Correlation of preoperative and intraoperative motor cortex excitability in brain tumor patients using regressions and machine learning.
Publisher DOI
PubMed ID
41167036
Abstract
Objective
We compared the preoperative navigated transcranial magnetic stimulation (nTMS) resting motor threshold (RMT) with the intraoperative neurophysiological monitoring (IONM) motor threshold (MT) in relation to clinical and tumor characteristics.Methods
We retrospectively analyzed the data of 65 patients with motor eloquent supratentorial tumors, preoperative nTMS and resection with IONM. An nTMS and IONM threshold ratio was defined by dividing the motor evoked potential (MEP) (R)MT of the tumor-affected side by that of the healthy side. A random forest (machine learning (ML) algorithm) was implemented to distinguish the tumor side from the healthy side based on MEPs.Results
Tumor side nTMS-RMT and IONM-MT were significantly higher than that of the healthy side, most notably for tumors in the precentral gyrus. A positive linear correlation between nTMS-RMTratio and IONM-MTratio was observed. The random forest classifier achieved an accuracy of >90 % with the nTMS and IONM MEP data.Conclusion
We observed lower excitability in the tumor-affected hemisphere in pre- and intraoperative data, particularly with tumors in the precentral gyrus. Low preoperative excitability during nTMS correlated with lower intraoperative excitability.Significance
Preoperative nTMS mapping might influence the selection of intraoperative stimulation intensities to reduce the risk of false negative mapping.
We compared the preoperative navigated transcranial magnetic stimulation (nTMS) resting motor threshold (RMT) with the intraoperative neurophysiological monitoring (IONM) motor threshold (MT) in relation to clinical and tumor characteristics.Methods
We retrospectively analyzed the data of 65 patients with motor eloquent supratentorial tumors, preoperative nTMS and resection with IONM. An nTMS and IONM threshold ratio was defined by dividing the motor evoked potential (MEP) (R)MT of the tumor-affected side by that of the healthy side. A random forest (machine learning (ML) algorithm) was implemented to distinguish the tumor side from the healthy side based on MEPs.Results
Tumor side nTMS-RMT and IONM-MT were significantly higher than that of the healthy side, most notably for tumors in the precentral gyrus. A positive linear correlation between nTMS-RMTratio and IONM-MTratio was observed. The random forest classifier achieved an accuracy of >90 % with the nTMS and IONM MEP data.Conclusion
We observed lower excitability in the tumor-affected hemisphere in pre- and intraoperative data, particularly with tumors in the precentral gyrus. Low preoperative excitability during nTMS correlated with lower intraoperative excitability.Significance
Preoperative nTMS mapping might influence the selection of intraoperative stimulation intensities to reduce the risk of false negative mapping.
Date Issued
2025-10-21
Publication Type
Article
Subject(s)
Subjects
Brain mapping
•
Excitability
•
Intraoperative neurophysiological monitoring
•
Machine learning
•
Motor evoked potentials
•
Navigated transcranial magnetic stimulation
Language(s)
en
Additional Credits
Journal
Clinical Neurophysiology
Publisher
Elsevier
ISSN
1872-8952
1388-2457
Access(Rights)
open.access