Skip to main content

ORIGINAL RESEARCH article

Front. Oncol.
Sec. Cancer Imaging and Image-directed Interventions
Volume 14 - 2024 | doi: 10.3389/fonc.2024.1474665

An MRI-based radiomics approach to predict post-operative facial nerve dysfunction in small and medium-sized vestibular schwannomas

Provisionally accepted
Jun Yang Jun Yang 1,2Yan Liu Yan Liu 1Yangyang Wang Yangyang Wang 1Qin Xu Qin Xu 1Shucheng Jin Shucheng Jin 1Jianlong Zhouwen Jianlong Zhouwen 1Wang Jia Wang Jia 1*
  • 1 Beijing Tiantan Hospital, Capital Medical University, Beijing, China
  • 2 Beijing Luhe Hospital, Capital Medical University, Beijing, Beijing Municipality, China

The final, formatted version of the article will be published soon.

    Facial nerve dysfunction remains one of the postoperative complications following surgery for small and medium-sized vestibular schwannomas (VSs). With the advancement of magnetic resonance imaging (MRI) technology, radiomics has emerged as a novel approach that can extract imaging features for disease prediction. This study aims to establish a predictive model for facial nerve dysfunction using radiomics methods. The 256 patients who fulfilled the inclusion criteria were randomly assigned to training (n = 179) or test (n = 77) cohorts. Radiomics features were extracted. The intra-class correlation coefficients were used to select high-stability features within the training cohort and redundant features were screened. Finally, six radiomic models (logistic regression [LR], support vector machines [SVM], random forest [RF], decision tree [DT], light gradient boosting machine [LGB], and extreme gradient boosting [XGB]) named after the classifier name were established based on the selected features. The predictive performance of each model was evaluated in the test cohort. Ten radiomics features were selected from thetraining cohort to build six predictive models. In the test cohort, the area under the curve values of all models were >0.80, and were highest for the LR, LGB, and XGB models. The decision curve analysis of the test cohort showed that when the threshold probability was >10%, the LR, SVM, RF, DT, and XGB models showed better predictive performance and efficiency compared to theLGB model. The six prediction models based on machine learning of radiomic characteristics showed good prediction ability for post-operative facial nerve dysfunction of small-and medium-sized VSs. The LR and XGB models performed better and may help patients make preoperative decisions to achieve better facial nerve function.

    Keywords: Magnetic Resonance Imaging, Radiomics, Nerve dysfunction, vestibular schwannoma, machine learning

    Received: 02 Aug 2024; Accepted: 18 Dec 2024.

    Copyright: © 2024 Yang, Liu, Wang, Xu, Jin, Zhouwen and Jia. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

    * Correspondence: Wang Jia, Beijing Tiantan Hospital, Capital Medical University, Beijing, China

    Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.