An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Noebels JL, Avoli M, Rogawski MA, et al., editors. Jasper's Basic Mechanisms of the Epilepsies. 5th edition. New York: Oxford University Press; 2024. doi: 10.1093/med/9780197549469.003.0036
Insults to the brain (e.g., stroke, head trauma, infection, status epilepticus) can precipitate the development of epilepsy—months or even years after the initial insult. While not all patients with such injuries will go on to develop epilepsy, in a subgroup of individuals the injury will trigger epileptogenesis—a complex chain of events that transforms a nonepileptic brain into one that is prone to generating spontaneous seizures. To date, there are no clinically available tools capable of predicting which individuals are at high risk of post-insult epilepsy. A growing number of studies attempt to bridge this gap, by evaluating the utility of different diagnostic modalities (e.g., neuroimaging, molecular testing, electroencephalogram [EEG] analysis) in capturing changes reflective of the epileptogenic process. EEG holds particular promise as a modality for reflecting epileptogenic changes, due to its high temporal resolution and subsequent ability to capture a wide variety of brain-activity signatures. With the added advantages of portability, relative ease of use, safety, and low cost, a successful EEG-based biomarker is likely to become widely accessible in clinical settings. This chapter will discuss the most promising EEG signatures of epileptogenesis (interictal spikes, high-frequency oscillations, theta dynamics, and nonlinear dynamics), and the challenges of translating these biomarkers into clinically applicable tools. We conclude by highlighting the potential value of biomarker paradigms that (1) combine several signatures and (2) examine dynamic changes in these signatures over time.
The ability to study the electrical activity of the human brain was first reported in 1929 by German psychiatrist Hans Berger (Berger, 1929). Using electrodes placed on the surface of the head, Berger found correlates between electrical activity and functional states such as attention, sleep, and coma (Berger, 1929). Berger’s technology became the first approach successful in reflecting the function of the brain, and he was the one to give it its now well-known name—the electroencephalogram.
The utility of the electroencephalogram (EEG) in the understanding of epilepsy became apparent with a series of reports between 1934 and 1936, demonstrating epileptiform spikes (Fischer and Löwenbach, 1934), the pattern of absence seizures (Gibbs, Davis, and Lennox, 1935), and focal interictal spikes (Gibbs, Lennox, and Gibbs, 1936; Jasper, 1936). Despite the emergence of modalities that utilize alternative (nonelectrical) indicators of brain activity (e.g., vascular dynamics imaged using functional magnetic resonance imaging [MRI], or metabolic function imaged using positron emission tomography [PET] and single-photon emission computed tomography [SPECT]), the EEG continues to be the first-line method for diagnosing epilepsy, characterizing seizures, and identifying the seizure-onset zone in candidates for resective surgery. The widespread use of EEG in epilepsy is primarily attributed to its superior temporal resolution (millisecond scale) and its ability to capture the dynamics of neuronal networks in real time. Additional advantages of EEG over other imaging modalities are its portability, relative ease of use, safety, and low cost.
The main limitation of EEG is its poor spatial resolution. Each electrode captures a summation of electrical activity generated in tissue spanning several square centimeters and containing ~30–500 million cortical neurons (pyramidal neurons in superficial layers of the cortex; Buchtel, 2002). However, with recent technological advances that combine high-density EEG systems, head anatomy mapping, and advanced postprocessing algorithms—the spatial resolution of EEG stands to improve greatly in the near future (Michel and Brunet, 2019).
While EEG has a well-established utility in capturing hallmarks of an epileptic brain (e.g., seizure activity and interictal spikes), its usefulness in detecting epileptogenesis—the dynamic transformation of a healthy neural network into one that is prone to generating spontaneous seizures—remains the subject of ongoing exploration. Accumulating evidence indicates that the process of epileptogenesis involves changes in interneuronal connectivity and plasticity that occur over a period of months and years after an initial brain insult (e.g., stroke, head trauma, infection, status epilepticus; Pitkänen et al., 2007). This raises the hypothesis that these changes and the associated alterations in network activity may also have EEG signatures that can be detected prior to the onset of spontaneous seizures.
To demonstrate the validity of this hypothesis, studies must show that a specific signature allows reliable differentiation between (1) patients who will recover from the insult and (2) those who will eventually develop epilepsy. Appropriate studies in humans require years-long follow-up of patients in study designs that are often costly to conduct. Hence, as we will see throughout this chapter, most of the existing biomarker candidates were identified in prospective animal studies, and their performance was evaluated by comparing biomarker presence in animals that develop post-insult epilepsy versus those that do not. However, while animal studies can offer valuable insights into epileptogenic processes, they have several major drawbacks (Table 36–1). These include (1) their limited relevance to epilepsy in humans (Löscher, 2017); (2) the lack of consensus regarding the definition of epilepsy in animals (Simonato et al., 2017); and (3) the lack of standardized procedures for the acquisition and analysis of animal EEG (Moyer et al., 2017).
Advantages and Disadvantages of Human and Animal Studies of EEG Biomarkers for Epileptogenesis: Video-EEG, Simultaneous Monitoring of Brain Activity (Using EEG) and Behavioral Changes (Using a Video Camera).
The term “EEG signatures” refers to patterns within the signal that are characteristic to specific brain activity. The inventor of the EEG—Hans Berger—is also considered to have established the first “EEG signature,” when noticing that a periodicity of ~10 cycles per second (Hz) occurs during eyes-closed wakeful relaxation (Berger, 1929). He was the one to give this periodicity its name—alpha waves. Since then other patterns have been discovered, establishing EEG signatures of different physiological and pathological states. In 2021, the state-of-the-art EEG signatures of epileptogenesis include the following (see Table 36–2).
Electroencephalographic Signatures of Epileptogenesis.
The occurrence of interictal spikes—brief events (lasting <250 milliseconds) of epileptiform spikes and sharp waves between seizures—have been considered to be an important EEG signature of established epilepsy since the 1970s (Gibbs, Lennox, and Gibbs, 1936; Jasper, 1936; Gloor, 1975; Staley and Dudek, 2006). However, their utility in predicting epilepsy is a matter of ongoing debate, as traditionally spikes are considered to appear after at least one spontaneous seizure (Staley and Dudek, 2006). Contrary to this early belief, in 2017 our group demonstrated that spikes do precede the development of spontaneous seizures in five murine models of acquired epilepsy (Milikovsky et al., 2017). However, having also found spikes in animals that did not develop post-insult epilepsy, we concluded that this EEG signature has poor specificity to an epileptic outcome in rodents. A 2018 study in mice reported that the specificity of spikes as predictors of epilepsy can be improved—by focusing on spikes that are coupled with cardiac-rhythm abnormalities (Bahari et al., 2018). The same year, a study of stroke patients demonstrated that patients with interictal spikes within 72 hours post stroke, are at a 3.8 times higher risk of developing epilepsy within a year’s time than patients without this EEG signature (Bentes et al., 2018). Additional large-cohort studies are needed to better understand the epileptogenic role of interictal spikes and validate their utility as biomarkers for epileptogenesis.
In 1999, another type of interictal activity was observed. Using intracranial EEG electrodes, Bragin et al. found high-frequency oscillations (HFOs; defined as at least four oscillations at a frequency between 100 and 500 Hz) originating from epileptic tissue in animals with experimental epilepsy and in patients with mesial temporal lobe epilepsy (Bragin et al., 1999a, 1999b). This discovery established a new avenue in epilepsy research, investigating (1) the mechanisms of HFO generation (Jiruska et al., 2017); (2) the co-occurrence of HFOs and interictal spikes (Jacobs, Kobayashi, and Gotman, 2011; Salami et al., 2014); (3) the utility of HFOs in identifying the seizure-onset zone (Jacobs et al., 2008; Cimbalnik, Kucewicz, and Worrell, 2016; Sagi and Evans, 2016); and (4) the potential of HFOs to serve as a signature of epileptogenesis (Bragin et al., 2004, 2016).
Bragin and colleagues have demonstrated the utility of HFOs as a signature of epileptogenesis in two models of insult-induced epilepsy: (1) epilepsy induced by status epilepticus (Bragin et al., 2004); and (2) epilepsy induced by traumatic brain injury (TBI; Bragin et al., 2016). Both studies showed that all animals with HFOs in the first 2 weeks following the insult went on to develop recurrent spontaneous seizures, while animals with no HFOs remained seizure-free (Bragin et al., 2004, 2016).
However, there remains a paramount challenge in using this signature as a biomarker of epileptogenesis: HFOs are also involved in normal brain activity and are routinely observed in control animals (Engel et al., 2009; Bruder et al., 2021). For this biomarker to prove successful, there is therefore a critical need to first distinguish between “physiological” and “pathological” HFOs.
The clinical translation of this biomarker is also limited by the difficulty to reliably detect HFO in noninvasive recordings. While studies demonstrating the ability to detect HFOs noninvasively emerged over a decade ago (Xiang et al., 2009; Kobayashi et al., 2010; Andrade-Valenca et al., 2011; Ferrari-Marinho et al., 2020), these approaches are susceptible to frequent false negatives (true HFOs that fail to be detected due to signal loss) and false positives (muscle activity and background noise that are mistaken for HFOs; Zijlmans et al., 2012, 2017). The establishment of HFOs as a robust noninvasive biomarker for epileptogenesis, therefore, awaits the development of techniques for accurate distinction between (1) HFOs originating from brain and non-brain sources and (2) “physiological” and “pathological” brain HFOs.
Shortly after Hans Berger gave the name “alpha-waves” to EEG activity of ~10 Hz in 1929, other Greek letters were assigned to EEG oscillations at specific frequency ranges. Today, the nomenclature typically associates the letter alpha with oscillations at 8–12 Hz, beta with oscillations at 15–30 Hz, gamma with oscillations above 30 Hz, delta with 0.5–3 Hz, and theta with 3–8 Hz. As each of these waves becomes more or less dominant with specific brain-states (Akila et al., 2020), in 2017 our group sought to examine their relevance to the epileptogenic process (Milikovsky et al., 2017). We used long-term EEG recordings from five animal models of insult-induced epilepsy to evaluate wave-based signatures in three experimental paradigms: (1) the intensity of specific waves at a single time-point 4 days post insult (1-hour EEG recording); (2) the change in wave intensity from 1 to 4 days post injury (1-hour EEG recording every 24 hours for 4 days); and (3) the change in wave intensity over a continuous 3-day recording. The third paradigm was found to be the most successful, revealing that a change in theta-wave intensity predicts epilepsy with an accuracy of >90%. The second paradigm yielded an accuracy of >65%, while the first failed to perform better than chance.
Together, these findings suggest that a single EEG exam may be insufficient for reliable prediction of epilepsy (primarily due to large interindividual variability), and they highlight the potential value of monitoring within-subject changes in cortical activity over time. Future studies are awaited to demonstrate the utility of such paradigms in humans.
Nonlinear analysis aims to characterize and predict complex dynamical features in time-series data. The use of nonlinear analysis rose to prominence in the late 1980s with the introduction of techniques for visualizing (Eckmann, Oliffson Kamphorst, and Ruelle, 1987) and quantifying (Webber and Zbilut, 1994) recurrences (time correlations) in noisy and nonstationary data. The nonstationary nature of EEG signals (having means, variances, etc. that change over time) has led several groups to explore whether nonlinear EEG patterns can be used to detect and predict seizures (Lehnertz and Elger, 1998; Ouyang et al., 2008; Acharya et al., 2011; Ngamga et al., 2016).
The utility of nonlinear EEG analysis in the prediction of epilepsy was explored in two animal studies by Rizzi et al., showing that a gradual and long-lasting increase in nonlinear dynamics is predictive of epilepsy (Rizzi et al., 2016, 2019). The authors demonstrated the phenomenon in three animal models of insult-induced epilepsy and proposed two measures of increased nonlinearity: (1) increased laminarity (appearance of slow-changing dynamics; Rizzi et al., 2016) and (2) reduced EEG dimensionality (Rizzi et al., 2019).
Notably, similar to the theta-dynamics signature proposed by Milikovsky et al. (2017), the signatures proposed by Rizzi et al. (2016, 2019) capture changes in days-long EEG recordings. The ability to translate these signatures into clinically applicable EEG paradigms requires future validation.
Accumulating insights into acquired epilepsy highlight the variability in both epilepsy phenotypes and epileptogenic trajectories—which can vary in time-scale, spatial distribution, molecular and cellular alterations, and overall network modifications. All these factors make it challenging to rely on a single biomarker in a one-size-fits-all approach. To overcome this challenge, future studies ought to (1) focus on per-patient changes in cortical activity by monitoring patients repetitively/continuously post insult (Fig. 36–1); (2) identify a multifeature biomarker as a combination of best-performing EEG signatures, imaging hallmarks, and molecular markers of epileptogenesis, toward a comprehensive characterization of changes associated with different epileptogenic trajectories; and (3) determine the optimal time intervals between repeated patient evaluations post-insult.
Proposed paradigm for monitoring patient-specific changes in EEG properties. Future studies should examine the predictive power of biomarkers that combine several EEG signatures and examine changes in these signatures over time. Compared to traditional (more...)
Animal models of epilepsy can be useful for identifying epileptogenic mechanisms and mechanism-specific EEG signatures. The chief drawback of this approach is the limited ability of animal models to mimic the conditions of acquired epilepsy in humans (e.g., insult characteristics, acute post-insult treatment, time course of epileptogenesis, neocortical complexity, existing comorbidities, medications, and risk genes for epilepsy). To overcome this limitation, there is a critical need for comprehensive mining of human EEG data—aimed at (1) optimizing the de-noising of scalp EEG; (2) demonstrating the clinical utility of EEG signatures identified in animal studies; and (3) identifying alternative signatures, specific to insult, epilepsy phenotype, age, gender, comorbidities, medications, and so on. This endeavor will require large multicenter studies with robust prospective/retrospective study designs and a rigorous evaluation of predictive power.
A reliable biomarker (or biomarker combination) stands to greatly boost the development of new antiepileptogenic drugs. By allowing the identification of animals/patients who are at high risk of developing post-injury epilepsy, such a tool will drastically reduce the sample size required for sufficiently powered drug studies and the cost of evaluating new treatments.
A robust and quantitative biomarker for epileptogenesis also has the potential to serve as a standardized measure of treatment efficacy, facilitating data comparison between research groups and improving the reproducibility of research findings (Harte-Hargrove et al., 2017; Moyer et al., 2017). Moreover, it may shorten the duration of drug trials by allowing biomarker-based evaluation of response to treatment within weeks/months of treatment initiation (obviating the need to wait for the appearance of spontaneous seizures, which can take years in humans).
The quest for an EEG biomarker for epileptogenesis is also revealing new insights into the mechanisms of this pathological process. Several research groups are working to illuminate the molecular mechanisms that generate interictal spikes, HFOs, and changes in theta and nonlinear dynamics (Salami et al., 2014; Jiruska et al., 2017; Milikovsky et al., 2017; Rizzi et al., 2019; Vera and Lippmann, 2021). Such insights stand to reveal novel therapeutic targets and lead to the development of mechanism-specific antiepileptogenic interventions. Moreover, the coupling of mechanism-specific biomarkers with mechanism-specific treatments would pave the way towards early and effective interventions.
As presented in the introduction, one of the advantages of EEG over other clinically used neuroimaging technologies is its portability. However, the devices in clinical use today remain bulky, and EEG monitoring is usually restricted to stationary conditions in lab/clinical settings. Excitingly, the past decade has seen considerable efforts to develop “wearable EEG,” intended to allow untethered data collection during day-to-day activities, in a natural environment, and over prolonged periods of time. Some devices utilize helmets with embedded electrode arrays (Casson, 2019; Boto et al., 2021), while others propose subcutaneous implantation of EEG electrodes (between the scalp and the skull; Duun-Henriksen et al., 2020). Wearable devices also vary in terms of data handling (storing data within device-embedded hardware or transmitting data to a body-worn unit) and data-analysis (real-time versus offline signal processing). Once these technologies overcome several standing technical challenges (e.g., battery capacity, sampling frequency, and electrode performance), they are likely to increase the translational potential of electrographic biomarkers of epileptogenesis.
EEG remains an important tool in epilepsy evaluation/monitoring in clinical settings, and considerable research efforts continue to focus on better understanding EEG patterns and the mechanisms that generate them. Research into EEG signatures of epileptogenesis has primarily relied on animal studies and awaits robust clinical validation. To address the high variability in epileptogenic trajectories, future research should explore the utility of multimodal biomarkers (combining EEG, neuroimaging, and molecular signatures) and dynamic paradigms (assessing subject-specific changes in signature properties over time). A reliable biomarker is likely to accelerate the development of antiepileptogenic drugs and allow biomarker-coupled therapeutic strategies.
This is an open access publication, available online and distributed under the terms of a Creative Commons Attribution-Non Commercial-No Derivatives 4.0 International licence (CC BY-NC-ND 4.0), a copy of which is available at https://creativecommons.org/licenses/by-nc-nd/4.0/. Subject to this license, all rights are reserved.
Your browsing activity is empty.
Activity recording is turned off.
See more...
