MRI-based texture analysis for differentiate between pediatric posterior fossa ependymoma type A and B

Abstract

Purpose

The aim of the study was to evaluate the feasibility of texture analysis in differentiating between posterior fossa ependymoma type A (PF-EPN-A) and type B (PF-EPN-B) among children.

Materials and Methods

Our retrospective study included 43 patients (37 PF-EPN-A and 6 PF-EPN-B) who were pathologically diagnosed with ependymomas in the posterior fossa. The texture features were extracted automatically from the volume of interests (VOIs), which were manually delineated on fluid-attenuated inversion recovery (FLAIR), contrast-enhanced T1-weighted (T1C), and diffusion-weighted imaging (DWI) MRI sequences. A receiver operating characteristic curve (ROC) was built to assess the diagnostic value of the texture parameters, and the prognostic value was evaluated by survival analysis.

Results

Texture parameter [Wavelet-LHH (H: High pass filter, L: Low pass. filter)_glcm (gray-level co-occurrence matrix)_Idn (Inverse difference normalized)] provides valuable information in distinguishing subgroups of ependymomas with higher specificity and positive predictive value (PPV). A total of 27 patients were divided into a high-risk group (IDN value＞0.916) and a low-risk group (IDN value＜0.916) with the most optimistic cut-off value (0.916). The Kaplan–Meier analysis of the survival curves showed significantly longer disease-free survival for low-risk groups compared to high-risk groups [hazard ratio (HR): 0.28, 95% confidence interval (CI): 0.11–0.69; p = 0.017].

Conclusion

Our results suggested that the texture parameters based on DWI images can be used to differentiate PF-EPN-A from PF-EPN-B. Texture analysis could be used as a noninvasive tool in distinguishing subgroup pediatric posterior fossa ependymomas and provide reliable prognostic information upon the verification of its reproducibility and feasibility by further studies.