Differentiation of epileptic regions from voluntary high-gamma activation via interictal cross-frequency windowed power-power correlation

Restricted access

OBJECTIVE

Electrocorticography is an indispensable tool in identifying the epileptogenic zone in the presurgical evaluation of many epilepsy patients. Traditional electrocorticographic features (spikes, ictal onset changes, and recently high-frequency oscillations [HFOs]) rely on the presence of transient features that occur within or near epileptogenic cortex. Here the authors report on a novel corticography feature of epileptogenic cortex—covariation of high-gamma and beta frequency band power profiles. Band-limited power was measured from each recording site based on native physiological signal differences without relying on clinical ictal or interictal epileptogenic features. In this preliminary analysis, frequency windowed power correlation appears to be a specific marker of the epileptogenic zone. The authors’ overall aim was to validate this observation with the location of the eventual resection and outcome.

METHODS

The authors conducted a retrospective analysis of 13 adult patients who had undergone electrocorticography for surgical planning at their center. They quantified the correlation of high-gamma (70–200 Hz) and beta (12–18 Hz) band frequency power per electrode site during a cognitive task. They used a sliding window method to correlate the power of smoothed, Hilbert-transformed high-gamma and beta bands. They then compared positive and negative correlations between power in the high-gamma and beta bands in the setting of a hand versus a tongue motor task as well as within the resting state. Significant positive correlations were compared to surgically resected areas and outcomes based on reviewed records.

RESULTS

Positive high-gamma and beta correlations appeared to predict the area of eventual resection and, preliminarily, surgical outcome independent of spike detection. In general, patients with the best outcomes had well-localized positive correlations (high-gamma and beta activities) to areas of eventual resection, while those with poorer outcomes displayed more diffuse patterns.

CONCLUSIONS

Data in this study suggest that positive high-gamma and beta correlations independent of any behavioral metric may have clinical applicability in surgical decision-making. Further studies are needed to evaluate the clinical potential of this methodology. Additional work is also needed to relate these results to other methods, such as HFO detection or connectivity with other cortical areas.

ABBREVIATIONS CPS = complex partial seizure; ECoG = electrocorticography; EEG = electroencephalography; GTCS = generalized tonic-clonic seizure; HFO = high-frequency oscillation; HG = high gamma; SAH = selective amygdalohippocampectomy; SPS = simple partial seizure.

Downloadable materials

  • Supplemental Fig. S1 (PDF 574 KB)

Article Information

Correspondence Michael Kogan: University at Buffalo Neurosurgery, Buffalo, NY. mkogan@ubns.com.

INCLUDE WHEN CITING Published online May 10, 2019; DOI: 10.3171/2019.2.JNS181991.

Disclosures The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

© AANS, except where prohibited by US copyright law.

Headings

Figures

  • View in gallery

    Summary of methods. A and B: Log HG and beta power in one epoch from a hand task in one channel. A 250-msec average sliding window was used to smooth the signal. C: Smoothed HG and beta signals. A 500-msec sliding window correlation was used to calculate a rho correlational coefficient over the course of an epoch. D: Resulting rho value over one epoch. E: Mean correlation over trials with a significant rho from all epochs at every individual correlation window; one paired t-test was used. Instances of significant correlation are superimposed on the mean signal from all epochs (red, positive; blue, negative). F: Mean signal over trials with significant correlations after p values were adjusted using a false discovery rate correction. G: Spectrogram of all epochs in the channel. H: Adjusted p values were summed for each channel and then converted to z-scores for the entire grid (red, positive; blue, negative). I: Plot of only positive z-scores over a value of 1. Figure is available in color online only.

  • View in gallery

    Tongue (left) and hand (right) motor task maps for significant positive (+ Rho) and negative (− Rho) correlations in four patients (red, positive; blue, negative). Clinical stimulation of motor and/or sensory electrodes circled yellow over corresponding channels. Figure is available in color online only.

  • View in gallery

    Maps of significant positive correlation electrodes. A: Five patients underwent resections of presumed focus at the area of the circle. B: Three patients underwent temporal lobectomies. C: Three patients had selective temporal resections or laser ablation. D: Two patients were diagnosed with multifocal epilepsy and did not undergo resection. For patients with available follow-up, Engel scores are shown. LITT = laser interstitial thermal therapy; NA = not available; Pt = patient. Figure is available in color online only.

  • View in gallery

    Maps of spike distribution based on the z-score threshold method. A: Five patients underwent resections of presumed focus at the area of the circle. B: Three patients underwent temporal lobectomies. C: Three patients had selective temporal resections or laser ablation. D: Two patients were diagnosed with multifocal epilepsy and did not undergo resection. For patients with available follow-up, Engel scores are shown. Figure is available in color online only.

  • View in gallery

    A: Case 13. Nonfocal epilepsy. B: Case 7. Left temporal lobectomy. C: Case 9. Left SAH with clinical improvement. D: Case 11. Left SAH with no clinical improvement. Clinical seizure onset zone (upper) and positive correlation map (lower). E: Example spike detection; there is broadband signal in the densest spike channel. Detected spikes are superimposed (red asterisks). F: Broadband signal power for the most significant positive channel with spikes superimposed. G: Postresection MRI. Figure is available in color online only.

TrendMD

Metrics

Metrics

All Time Past Year Past 30 Days
Abstract Views 143 143 62
Full Text Views 30 30 13
PDF Downloads 16 16 7
EPUB Downloads 0 0 0

PubMed

Google Scholar