Epileptology as a field still suffers from a measurement issue. To unlock the full potential of monitoring the biomarkers of epilepsy—electrographic seizures and interictal epileptiform activity (IEA)—the ideal measurement method should be always available to continuously record epileptic brain activity. Additionally, the method should be safe, unobtrusive, noninvasive or minimally invasive, and scalable for the broad patient population with epilepsy.

Self-reporting in epilepsy

To this date, the gold standard outcome in epilepsy is patient-reported seizures. Yet, it is well known that patients tend to over- or underreport seizures [5,6,7]. Even well-kept diaries may document symptoms that are not actually epileptic seizures, an overreporting problem. Conversely, patients tend to underreport seizures occurring out of sleep or seizures with loss of awareness [5, 7]. Despite the inaccuracy of self-reports, studies based on seizure calendars have shown that seizures occur inhomogeneously over time in many adults, sometimes with striking about-weekly to about-monthly periodic patterns detectable by eye [1], and sometimes with patterns evident only with suitable statistical methods [8]. Other studies quantified catamenial seizures in the diaries of women with epilepsy [9, 10], but about-monthly seizure cycles also exist in men (Fig. 1). Across studies and patients, the patterns observed were generally weak, due to the use of techniques that did not account for nonstationarity in epilepsy cycles [8,9,10], as detailed below. Thus, cycles of self-reported seizures may represent the “tip of the iceberg” and are confounded by nebulous reports, such that there is a need to quantify seizures reliably in the daily lives of people with epilepsy.

Fig. 1

Seizure diary kept by a male patient. a Note the striking about-monthly regularity with which single seizures or seizure clusters occur. In this case, a chronotherapeutic reinforcement of the antiseizure medication during or even before the expected cluster occurrence may prove helpful. b Each vector represents one seizure and points to its time of occurrence on the 24‑h clock (with approximate day and night as shading). Seizures were reported throughout the day except between midnight and 06:00, raising the question of a protective or amnestic effect of sleep

Undersampled conventional EEG

In the opinion of many practicing neurologists, EEG remains the best tool for quantifying seizure frequency because it is more objective than reports of patients’ experiences. Although electrographic seizures may not always be perceived by patients or be clinically salient, it is reasonable that even these “silent” seizures should be treated. Indeed, electrographic seizures can propagate throughout the brain and lead to cognitive issues that are only identifiable with in-depth neuropsychological assessments. Yet, the chances of capturing electrographic seizures with short-term EEGs (e.g., < 1 h) are low [11]. This has led many hospitals to develop costly seizure monitoring units that offer long-term EEG (recording duration of 2–3 weeks). Although these units are invaluable for localizing seizures in preparation for surgical treatments, they do not represent a scalable solution to monitor epileptic brain activity chronically. Thus, currently, conventional scalp EEG of greater duration than 3 weeks is highly limited, and this severe temporal undersampling impedes progress in clinical epileptology [12].

Chronic EEG

Chronic EEG recording devices are surgically implanted and can record EEG over months to years, overcoming the limitations of conventional scalp EEG. To date, only three devices have had the capability of recording intracranial EEG continuously in human patients.

1.

A decade ago, a trial of the NeuroVista device showed the possibility of continuously recording and storing intracranial EEG (16 channels) in 15 participants. However, the device was never commercialized due to a lack of investment beyond 2013, but analyses of the recordings have shown the presence of circadian and multidien cycles [13, 14].

2.

In 2013, the RNS System (NeuroPace) received approval in the United States for commercialization as a therapeutic intracranial cortical neurostimulator with eight electrodes and to date has been implanted in more than 6000 people with epilepsy. Although the benefit of this therapy is palliative (as opposed to curative), it has limited recording capability that has been instrumental in uncovering long-term cyclicity in focal epilepsy [2, 15].

3.

Medtronic has conducted a trial of a device made for epilepsy (RC + S device) in five patients and then broadened the indication of the commercially available Percept PC deep brain stimulator from movement disorders to epilepsy. Unfortunately, the device has very limited recording capability (average local field potential in defined frequency ranges) in the ambulatory setting.

The concept of subscalp EEG has placed itself as the next frontier for a broader deployment of implantable EEG recording systems [16]. Subscalp EEG is a minimally invasive method in which electrode leads are placed beneath the scalp, but above the skull, minimizing the risk of intracranial complications [17]. The quality of the signal is comparable to that of scalp EEG, but the subcutaneous electrodes do not require maintenance. To date, three types of subscalp electrodes have been implanted in patients.

1.

One such device (SubQ 24/7, UNEEG) is on the European market and is in trials in the United States. The device offers a “keyhole” view on brain epileptic activity, sampling from one side of the head with three electrodes [18].

2.

The Epiminder subscalp EEG system (Minder) is a four-electrode linear lead with two bihemispheric channels and has yielded ongoing continuous recordings up to 1 year [19]. The device is being trialed on 16 patients in Australia.

3.

The Epios system offers full-head coverage (28 channels) with a modular trident-shaped electrode lead that was recently tested in a trial in Switzerland [20].

Seizure cycles have been found with these recording devices, regardless of whether they also acted as neurostimulators [2, 21] or were implanted into the skull [13] or beneath the scalp [19, 22]. This independence from the recording modality lends credence to the generalizability of seizure cycles.