was used to assess inter-reader agreement. For diagnostic performance metrics, we assessed sensitivity, specificity, and positive and negative predictive values. For predictive performance, all 18 signs were included in bivariate and stepwise lasso multivariate logistic regression models to diagnose cecal volvulus. Performance was assessed by ROC curves. Results: 191 patients (mean age: 63 years +/− 15.5 [SD]; 135 women) were included in the study. Nine of the 18 CT signs of cecal volvulus demonstrated good or better (> 0.6) inter-reader agreement. Individual CT signs with sensitivity, specificity, positive and negative predictive values all above 70% for diagnosing cecal volvulus were transition point, bird beak, and X-marks-the-spot. A lasso regression model determined four CT features: transition point, bird beak, coffee bean, and whirl had excellent prediction (AUC =.979) for cecal volvulus if all present. Conclusion: CT signs for cecal volvulus that have high sensitivity and specificity include: transition point, bird beak, and X-marks-the-spot and were reliable in distinguishing non-volvulus mimickers. If the following four features were present: transition point, bird beak, coffee bean, and whirl, there was excellent prediction (AUC =.979) for cecal volvulus.