1. Development of a novel prediction model for differential diagnosis between spinal myxopapillary ependymoma and schwannoma
- Author
-
Chorog Song, Hyun Su Kim, Ji Hyun Lee, Young Cheol Yoon, Sungjoon Lee, Sun-Ho Lee, and Eun-Sang Kim
- Subjects
Medicine ,Science - Abstract
Abstract Spinal myxopapillary ependymoma (MPE) and schwannoma represent clinically distinct intradural extramedullary tumors, albeit with shared and overlapping magnetic resonance imaging (MRI) characteristics. We aimed to identify significant MRI features that can differentiate between MPE and schwannoma and develop a novel prediction model using these features. In this study, 77 patients with MPE (n = 24) or schwannoma (n = 53) who underwent preoperative MRI and surgical removal between January 2012 and December 2022 were included. MRI features, including intratumoral T2 dark signals, subarachnoid hemorrhage (SAH), leptomeningeal seeding, and enhancement patterns, were analyzed. Logistic regression analysis was conducted to distinguish between MPE and schwannomas based on MRI parameters, and a prediction model was developed using significant MRI parameters. The model was validated internally using a stratified tenfold cross-validation. The area under the curve (AUC) was calculated based on the receiver operating characteristic curve analysis. MPEs had a significantly larger mean size (p = 0.0035), higher frequency of intratumoral T2 dark signals (p = 0.0021), associated SAH (p = 0.0377), and leptomeningeal seeding (p = 0.0377). Focal and diffuse heterogeneous enhancement patterns were significantly more common in MPEs (p = 0.0049 and 0.0038, respectively). Multivariable analyses showed that intratumoral T2 dark signal (p = 0.0439) and focal (p = 0.0029) and diffuse enhancement patterns (p = 0.0398) were independent factors. The prediction model showed an AUC of 0.9204 (95% CI 0.8532–0.9876) and the average AUC for internal validation was 0.9210 (95% CI 0.9160–0.9270). MRI provides useful data for differentiating spinal MPEs from schwannomas. The prediction model developed based on the MRI features demonstrated excellent discriminatory performance.
- Published
- 2024
- Full Text
- View/download PDF