3.1 Introduction

Epilepsy and sleep are strongly interrelated at many levels [1–3]. Several epilepsy syndromes present seizures only during sleep, like the nocturnal frontal lobe epilepsy (NFLE), while others like the Landau-Kleffner syndrome, electrical status epilepticus during slow wave sleep, childhood epilepsy with occipital paroxysms and benign epilepsy with centrotemporal spikes, are correlated with seizures occurring during sleep or around awakening [4, 5].

This close relationship between sleep and epilepsy has been recognized since the nineteenth century. William Gowers subdivided epilepsies with regards to their periodicity and highlighted the effects of sleep on epilepsy recognizing the different seizure distribution and semiology of epileptic syndromes during the sleep wake cycle [6].

Seizures or epileptiform activity during sleep, lead to sleep fragmentation and increased arousals, which has serious implications not only to the subjective sleep quality leading to morning tiredness and increased daytime sleepiness or even insomnia but also to the general patient’s quality of life [7]. Results of epileptiform activity during sleep often extend to deterioration of cognitive functions as well, such as memory consolidation [8].

Furthermore, parasomnias, a common comorbidity, seem to infer to sleep quality but to also pose an extra diagnostic challenge in their distinction from nocturnal epilepsies [9, 10].

Antiepileptic medications are also major contributors to sleep disturbances, even though their exact role remains unclear. In studies in patients with epilepsy the direct effects of the drug used are hard to be deciphered from the effects of altered epileptic control or the underlying findings of the condition per se. Yet, antiepileptic drugs modulating GABA and glutamate neurotransmission and altering sodium and calcium channels functions seem to have direct effects on sleep other than their anticonvulsant actions. Newer AEDS though, such as levetiracetam have fewer negative effects and some such as gabapentin or lamotrigine even show some positive effects stabilizing sleep [11–13].

Sleep deprivation aggravates seizures and is known to be an activator of EEG epileptiform discharges, which has led to its use as a diagnostic aid [14]. Yet, since sleep deprivation usually occurs during physical overactivity or psychological distress, one cannot be certain whether the seizure provoking characteristics are due specifically to the lack of sleep per se or are caused by mechanisms underlying stress non-specifically [15].

Because of this reciprocal relationship between sleep and epilepsy, research related to the basic mechanisms of sleep–wake regulation and the anatomic regions involved, is essential for our understanding of the mechanisms that govern epileptogenesis and its clinical expression. Electroencephalographic findings, seizure types and frequency, vary depending on the sleep–wake cycle.

Epileptic seizures and EEG discharges alter the architecture of sleep [16]. They can influence both sleep’s macrostructure, such as increasing stage shifts and the number and the duration of awakenings while reducing and fragmenting REM sleep, but also sleep’s microstructure, such as the Cyclical Alternating Patterns (CAPs) and even phasic events such as K-Complexes and sleep spindles [17].

Epileptic seizures tend to occur during NREM sleep [18] suggesting their relation to hypersynchronization. Furthermore they tend to occur during CAP phases (mainly CAP-A) indicating their correlation to arousals or sudden changes in excitability during sleep [19]. Indeed sudden provoked awakening during sleep is noted as a precipitant of seizures [20].

3.2 Sleep Physiology

In order to shed light on the relationship between sleep and epilepsy, is clearly important to understand the dynamics underlying sleep and epilepsy’s EEG correlates.

Sleep consists of distinct stages which result from a complex interplay between brain regions that are responsible for the transition from the waking to the sleeping state and vice versa [21]. Sleep onset is controlled by two factors (Fig. 3.1a): a circadian influence by the hypothalamic suprachiasmatic pacemaker nucleus and homeostatic influences exerted in hypothalamus by substances like adenosine accumulating in hypothalamus with increasing wake time. According to the proposed “flip-flop” scheme (Fig. 3.2a), this hypothalamic activation inactivates the centers of the upper brainstem which are known to tonically maintain wakefulness, i.e. the ascending reticular activating system containing aminergic (noradrenaline, serotonin and dopamine) and cholinergic neurons, which project diffusely to the entire nervous system (Fig. 3.2b). The homeostatic factor constituting the sleep pressure will gradually decrease as sleep progresses and activation of brainstem arousing centers will start inactivating the hypothalamic ones leading to awakening (Fig. 3.2a). For the duration of sleep, the patterning of sleep stages is additionally influenced by a daylong oscillation of arousal with a period of about 90 min and by the interaction of specific REM-on and REM-off neurons in the brainstem to produce 5–6 cycles in a night’s sleep (Fig. 3.1b), each of which consist of NREM and REM periods [22]. The above sleep promoting centers are located in the anterior hypothalamus preoptic area and contain mostly neurons releasing as neurotransmitters GABA and the neuropeptide galanin. In contrast, posterior hypothalamus is the home of two types of neurons which functionally belong to the activating ones like those in the brainstem: and respectively release histamine and orexin. It is quite impressive that the distinction between hypothalamic sleep promoting anterior and wake promoting posterior areas, was proposed as early as 1920 by Constantin von Economo, based on the lesions he observed in patients with lethargic encephalitis.

Fig. 3.1
figure 1

Sleep macro- and microstructure. (a) Sleep time (S) is controlled by both circadian and homeostatic factors, the latter building up as waking (W) progresses and diminishing slowly after sleep onset. (b) On the basis of EEG, EOG and EMG features a whole night’s sleep is described by a hypnogram (blue) denoting about 6 cycles of about 90 min duration, each containing a NREM and a REM part. NREM includes four stages (N1–N4). Elements of sleep microstructure are shown as K-complexes (KC, green dots), microarousals (MA) or rare awakenings (AW). (c) Time frequency analysis (hypnospectrogram color coding EEG power) for the same sleep reveals the frequent presence of spindles (S, 12–15 Hz) mainly during stage 2 of NREM sleep

Full size image
Fig. 3.2
figure 2

Mechanisms promoting sleep. (a) According to a seesaw model sleep starts when hypothalamic centers like nucleus posterolateralis (VLPO) inactivate those centers of brainstem which maintain wakefulness like locus coeruleus (LC, releasing norepinephrine), tuberomamillary n. (TMN, releasing histamine), raphe n. (RN, releasing serotonin) and others. Additionally VLPO inactivates orexinergic (ORX) neurons which also promote arousal. Arrows and filled circles denote respectively excitatory and inhibitory connections. (b) Simplified diagram of brain circuits involved in sleep onset. AH and L&DH indicate respectively anterior, lateral and dorsal hypothalamus. Ach: acetylcholine, NE: norepinephrine, 5-HT: serotonin, HA: histamine, DA: dopamine, glut: glutamate. While all monoamines (MA, blue curve) decrease throughout sleep, Acetylcholine (red curve) is the only neuromodulator supporting REM sleep. The induced by hypothalamus diminution of ascending inputs to thalamus and cortex promotes oscillatory activity in the frequency of spindles or delta waves which blocks sensory processing from reaching the cortex

Full size image

Additionally, ‘external’ influences (i.e. circadian rhythms and sensory stimuli on the Reticular Activating System—RAS) can ‘switch’ between sleep and wakefulness.

Wakefulness is maintained through the tonic activity in the RAS from the upper brain stem and the thalamocortical and cortical projections from the posterior hypothalamus and the basal forebrain. The transition to sleep is caused by both a withdrawal of external stimuli from the RAS and an activation of the hypothalamus’ preoptic area [23]. This leads to a widespread increase in GABAergic activity that gives rise to the relative synchronous oscillations of NREM sleep such as sleep spindles, delta activity and the slow cortical oscillations.

At the microstructure level of NREM sleep organization, the cyclic alternating pattern (CAP) alternates with periods of quiescence (non-CAP periods) (Fig. 3.4c); NREM sleep is thus divided into periods showing CAP and periods not showing CAP. During the CAP periods, phases containing arousal phenomena and therefore indicate increased vigilance (phase A of CAP) alternate with phases without arousal phenomena, which correspond to reduced vigilance (phase B of CAP) [24].

Sleep spindles (Figs. 3.1c and 3.4a) generated by the reticular nucleus of the thalamus (RE) and its connections to the dorsal thalamus (Fig. 3.2b), consist of a 10–14 Hz activity lasting 2–3 s occurring especially in stages 2 and 3 of NREM sleep. The spindle oscillation is transferred to the thalamocortical relay cells, traveling to the cortex, generating rhythmic excitatory postsynaptic potentials and a synchronous activity over widespread cortical regions that can be observed at the macroscopic/EEG level [25]. Inhibitory RE to RE neurons collaterals desynchronize spindle activity limiting both the amplitude and the duration of sleep spindles [26].

Modeling of spindles from the EEG and MEG signal was remarkably unsuccessful in illuminating the nature of the generators that appeared to be dispersed in different areas with little consistent regularity within and across events. This was rather surprising given the rather widespread and consistent spindle imprint on the EEG (in terms of signal morphology of the sigma-band-pass filtered signal of central and dorsal EEG sensors. One plausible explanation was that the problem arose because the source localization algorithms did not have sufficient power to disentangle activity from a mixture of superficial and deep generators. In recent years a series of studies focused on the problem of identifying the generators of spindles from EEG, MEG, simultaneous EEG and MEG [27, 28], EEG triggered fMRI and invasive measurements in humans [29, 30] begin to converge to a rather surprising but consistent picture: spindles are sporadic events that at least in the cortex occur asynchronously over wide areas and with high variability from event to event. Our own tomographic analysis confirms the high variability of generator loci from spindle to spindle but provides a unified picture of the evolution of activity from the awake state through core periods of light sleep and the periods before and during spindles and K-complexes (Ioannides et al., in preparation). All of the established features of spindles are recovered by the tomographic analysis, but they appear strongly when the appropriate comparison is made between conditions. An example is demonstrated in Fig. 3.3: comparing the spectral density of individual events during and just before spindles shows the anterior foci of the low sigma activity and the more posterior foci in the high sigma periods. In addition and in a more ventral level increases during the spindles are seen for the entire sigma band in posterior cingulate and in the right caudate [31].

Fig. 3.3
figure 3

Spectral statistics of MEG tomographic data showing statistical significant changes of activity common to all four subjects in three narrow frequency bands, alpha and low sigma band (9.6–12.8 Hz) statistical parametric, main sigma band (11.2–14.4 Hz) and high sigma band (12.8–16 Hz). Yellow outline marks common for all 4 subjects statistics at p < 0.05. The green line indicates the central sulcus. The left hemisphere is on the left. The MRIs at the more dorsal/upper (15) and more ventral/lower (11) level are slightly off scale to fit the space provided

Full size image

Spindles appear to play at least three roles. (a) shaping the early organization of sensory-motor cortical circuits [32] (b) maintaining sleep by increasing the threshold perceiving sensory stimuli [33] and (c) helping memory consolidation through a cortico-hippocampal interaction [34].

The delta oscillation (1–4 Hz) has both a thalamic [35] and a cortical component [36] and seems to result from the same circuit that generates sleep spindles. However, spindle and delta activity exclude each other within the thalamocortical neuron, with spindle activity arising at relatively depolarized resting membrane potentials and delta activity arising at more hyperpolarized membrane potentials [37].

Slow cortical oscillations at frequencies bellow 1 Hz, observed during slow wave sleep, are generated in the cerebral cortex with different cortical areas synchronizing in a widespread manner through corticocortical connections. K complexes (Fig. 3.1b), a fluctuating arousal occurring at periodic intervals during NREM sleep, can be triggered by both external sensory stimuli as well as internal very slow sleep oscillations [38]. They appear to have a very dynamic relationship to spindles and to theta bursts (Fig. 3.4a, b) [39, 40].

Fig. 3.4
figure 4

Elements of sleep microstructure putatively involved in seizure expression. (a) K-complex followed by spindles in central EEG electrodes FPZ to OZ. (b) Averaged spectrogram of many K-complexes. Note the interruption of preceding spindle, development of a theta burst and the increased frequency of the spindles following the K-complex. (c) Cycle alternating patter in EEG during NREM with most power concentrating in periods “A”

Full size image

K complexes, sleep spindles and arousals but also epileptiform discharges and seizures (more about this later) seem to occur during the A phase of CAP and seem to be inhibited during the B phase of CAP (Fig. 3.4c) [41].

During REM, the selective reactivation of the RAS cholinergic system with neurons releasing serotonin or norepinephrine on the thalamocortical system remaining inactive, leads to the observed EEG activity which is similar to what it is observed during wakefulness but with muscle atonia. Lastly, the transition to waking, is caused by the activation of the cholinergic as well as monoaminergic nuclei in the brainstem (Fig. 3.2b) and a resultant activation of the thalamocortical cells [42].

Thanks to the tomographic analysis of MEG data the changes during sleep can be studied non-invasively using MEG. The first tomographic analysis of whole night MEG data revealed the complexity of the processes involved [43] but also provided hints of a global organization and a relatively smooth transition from one stage to the next when the quiet rather than active periods of each sleep stage were compared [44]. The most consistent change running through all sleep staged was identified in near-midline areas on the left hemisphere, one in the frontal lobe and the other in the posterior mid-parietal area. In these two areas the gamma band activity was higher during REM sleep than other sleep stages and active wakefulness. A meta-analysis of recent neuroimaging studies showed that these two areas are at the centre of foci of increased activity identified in experiments with increased resting state activity compared to task periods (areas belonging to the “default system”) and areas identified in experiments requiring understanding own and other peoples intention and introspection (areas of the “Theory of Mind” system). It seems significant that the areas identified in the earlier sleep study, the default mode areas and the Theory of mind area were not just randomly placed with respect to each other. The areas of increased gamma band activity were at the centre and not overlapping with the areas of the two other systems, suggesting they play a pivotal role in maintaining the person’s identity through sleep. This makes us dare pushing beyond the suggestion of Domhoff that the default network might be involved in dreaming [45, 46]; we could suggest that the areas identified as super-active in the gamma band during sleep, and especially the dorsal medial prefrontal cortex are acting like a “Dream Box” releasing consciousness while the rest of the brain is largely subdue. Our recent research, focusing on light sleep suggests that an area close to the “Dream Box” areas is also important for spindles (Ioannides et al., in preparation).

3.3 The Interictal State and Sleep

Abnormal electrical discharges occurring in the time between seizures are referred to as ‘interictal spikes’. The summation of this abnormal electrical activity along with the signs and symptoms, within a broad psychiatric and behavioral range, is referred to as the ‘Interictal State’ [47].

Interictal spikes consist of abnormal discharges of an inappropriately synchronized population of neurons from focal brain areas. Unlike seizures, interictal discharges do not spread across large brain areas and they do not cause clinical symptoms [48].

Generalized spike-wave discharges in idiopathic generalized epilepsy aggravate as NREM sleep progresses and after awakening, while they diminish during REM sleep. These discharges tend to occur during CAP-A phases of NREM sleep. A study on patients with JME showed that their distribution in CAP phases (A or B) may relate to (or even predict) seizure control. Increased epileptic pressure may cause disruption of the inhibitory mechanisms of phase B, increase the CAP rate by contributing to more A phases, and thereby foster more epileptiform discharges through the CAP A window. In other words there seems to be an epileptic positive feedback with clinical correlates: increased seizure activity is associated with enhanced intrusion of spike wave activity into phase B of CAP sleep, increased CAP rate, more epileptiform discharges and by implication higher probability of having more seizures. Increased electrographic awakenings fragment sleep and may independently contribute to the clinical deterioration by impairing sleep quality [49]. Similar findings were noted in childhood absence epilepsy [50]. Also, generalized spike–wave discharges alter morphologically during NREM sleep [51, 52].

Also, in partial epilepsies NREM activates interictal discharges while REM suppresses them without being clear though, which stages of NREM aggravate discharges; some studies suggest that interictal discharges increase during NREM stages 1–3 [53] and others suggest that they increase during NREM stages 3–4 instead [54].

Besides NREM’s effect on the rate that interictal discharges occur, discharges seem to also be more widespread in terms of their partial extent during NREM than during wakefulness or REM sleep. Yet, it has been debated that interictal discharges observed during NREM are of a lower value to localizing the true epileptic focus in contrast to activity observed during REM and wakefulness [54], since there is a large percentage of novel epileptiform foci, unrelated to the true epileptic focus, observed during NREM even from the contralateral hemisphere.

Lastly, ripples (80–250 Hz) and fast ripples (>200–250 Hz), which seem to be generated within the neocortex or the hippocampus, and have been correlated to the interictal to ictal transition [55] also seem to be enhanced by NREM sleep [56]. Schevon et al. [57] suggested that fast ripples are produced by cortical domains near locally excitable clusters that produce microdischarges and are better correlated to interictal epileptiform events. High frequency oscillations, the increasingly recognized biomarkers of the epileptogenic zone [58, 59] are facilitated by slow waves in NREM sleep [60].

All mentioned above, indicate the relation of epileptiform activity to the increased neural synchronization that is observed during NREM sleep as it is to be discussed below.

3.4 Epileptic Seizures and Sleep

Similarly to interictal discharges, seizure activity is also correlated to sleep and it is modified both by the sleep wake cycle as well as the sleep stage. Early observations since the nineteenth century have tried to classify epilepsies into diurnal vs. nocturnal or diffuse [61] and Janz at 1962 [62] described the awakening epilepsies, which occur upon or soon after awaking and seem to be primary generalized in nature.

Idiopathic generalized epilepsies (IGE) include awakening epilepsies while only a minority of them may present with EEG abnormalities seen only during sleep; even then, seizures (absences, myoclonic and tonic-clonic) occur during wakefulness. Idiopathic focal epilepsies can occur during sleep in as much as 80 % of cases [63]. Frontal lobe seizures typically occur during NREM sleep, particularly during stage 2 (N2) (Fig. 3.5—nocturnal seizure with genital automatisms) while temporal lobe seizures are more likely to generalize during sleep [64]. In NFLE the vast majority of seizures will occur during sleep with NREM sleep enhancing focal epileptiform activity in association to the A phase of CAP A [65].

Fig. 3.5
figure 5

Nocturnal frontal lobe seizure, seemingly arising from N3. The actual seizure discharge starts at the blue arrow, but sleep stage has already changed from N3 to lighter 3 s before (green arrow) as evidenced by the diffuse increase of faster rhythms. The patient engages in bilateral genital automatisms of which he has no recollection

Full size image

Epilepsy with continuous spike wave during slow wave sleep (CSWS) and Landau–Kleffner syndrome, are also strictly correlated to the sleep wake state, both showing a continuous spike wave activity during sleep [66].

Sleep stage is a key factor in seizure occurrence and characteristics. NREM sleep facilitates seizure onset and seizure spread whereas REM suppresses them [14]. Slow wave sleep is the state of maximum synchronicity in the brain while K complexes and sleep transients that are often correlated to epileptiform activity during lighter stages of sleep, are related to patterns of periodic arousal instability as it is described by the CAP [65], both indicating the relation to the hypersynchronization and hyperexcitability which characterize epileptiform activity.

Finally, sleep deprivation has been correlated to seizure inducing and precipitating epileptiform discharges possibly by inducing NREM sleep but also through affecting cortical excitability [14].

3.5 Mechanisms Underlying the Effect of Sleep on Epilepsy

3.5.1 Sleep Mechanisms Altering Brain Synchrony and Excitability

The theoretical background of the mechanisms involved in the interaction of sleep and epilepsy include the existence of shared neuronal circuits between sleep and epilepsy, the increased synchronization that is evident during certain sleep stages and seems to facilitate epileptiform activity, and various intrinsic characteristics of the epileptic focus which are influenced by sleep related activity.

Epileptiform activity is generally characterized by hypersynchronization and hyperexcitability. During sleep and wakefulness the level of neural synchronization varies. During NREM sleep there is increased neural synchronization compared to REM sleep or wakefulness. This suggests that different levels of neural synchronization can interfere with the abnormal synchronization occurring during ictal and interictal discharges, promoting it during NREM or suppressing it during REM or wakefulness [14].

Yet neural synchronization during NREM varies [67] and frequencies observed in the delta, theta and sigma band fluctuate during the course of sleep. There are conflicting data on what aspects of the neural activity during NREM promotes epileptiform discharges. The depth of sleep and the log delta power [68], the progression through deeper stages of NREM sleep or lighter stages of NREM sleep [69] and changes in the power in the sigma or theta band [53, 70–73], all have been correlated with the rate of discharges but fail to clearly predict which of the activity seen during NREM promotes discharges. A possible explanation of these varied data lies in the need of a necessary concurrence of neuronal activity of the epileptic loci with the activity of the rest of the brain temporally and spatially. Sigma band is best observed in the frontocentral regions; therefore it is expected to predict discharges coming from these areas [14]. Similarly the epileptic loci may carry some intrinsic rhythmicity itself which could fall within several frequencies normally occurring in the brain. Therefore, a similar extrinsic frequency of the brain would boost and be related to the intrinsic frequency of the epileptic loci. If the intrinsic frequency falls within sigma band for example, it is expected to be better correlated to power in the sigma band or sleep spindles [74].

Α recent study showed that during NREM sleep epileptic spike discharges and high frequency oscillations appeared not continuously but with highest rate in association to larger slow waves and particularly during the highly synchronized transition from “down” to “up” states underlying these waves; not during the “up” states, which is the case with physiological activity [60]. The authors concluded that the activation of epileptic discharges during NREM sleep is not a state-dependent phenomenon but is predominantly associated with specific events, and apparently facilitated by increased synchronization rather than by increased excitability. Understanding the prime cause however remains as challenge, because dynamic bistability of neuronal membrane potentials widespread synchronization are mutually dependent and reinforced (see [3]).

On another note, intrinsic features of the epileptic focus may be responsible for the varied behavior regarding the enhancement or otherwise a relative immunity to the extrinsic synchronized rhythm which is present during NREM sleep [74]. There is data in literature where epileptic foci can be selectively enhanced by NREM in contrast to the rest of the brain or, in other cases, epileptic foci may lose the ability to be modulated by sleep related activity. This could explain the persistence of local hyperexcitability and hypersynchronization of a local area even in the absence of similar extrinsic activity. Yet, for the epileptiform activity to be spread to other regions, as seen to the intractable symptomatic and secondary generalized epilepsies, the synchronized activity that is present during NREM can facilitate the transmission of the epileptiform activity to the proximate normal cortex [74].

Additionally, in terms of generalized epileptiform discharges, it is speculated that sleep events and rhythms such as spindles, K complexes and delta oscillations share common or overlapping neuronal circuits involved in the generation of generalized discharges.

Spike and wave discharges (SWDs), the electrographic hallmark of typical absence seizures, which are an integral component of several idiopathic generalized epilepsies [5], have been reported to occur preferentially during the light stages of NREM sleep, where the majority of sleep spindles are observed and in a reverse relationship to their rate throughout the night. Gloor in 1978 [75] proposed that the same TC circuit producing sleep spindles could generate SWDs in states of cortical hyperexcitability. The hypothesis found support and mechanistic explanation in a series of experiments in the model of feline generalized epilepsy with penicillin (Fig. 3.6a) and developed further on the basis of in vitro and in vivo experiments, especially those revealing the neuronal mechanisms of spindles generation [76] and those very fruitfully using rodent genetic models of absence seizures [77–83].

Fig. 3.6
figure 6

(a) Diagram of suggested successive changes in membrane potential of PN during the transition from spindles to SWDs in the feline model absence seizures, as deduced from different experiments at the intracellular (i.c.) level, at successive times following small doses of barbiturates and penicillin. The observation that under these conditions spindles appear to be gradually replaced by SWDs of half or a third of the spindles frequency (in a step-like transition), lead to the following hypothesis: SWDs may be generated because the pyramidal neurons’ usually subthreshold excitatory postsynaptic potentials (EPSPs) corresponding to each EEG spindle wave (top trace), upon an increase in cortical excitability, progressively (from top to bottom traces) become more and more effective in firing and so they activate cortical interneurons whose increasing inhibitory postsynaptic potentials annul the next EPSP and further on the next two EPSPs. (b) Diagram of intracellular (i.c.) events underlying EEG characteristics of generalized tonic-clonic seizures (produced by higher doses of penicillin in the same experiment), serves to show the fundamental difference in mechanisms underlying the two seizure models: Excitability of interictal spike (1) is not just increased passed threshold but clearly abnormal (paroxysmal depolarization shift). Inhibition, while instrumental in SWDs, is maintained only in (1) and is completely lost during the tonic (2) and clonic (3) phases. Finally there is postictal depression (4). Modified from [79]

Full size image

One of the most important recent discoveries in the field has been the identification of a cortical ‘initiation site’ of SWDs of rodent absence seizures [84]. Also high density EEG as well as MEG and fMRI studies in patients with different types of idiopathic generalized epilepsy have shown SWDs in discrete, mainly frontal and parietal cortical regions before they appear over the rest of the cortex [85–89]. This novel view is consistent with the above hypothesis (Fig. 3.6a) because they describe variable expressions of—the inherent in this hypothesis—cortical hyperexcitability/instability, not with the readiness of the thalamocortical system to so accurately synchronize the entire brain at specific frequency, which can be presumed to be a product of long latent period epileptogenic process on a specific genetic background. This epileptogenic process appears to cardinally involve changes in the thalamocortical mechanisms which generate sleep spindles.

These studies emphasize the importance of (a) mutual interaction between the sleep and epilepsy, (b) recognizing that different types of epilepsy may have fundamentally different mechanisms (compare Fig. 3.6a to Fig. 3.6b) and (c) the significance of observing the “bigger picture” in both time (i.e. CAP periods) and brain space, since both sleep and epilepsy by definition involve dynamic changes in large brain circuits.

Εpileptiform discharges can be an expression of corrupted normal sleep rhymes [26]. Both sleep spindles and spike-wave discharges can be generated by modulating the degree of GABAergic inhibition in the thalamus [74].

The reticular nucleus (RE) intrinsic oscillatory properties generate the sleep spindles, which are then amplified by the reciprocal connections from the thalamocortical neurons to the RE. In addition downward projections from the cortical neurons to the thalamocortical and RE trigger and synchronize them across cortical neurons. GABAergic neurons inhibit the thalamortical neurons which project via glutamatergic synapses to cortical neurons and back to the RE. Also, inhibition among RE neurons limits synchronization within the RE thus limiting size and duration of spindles. RAS and sensory excitatory input can further modulate activity of the cortical, thalamocortical and reticular neurons [74].

In the theory of the spindle generating thalamocortical circuit, the reduced GABAergic inhibition within the reticular nucleus which would increase the degree of synchronization in the RE—thalamocortical neuron circuit or otherwise the increased excitation of RE and/or thalamocortical neurons, can lead to the replacement of sleep spindles by thalamic oscillations that are synchronized and epileptiform. Increased excitation to the RE can also explain the alteration of the frequency from the spindle frequency (~10 Hz) to the spike-wave frequency (~3 Hz) because of a possible alteration of GABAb to GABAa response to the RE-thalamocortical inhibitory synapse [74].

Indeed, for some cases, seizures are better correlated to sudden excitability during sleep as it is noted during CAPs and arousals or microarousals than increased synchronization during NREM sleep [19, 49, 50]. Sudden provoked awakening during sleep is a well described trigger of epileptiform discharges and seizures [20]. Also this includes seizures occurring upon awakening, seizures correlated with K complexes and seizures during the active A phase of the cyclic alternating patterns. Sudden excitability can concern a local increase of excitability within the epileptic focus [19] or otherwise a diffuse excitation caused by glutamatergic and cholinergic projections upon arousal/ awakening.

3.5.2 Neuromodulation in Sleep and Epilepsy

Both pro- and anticonvulsive properties have been attributed to monoaminergic systems involved in sleep (Fig. 3.2b). Serotonergic and histaminergic systems involved in arousal are considered to have antiepileptic actions, while changes in the norepinephrine system have been implicated in absences [3, 90]. Melatonin, also a monoamine, has been found to be mostly anticonvulsive (see later discussion). The cholinergic system, which is involved in “REM-on” and arousal, has been shown to have anticonvulsive properties [3].

Adenosine, an endogenous hypnogen, is under investigation as a therapeutic target for seizure control. Adenosine is an inhibitory neuromodulator [91] which through A1 receptors that can act by blocking glutamate release but also independently of GABA and glutamatergic systems [92]. Its concentration as well as its receptor density has been shown to rise after seizures [93], with the latter remaining increased for a prolonged period of time. Endogenous adenosine blocks epileptiform discharges in human epileptogenic cortex maintained in vitro [94]. This endogenous anticonvulsive role is further increased by its neuroprotective properties [3].

On the other hand, the peptide orexin, which is involved in arousal from sleep (Fig. 3.1a, b), may favor epileptic activity. Orexin has been classically linked to the parasomnias (narcolepsy in particular) [95]. It has been reported to induce behavioral seizures, while an increase in its levels has been shown in both animal models of epilepsy [96] and in patients who experienced seizures during polysomnographic examination.

3.5.3 Circadian Rhythms and Epilepsy

Although the relationship between sleep and epilepsy is a well-established one, it is crucial to understand that the sleep-wake cycle (Fig. 3.1a) is, albeit the most pre-eminent and the easiest to identify, just one of many endogenously controlled physiologic processes, cycling at a 24-h period. Those circadian rhythms also include hormone production, temperature variation and cardiovascular pattern regulation, and are controlled by both photic and non-photic environmental time cues (zeitgebers) [95]. At the core of the circadian clock lie pacemaker neurons in the suprachiasmatic nucleus of the hypothalamus. These receive input from photosensitive ganglion cells in the retina (retinohypothalamic pathway), and are thus entrained to light–dark cycles [97]. Other afferent connections are the geniculohypothalamic and the median raphe serotonergic pathway, while efferent connections include output to the pineal (which produces melatonin), the subiculum and the hippocampus [98]. On a molecular level, the rhythm of SCN cells is regulated by a negative feedback loop of gene expression (such as CLOCK and BMAL1), is independent of action potentials and remains intact in zeitgeber deprivation. SCN neurons mainly use GABA and inhibit the neurons that they innervate [97, 98]. The molecular mechanisms of circadian rhythms have been closely linked with epilepsy. Membrane excitability is altered due to changes in potassium, regulated by clock gene products. The expression of hippocampal genes and ligand-binding activities (α1 adrenergic and benzodiazepine receptors) also oscillate depending on the SCN. In addition, transcription factors involved in epilepsy (such as those regulating the expression of the GABAα1 receptor subunit) have been shown to be important in circadian rhythms. Lastly, the SCN has been reported to be involved in a major signaling pathway of the acquired epilepsies (mTOR pathway, reviewed in [97]). Recently, a reduced seizure threshold was found in a knockout mouse model for the BMAL1 clock gene [99]. The link between circadian rhythms and epilepsy, is responsible for the effect of epilepsy on various circadian functions. People with epilepsy have limited diurnal pressure and heart-rate variability, a fact linked to sudden unexpected deaths (SUDEP). Recent findings suggest that SUDEP, results from depolarization spreading to and inactivating brainstem nuclei supporting vital cardiorespiratory functions [100].

Seizures have been shown to disrupt normal temporal rhythms of hormone production, with postictal elevations of cortisol, prolactin and growth hormone [95]. Melatonin, whose production is directly influenced by the light–dark cycle, has been extensively investigated regarding its relationship with seizures. Finally the monoamine melatonin reduces glutamatergic neurotransmission, while enhancing GABA-related activity. One of its metabolites (kynurenic acid), has also been shown to have antiepileptic properties, with others also being neuroprotective. While the majority of animal and patient studies have reported in vivo anticonvulsive effects, there are also data that contest the efficacy of melatonin as an add-on treatment for seizures [101].

3.6 Polysomnography in Long Term Monitoring of Epilepsy

Monitoring brain and body activities of patients with Epilepsy is a very important procedure that has proven to be very useful for clinical and research purposes. Polysomnography recordings is the most suitable mean to monitor these activities for a long period as it integrates many modalities such as Electroencephalogram (EEG), Electro-oculography (EOG), Electrocardiogram (ECG), Electromyogram (EMG). Each polysomnography modality is specialized to monitor effectively a single activity of the patient and many studies [102–105] have revealed modality specific patterns in sleep which could be used as biomarkers for epilepsy prognosis and diagnosis or could enable the accurate epileptic seizure detection.

3.6.1 Open Problems in Polysomnographic Monitoring of Epilepsy

Despite the great advancements in long term monitoring of epileptic patients there exist many unsolved problems until now.

First of all, there exist many different views on how much of the polysomnographic data should be recorded and stored in view of the space and power limitations of the existing devices. Some researchers and clinicians claim that only recording and keeping the data identified by the seizure and spike detection programs (usually supplemented by periodic sampling of the wake and sleep background activity) is sufficient. However, this may mean that atypical and subclinical seizures are not recorded. Moreover, any subtle changes (in e.g. EEG or ECG) that may occur before sustained ictal electrographic activity may be permanently lost. With the availability of large capacity and relatively inexpensive digital storage media, the continuous recording of the polysomnography signals and subsequent archiving of the data has become a more practical and economic option. However, new advancements in low power consumption devices and storage devices alongside with the incorporation of compression techniques in them are required to enable the efficient recording of polysomnographic “big data” [106].

Another problem that remains unsolved in spite of the efforts of the scientific community is the integration of different types of modalities. Integrating different types of modalities can be more effective than the utilization of any modality alone, for problems such as spatiotemporal pattern discovery. Effective integration aims to exploit the advantages of every modality in order to achieve accurate results, and is considered to be an essential task in modeling human brain activity. However, the integration of more than two modalities in a single monitoring device is not provided by most of the current monitoring devices and methods.

Existing monitoring devices for epileptic patients have emphasized so far to the problems of seizure prediction, and focus localization. However, much progress have been accomplished so far in the fields of algorithmic epilepsy differential diagnosis and therapy advice and these advancements should be included in monitoring devices to improve their functionality. A striking example in this direction is therapeutic devices for epileptic patients. Fisher [107] presented the so far limited available devices that not only attempt to detect seizures but also try to treat them using Responsive neurostimulation techniques or Optogenetics techniques (the latter in experimental animals). In this article it is mentioned that optimizing the place of devices in therapy for epilepsy will require further development and clinical experience.

Finally, another open issue is the incorporation of other non-standard modalities in monitoring devices for epileptic patients. For example, in the work of Beniczky et al. [108], a wireless device and a method were proposed for the detection of generalized results using only wrist accelerometer recordings. The experimental results indicated that this methodology achieved accuracy similar to a video–electroencephalography method, indicating that simpler alternative modalities may perform difficult tasks as good as more complex, time consuming and energy intensive techniques. Thus, an interesting future direction is the exploration of the potential of these simple alternative modalities and their incorporation in long term monitoring devices for epileptic patients.

3.6.2 The ARMOR Approach

As already mentioned in the present chapter sleep and epilepsy are highly related. In order to efficiently monitor patients with epilepsy it is extremely crucial to record and analyze polysomnographic signals during their sleep. It is even more important for every monitoring system not to disturb sleep or deteriorate its quality.

ARMOR platform, is a wireless, power efficient device which could incorporate many polysomnographic modalities. By design it does not disturb sleep significantly avoiding to use modalities such as video recordings which are believed to stress subjects. Multiple modalities recordings are integrated, streamed and stored to a specially designed database. The quality of the EEG signal of ARMOR platform is more than adequate to locate and analyze essential sleep microstructure events, such as sleep spindles, k-complexes and microarousals, alongside with other sleep macrostructure events such as sleep stages and CAPS. A variety of online algorithms have been incorporated to ARMOR platform to enable the location of sleep onset with the detection of alpha rhythm block and the detection of epileptic interictal spikes and seizures. Moreover, ARMOR polysomnographic sleep recordings can be further analyzed using ARMOR offline algorithms to answer more complex clinical questions. Another advancement of ARMOR platform is its advanced alert system which could help clinicians interfere during the subject’s sleep when this is required, e.g. during an epileptic seizure. For all these reasons, the polysomnographic recording component of ARMOR project is an advancement compared to the state-of-the-art as it deals with most of the aforementioned open problems.