Spike Ripples: Hidden Clues to the Mystery of the Epileptogenic Zone

Abstract

Spike Ripples Localize The Epileptogenic Zone Best: An International Intracranial Study

Shi W, Shaw D, Walsh KG, Han X, Eden UT, Richardson RM, Gliske SV, Jacobs J, Brinkmann BH, Worrell GA, Stacey WC, Frauscher B, Thomas J, Kramer MA, Chu CJ. Brain. 2024;147(7):2496–2506. doi: 10.1093/brain/awae037.

We evaluated whether spike ripples, the combination of epileptiform spikes and ripples, provide a reliable and improved biomarker for the epileptogenic zone compared with other leading interictal biomarkers in a multicentre, international study. We first validated an automated spike ripple detector on intracranial electroencephalogram (EEG) recordings. We then applied this detector to subjects from 4 centres who subsequently underwent surgical resection with known 1-year outcomes. We evaluated the spike ripple rate in subjects cured after resection (International League Against Epilepsy Class 1 outcome, ILAE 1) and those with persistent seizures (ILAE 2-6) across sites and recording types. We also evaluated available interictal biomarkers: spike, spike-gamma, wideband high-frequency oscillation (HFO, 80-500 Hz). We also evaluated available interictal biomarkers: spike, ripple (80-250 Hz), and fast ripple (250-500 Hz) rates using previously validated automated detectors. The proportion of resected events was computed and compared across subject outcomes and biomarkers. Overall, 109 subjects were included. Most spike ripples were removed in subjects with ILAE 1 (International League Against Epilepsy Class 1) outcome (P < .001), and this was qualitatively observed across all sites and for depth and subdural electrodes (P < .001 and  < .001, respectively). Among ILAE 1 subjects, the mean spike ripple rate was higher in the resected volume (0.66/min) than in the nonremoved tissue (0.08/min, P < .001). A higher proportion of spike ripples were removed in subjects with ILAE 1 outcomes compared with ILAE 2-6 outcomes (P = .06). Among ILAE 1 subjects, the proportion of spike ripples removed was higher than the proportion of spikes (P < .001), spike-gamma (P < .001), wideband HFOs (P < .001), ripples (P = .009), and fast ripples (P = .009) removed. At the individual level, more subjects with ILAE 1 outcomes had the majority of spike ripples removed (79%, 38/48) than spikes (69%, P = .12), spike-gamma (69%, P = .12), wideband HFOs (63%, P = .03), ripples (45%, P = .01) or fast ripples (36%, P < .001) removed. Thus, in this large, multicentre cohort, when surgical resection was successful, the majority of spike ripples were removed. Furthermore, automatically detected spike ripples localize the epileptogenic tissue better than spikes, spike-gamma, wideband HFOs, ripples, and fast ripples.

Commentary

The epileptogenic zone (EZ) was defined by Lüders as “the area of cortex necessary and sufficient for initiating seizures and whose removal (or disconnection) is necessary for the complete abolition of seizures.” ¹ Prospectively, the EZ is estimated based on clinical judgment incorporating multimodal evidence about regions involved in ictal onset and propagation, interictal epileptiform activity, underlying lesion, and stimulation-evoked symptoms. The region involved in interictal epileptiform activity, the irritative zone, is usually larger than the EZ and relatively nonspecific among the factors available to inform this estimation. ²

The initial concept of the irritative zone was based on interictal spikes. Since then, transient bursts of oscillatory activity in the high gamma range have been suggested as more accurate intracranial biomarkers. Electroencephalogram (EEG) bursts above 80 Hz have been termed high-frequency oscillations (HFO), and those around 80–250 Hz “ripples.” High or broadband gamma oscillations in this range are also a widespread feature of physiologic activity. ³ In human hippocampus, sharp wave ripples in the 80–120 Hz range (higher in rodents) help mediate memory consolidation and reactivation. ⁴ In human neocortex, broadband gamma activity, which can extend as high as 250 Hz, is a key correlate of many physiological functions.

To increase specificity of HFOs as EZ biomarkers, additional criteria could be added. One suggestion is to increase the frequency threshold above the typical physiologic range. A “fast ripple,” higher than about 200 or 250 Hz, should exclude much physiological activity. Some patients exhibit oscillations even higher thresholds, exceeding 500 or even 1000 Hz. Modeling suggests phase shift synchronization of neuronal networks can yield oscillations above 1000 Hz. ⁵ As might be expected, however, increasing the frequency threshold increases specificity for EZ at the expense of sensitivity. For the higher ranges, no such oscillations might be recorded from a patient, not to mention the technical limitations. Another suggestion to increase specificity is to incorporate tasks that activate normal physiology to distinguish from pathology; this has for instance been proposed in hippocampus using an oddball task. ⁶

A recent suggestion has been to look for co-occurrence of oscillations with interictal epileptiform spikes. “Spike ripples” are defined as spikes with overriding oscillations in the 80–250 Hz range. Shi et al ⁷ have expanded on multiple prior reports that suggested this combined feature to be an accurate biomarker. Unlike some higher frequency oscillations, spike ripples were common enough that, on average, only a few minutes of sleep were sufficient to identify most channels containing them. With longer recordings, one could still envision other combinations involving higher frequency oscillations, which might be termed “spike fast ripples.” In one prior study, fast ripples overriding lower frequency oscillations were superior to ripples on spikes. ⁸

The usual method to assess performance of EZ biomarkers is retrospectively based on surgical outcome. Shi et al applied that standard paradigm by comparing relative rates of spike ripples from resected and nonresected locations between patients who became free and those that did not, in line with similar studies on other EEG-based biomarkers.

The authors addressed 3 hypotheses: (1) that the majority of spike ripples were removed in subjects who became seizure free, (2) that subjects who were seizure free had a higher proportion of spike ripples removed than in nonseizure-free patients, and (3) in patients who were seizure free, the proportion of spike ripples removed exceeded the proportion of other candidate biomarkers (spikes, spike-gamma, wideband HFOs, fast ripples, and ripples). These hypotheses were carefully tested in intracranial EEGs from 109 patients who underwent resection, taken from 4 centers, making this the largest evaluation of these biomarkers to date.

All 3 hypotheses were confirmed. Spike ripple removal correlated with seizure freedom. At the group level, though less clearly at the individual level, they were superior to spikes or the other HFO types evaluated. Accuracy of spike ripples exceeded fast ripples overall; spikes with fast ripples were not evaluated and might be too uncommon.

The findings lead to further questions. First, what are the circumstances in which spike ripples are most accurate and useful? It would be helpful to know how accuracy is affected by patient state, number and distribution of electrodes, ASM withdrawal, and temporal relationship to seizures. The total electrodes used were low in many cases, with an average of 59 contacts, and as few as 11 contacts, which may limit EZ discrimination.

Second, it would be useful more generally to evaluate spike ripples in terms of seizures. Recorded seizures, supplemented by stimulation, remain the main tool to determine what to resect. How well do spike ripples correlate with the electrode contacts involved in seizure onset, early versus late propagation, and other networked regions? Do spike ripples add any useful information about the EZ in comparison to what is deduced from the seizures themselves?

Third, the analysis evaluated each EEG channel independently, but there is sure to be important information based on their relative locations and combined activity. The spatial distribution of spike ripples may be valuable, including how tightly the spike ripple contacts cluster together. Regions with outgoing connectivity in the ripple and gamma bands correlate with resected contacts ⁹ ; a more accurate measure such as spike ripples could help elucidate such network activity.

Finally, like many observations in epilepsy, the physiological mechanisms underlying these oscillations are poorly understood. Some ways to identify these mechanisms could include further evaluation of single unit recordings in vivo and cellular imaging and electrophysiology on resected tissue.

Despite these questions, the validation provided by this study suggests that spike ripples are an accurate biomarker that should start entering standard clinical care. The next question will be how to make them practical. HFOs can be cumbersome to identify manually due to the magnified time scale required and easy to misidentify oscillations from filtering spikes or artifacts. ¹⁰ Therefore automated detections have become the standard in studies of HFOs regardless of subtype.

Another potentially useful contribution of the study is a new spike ripple detector. Two prior detectors from the same group were combined, one based upon time series features, and the other using spectrogram image classification (via a convolutional neural network). After tuning for high precision (positive predictive value), detector accuracy was good and sufficient to test the study aims, although one could imagine improvements with more modern methods. The analysis ran on a similar time scale as the recordings. Retrospective analysis of larger data segments (such as an entire EEG study) might currently be impractical clinically without specialized hardware, however only a few minutes of data may be required, and if desired, longer periods could be run during acquisition similar to other quantitative tools for spike identification and trending.