Related Experiment Video
Updated: Jul 12, 2026
03:53
An Orthotopic Sciatic Nerve Xenograft for Neurofibromatosis Type 1 Neurofibromas
Published on: October 10, 2025
A Radiomics-Driven Model to Distinguish Between Clinically Similar Myxopapillary Ependymomas and Lumbosacral
Adhith Palla1, Nicolas K Goff1,2, Blake Perdikis1
1Department of Neurological Surgery, NYU Langone Health, New York, New York, USA.
Neurosurgery Practice
|July 11, 2026
Summary
A new machine learning model accurately distinguishes myxopapillary ependymomas (MPE) from lumbosacral schwannomas using radiomics. This tool aids in preoperative planning for spinal tumors, improving patient care.
Area of Science:
- Neurosurgery
- Radiology
- Machine Learning
Background:
- Myxopapillary ependymomas (MPE) and intradural lumbosacral schwannomas present similar imaging characteristics.
- Accurate differentiation is critical due to MPEs' risk of cerebrospinal fluid dissemination and need for tailored management.
Purpose of the Study:
- To develop a radiomics-based machine learning model for distinguishing MPEs from lumbosacral schwannomas on preoperative imaging.
- To assess the model's accuracy and robustness in differentiating these spinal tumors.
Main Methods:
- Retrospective analysis of 101 spinal tumor cases (32 MPEs, 61 intradural schwannomas, 8 dumbbell schwannomas).
- Manual tumor segmentation from T1 postcontrast MRI and radiomics feature extraction using PyRadiomics.
- Ensemble machine learning classifiers (random forest, k-nearest neighbors, naive Bayes) trained with nested cross-validation.
Main Results:
- The radiomics model achieved high discriminatory performance: area under the ROC curve of 0.895 (cross-validation) and 0.984 (test set).
- Identified features like maximum 3D diameter, significantly larger in MPEs, likely due to longitudinal growth.
- Model demonstrated robustness to vertebral column malalignment.
Conclusions:
- A radiomics-based machine learning model shows excellent ability to differentiate MPE from lumbosacral schwannoma.
- The model's accuracy and robustness suggest its potential as a valuable tool for preoperative spinal tumor planning and patient counseling.