Preoperative Brain Mapping Predicts Language Outcomes after Eloquent Tumor Resection
Authored by Matthew Muir, Kyle Noll, Sarah Prinsloo, Hayley Michener, Jeffrey I. Traylor, Vinodh A. Kumar, Chibawanye I. Ene, Sherise Ferguson, Jeffrey S. Weinberg, Frederick Lang, Brian A. Taylor, Stephanie J. Forkel, and Sujit S. Prabhu
- Posted
- Server
-
bioRxiv
Introduction
Glioma patients with tumors near critical language regions present significant clinical challenges. Surgeons often lack the tools to understand how each unique surgical approach may impact linguistic ability, leading to subjective decisions and unpredictable outcomes.
Objective
We aim to develop an approach that uses data-driven preoperative brain mapping to quantitatively predict the impact of planned resections on long-term language function.
Methods
This study included 79 consecutive patients undergoing resection of language-eloquent gliomas. Patients underwent preoperative navigated transcranial magnetic stimulation (nTMS) language mapping to identify language-positive sites (“TMS points”) and their associated white matter tracts (“TMS tracts”) as well as formal language evaluations pre and postoperatively. The resection of regions identified by preoperative mapping was correlated with persistent postoperative language deficits (PLDs).
Results
The resection of TMS points did not predict PLDs. However, a TMS point subgroup defined by white matter connectivity significantly predicted PLDs (OR=8.74, p<.01) and exhibited a canonical group-level anatomical distribution of cortical language sites. TMS-derived tracts recapitulated normative group-level patterns of white matter connectivity defined by the Human Connectome Project (HCP). Subcortical resection of TMS tracts predicted PLDs independently of cortical resection (OR=60, p<.001). The resected TMS tract segments in patients with PLDs co-localized with normative, language-associated subcortical pathways, in contrast to the resected TMS tract segments in non-aphasic patients (p<.05). Accordingly, resecting patient- specific co-localizations between TMS tracts and normative tracts in native space predicted PLDs with an accuracy of 94% (OR=134, p<.001). Co-localization between individualized and normative tracts precisely predicted the linguistic performance of a patient intraoperatively in response to direct electrophysiological stimulation of subcortical brain.
Conclusion
This study outlines a data-driven brain mapping approach that provides surgical insight by preoperatively predicting the impact of individual glioma resection on long-term language function.
Key Points
White matter connectivity determines the long-term functionality of cortical language sites mapped by TMS.
Long-term deficits in language processing result from resecting individualized subcortical regions within language-associated white matter tracts.
Non-invasive TMS language mapping combined with routine preoperative imaging can predict language outcomes of individual surgical approaches with an accuracy of 94%.
