Authors :
Presenting Author: Andrew Michalak, MD, MS – NYU Langone
Leyao Yu, BS – NYU Langone
Amirhossein Khalilian-Gourtani, PhD – NYU Langone
Alia Seedat, BS – NYU Langone
Cassandra Kazl, MD – NYU Langone
Orrin Devinsky, MD – NYU Langone
Werner Doyle, MD – NYU Langone
Patricia Dugan, MD – NYU Langone Health
Chris Morrison, PhD – NYU Langone
Daniel Friedman, MD – Department of Neurology, New York University Grossman School of Medicine, NYU Langone Health
Adeen Flinker, PhD – NYU Langone
Rationale:
Direct electrocortical stimulation (DES) is the gold standard for mapping language cortex in brain surgery1-3, but DES during an awake craniotomy is limited by spatial coverage and time constraints. Extraoperative DES overcomes these limitations, but few studies have systematically aligned its results with functional atlases. Further, language mapping in tumor cohorts demonstrate “canonical” regions with high probability of language function and low variance3, but less is known about whether these findings extend to epilepsy populations. We leveraged a large extraoperative DES cohort to determine probabilities of brain regions based on a functional atlas.
Methods:
We retrospectively analyzed all patients (March 2008–January 2023) at NYU Langone who underwent extraoperative DES and consented for research. Language testing included continuous speech, visual naming, and auditory naming. Stimulation began at 1–2 mA during continuous speech and increased until a reproducible language deficit (≥ 2/3 trials) was observed, or threshold current was reached. Absence of deficit defined a “cleared” site. If motor phenomena or seizures occurred, the trial was excluded. Afterdischarges disqualified positive trials but not cleared trials if threshold current was reached.
Analyses were based on Glasser HCP atlas ROIs. To estimate the probability and variability of language disruption per ROI, we bootstrapped 1,000 samples of each ROI with replacement and computed the mean hit probability and standard deviation. High mean/low SD ROIs were considered more canonical.
Results:
We included 128 implants (125 patients; 3 re‐implants; 62 female, mean age 30 years, 103 right‐handed, 2 ambidextrous). Implant locations: L-83; R-32; B/L-10. Forty‐six had lesions—most commonly focal cortical dysplasia (Table 1). Unsupervised clustering using the Glasser atlas revealed language disruptions throughout frontal, temporal, and parietal lobes. The ROIs with the highest hit probability were 6r (66.7%), 8C (59.7%), and PFop (58.4%). Speech arrest was most frequent in 6r (59.8%), 43 (55.2%), and PFop (51.7%). Anomia was most frequent in 47s (41.5%), 8Av (40.4%), and 8C (40.2%). Despite some ROIs showing a high mean probability, no clusters of consistently high‐mean/low‐variance ROIs emerged, indicating high interpatient variability (Fig 1).
Conclusions:
This analysis provides a practical functional region-based atlas derived from a large extraoperative DES cohort. Several ROIs demonstrating high mean probabilities for language subdomains, but in contrast to tumor surgery populations, there was high variance seen across all ROIs in this epilepsy cohort. This underscores the importance of within-patient mapping in epilepsy surgery populations.
1. Ojemann G, et al. Cortical language localization in left, dominant hemisphere. J Neurosurg. 1989;71(3):316-326
2. Sanai N, et al. Functional Outcome after Language Mapping for Glioma Resection. N Engl J Med. 2008;358(1):18-27
3. Chang EF, et al. Stereotactic probability and variability of speech arrest and anomia sites during stimulation mapping of the language dominant hemisphere. J Neurosurg. 2016;126(1):114-121
Funding: R01NS109367
R01DC018805
R01NS115929
CRCNS1912286