Resetting of Brain Dynamics: Epileptic versus Psychogenic Non-Epileptic Seizures

Abstract

In this study, we investigated the possibility of differential diagnosis of patients with epileptic seizures (ES) and patients with psychogenic non-epileptic seizures (PNES) by an advanced analysis of dynamics of the patients' scalp electroencephalograms (EEG). The underlying principle was the presence of resetting of brain's pre-ictal spatiotemporal entrainment following onset of ES and the absence of resetting following PNES. Long-term (days) scalp EEGs recorded from five ES and six PNES patients were analyzed. It was found that: (a) Pre-ictal entrainment of brain sites was reset by epileptic seizures (p<0.05) in 4 out of the 5 patients with ES, and not reset (p=0.28) in the fifth patient. (b) Resetting did not occur (p>0.1) in any of the 6 patients with PNES. These preliminary results in patients with ES are in agreement with our previous findings from intracranial EEG recordings on resetting of brain dynamics at ES and it is expected to constitute the basis for the development of a reliable and supporting tool in the differential diagnosis between ES and PNES. Finally, we believe that these results shed a novel light on the electrophysiology of psychogenic epilepsy by showing that occurrence of PNES does not assist patients to overcome a pathological entrainment of brain dynamics.

1. Introduction

1.1. Epileptic seizures (ES)

Epileptic seizures are due to sudden development of pathological, synchronous neuronal firing in the cerebrum and can be recorded by scalp, subdural and intracranial electrodes. ES may begin locally in portions of the cerebral hemispheres (partial/focal seizures with a single or multiple foci), or simultaneously in both cerebral hemispheres (generalized seizures). After a seizure's onset, partial seizures may remain localized and cause relatively mild cognitive, psychic, sensory, motor, or autonomic symptoms, or may spread (secondarily generalized) to cause altered consciousness, complex automatic behaviors, or bilateral tonic-clonic convulsions, thus severely disrupting the brain's normal multi-task and multi-processing function. We have shown in the past [1] that epileptic seizures are not abrupt transitions into and out of an abnormal ictal state, but instead they follow a dynamical transition that evolves over minutes to hours. During this pre-ictal dynamical transition, multiple regions of the brain progressively approach a similar dynamical state.

Epileptic seizures typically reset the pre-ictal dynamical entrainment and lead to the disentrainment of dynamics of the focus from the rest of the brain. (When ES do not reset the established pathology in brain's dynamics, cluster of seizures and/or status epilepticus may result [2].) We have called this reversal of dynamics brain resetting at epileptic seizures and we have observed it in focal as well as generalized seizures, within and across patients [3-9]. We also have observed it using other measures of brain dynamics [10]. Furthermore, the observed dynamical resetting in patients with epilepsy is significantly sensitive (p<0.01) and specific (p<0.05) to seizures [7]. Other groups have independently observed similarly reversal trends using more classical methods of signal processing [11, 12]. This observation may reflect a passive mechanism (e.g., high electrical activity during a seizure depletes critical neurotransmitters and thus deactivates critical neuroreceptors in the entrained neuronal network). An alternative explanation is an active mechanism, that is, seizure activity releases neuropeptides that may subsequently contribute to the temporary repair of a pathological feedback network that allowed the dynamical entrainment to occur and last pre-ictally for tens of minutes. Such an explanation is analogous to mechanisms attributed to seizures associated with electroconvulsive therapy (ECT) [13, 14].

1.2. Psychogenic Non-Epileptic seizures (PNES)

In a recent critical review, Bodde et al. [15] presented a working definition of PNES as “an observable abrupt paroxysmal change in behavior or consciousness, that resembles an epileptic seizure, but that is not accompanied by the electrophysiological changes that accompany an epileptic seizure or clinical evidence for epilepsy”. Despite being discussed in the scientific literature for over a century, the etiology and mechanisms of PNES are not well understood. Although there is no consensus on the psychogenic features that lead to PNES, it is believed that a combination of psychogenic mechanisms and “trigger mechanisms” are at play [16, 17]. A traumatic experience in the past is found in 90% of PNES patients [18-21]. Combined with the lack of understanding of the physiological mechanisms of PNES is a lack of understanding of the clinical manifestations of PNES. The presentation of PNES can also be indistinguishable from epileptic seizures [22].

The average delay in diagnosis of psychogenic non-epileptic seizures (PNES) is more than 7 years. Neurologists diagnose PNES based on “seizure semiology, psychiatric history, seizure provocation techniques, postictal prolactin assay, and psychological testing” [22-24]. The gold standard for diagnosing PNES is currently video-EEG monitoring (VEM) [23, 25-26], but it has its limitations [27-29]. While VEM can misdiagnose PNES for epilepsy, epileptic seizures can also be misdiagnosed as PNES [17, 30]. Although the prevalence of PNES (2 to 33 per 100,000 [23, 31]) is much less than epilepsy (4-6 per 1,000 [23, 32]), 25% to 30% of the patients referred to epilepsy centers for long-term EEG monitoring for spell characterization are eventually diagnosed with PNES [15, 33-34]. The situation is complicated by the finding that 5% to 40% of PNES patients also have epileptic seizures [15, 35-36]. Due to the high number of PNES patients in epilepsy centers and epilepsy monitoring units (EMUs), the ability to differentiate reliably and quickly between epileptic and non-epileptic events is critical for proper treatment of PNES patients and for reduction of the associated economic burden.

There are at least three major concerns with the misdiagnosis of PNES as epilepsy. First, prognosis for PNES is worse if wrong diagnosis is perpetuated as the appropriate treatment is not prescribed. Second, if PNES is misdiagnosed for ES, and antiepileptic drugs (AEDs) are taken, unnecessary side effects may result. Third, the financial cost may be substantial [37]. The considerable social stigma attached to epilepsy can lead to patient hostility when the diagnosis is changed from epilepsy to PNES, especially in those patients who have been misdiagnosed for a long period of time [15, 38]. It is estimated that up to 75% of PNES patients, who do not have concomitant epilepsy, are initially treated with antiepileptic drugs (AEDs) [18, 39-40] and can suffer from debilitating side effects. Estimates of the annual cost of misdiagnosing PNES ranged from $0.5 to $4 billion in the 1990's [41]. Another attempt at such an estimation concluded that the cost of misdiagnosing and treating PNES is similar to the cost of treating intractable epilepsy, which in 1995 was approximately $231,432 per patient [17]. Sadly, these three broad categories of concern are so significant because of the general principal that once a patient has been diagnosed with epileptic seizures it is perpetuated and requires an unusual intervention before the initial diagnosis is overturned [41-43]. For many reasons (e.g., infrequent episodes, not witnessed by medical personnel, or superficially resembling epileptic seizures), there can be a delay in the diagnosis of PNES, estimated to be between 7 and 10 years [40, 43-44]. Delays in diagnosis have been prognostically associated with poorer likelihood of remission of PNES [43-46].

Our prior investigations with a small number of PNES patients [47] showed that their seizures do not reset pre-ictally entrained brain sites. In particular, we used measures from chaos theory to analyze available long-term scalp electroencephalograms (EEGs) recorded from two PNES patients in order to quantify the probability of resetting of the brain's spatiotemporal entrainment following PNES. We then compared this likelihood of resetting of brain's entrainment following PNES with ones at other times randomly selected from interictal period of the recording in the same patient. Our results showed no significant difference between brain resetting following PNES and at interictal periods in either patient (p-values of 0.71 and 0.24 respectively). We herein present results along similar but statistically stricter lines of analysis of long-term scalp EEG from 6 PNES patients that further support those initial findings. In addition, we compare these results with corresponding ones derived from analysis of long-term scalp EEG from 5 patients with epileptic seizures.

2. Methodology

2.1. Measures of Nonlinear Dynamics

The brain is inherently a nonlinear and nonstationary system. Among the important measures of the dynamics of a nonlinear system are the Lyapunov exponents that measure the average information flow (bits/sec) the system produces along local eigendirections through its movement in its state space [48-50]. Positive Lyapunov exponents denote generation of information while negative exponents denote destruction of information. A chaotic nonlinear system possesses at least one positive Lyapunov exponent, and it is because of this feature that its behavior looks random, even though as a system it is deterministic. Methods for calculating these measures of dynamics from experimental data have been published [51, 52]. Iasemidis et al. have shown that for a non-stationary system with transients like epileptic spikes, using the short-term maximum Lyapunov exponent (STL_max) is a more accurate characterization of the average information flow than the one using the regular maximum Lyapunov exponent (L_max) [48, 49]. STL_max is estimated from sequential EEG segments of 10.24 sec in duration per recording site over the entire EEG recording to create a set of STL_max profiles over time. The STL_max profiles computed over space and time characterize a spatiotemporal chaotic signature of the epileptic brain.

2.2. Dynamical Entrainment

Analyzing scalp, subdural, and depth EEG from patients with either temporal or frontal lobe epilepsy, we have shown that the STL_max profiles at brain sites systematically converge to similar values tens of minutes before a seizure. We have called these brain sites “critical sites” and their convergence “entrainment” or “synchronization” of the dynamics. In summary, the epileptic brain appears to be progressively entrained by the focal sites, leading to loss of relative independence of normal brain sites in processing of information long before a seizure develops (e.g., see Figure 1(a), (b), (c), left panels).

Figure 1.

Resetting of brain dynamics at seizures / events. Left panels: Epileptic seizure #3 from patient E5. Right panels: Event #1 from patient P5. (a) Scalp EEG recordings 1 minute pre-ictally (event at t=0). (b) Smoothed salp EEG recordings 1 hour pre-ictally (event at t=0). (c) Smoothed STL_max profiles from a pair of electrodes over the same time interval as in (b). (d) T-index profiles from the pair of electrodes in (c) and an additional pre-ictally entrained pair. The horizontal dotted line is the T_th for entrainment. First, we notice a longer entrainment period (about 20 minutes) prior to the epileptic seizure than the one (a couple of minutes) prior to PNES. Second, according to the rules for resetting (see text), for the epileptic seizure, resetting of both pairs of electrodes is observed in its postictal period resulting to a high number of resetting pairs (low SRP value; from Table 2, SRP=0.008). In PNES, none of the shown pairs resets and this explains the low number of resetting pairs at this event (high SRP value – see Table 2, SRP=0.389). While the F4-P3 pair disentrains within approximately 10 minutes postictally, thus barely satisfying the condition for disentrainment, its individual electrodes do not exhibit a significant change in their STL_max profiles postictally versus pre-ictally (e.g., see P3 or F4 across the event in the right panel of (c)). The F4-F10 pair did not even satisfy the disentrainment conditions as it becomes disentrained about 20 minutes (i.e., much later than 10 minutes) after this event's end.

A statistical measure of entrainment between two brain sites i and j, with respect to a measure of their dynamics (e.g., STL_max), has been developed in the past [7]. Specifically, the T_ij between measures at electrode sites i and j and at time t is defined as

$$T_{\mathit{\text{ij}}}\left(t\right)=\sqrt{m}\frac{\left|{\widehat{\text{D}}}_{\text{ij}}\left(t\right)\right|}{{\widehat{\sigma }}_{\mathit{\text{ij}}}\left(t\right)}$$ (1)

where ‖ is the absolute value, and D̂_ij(t) and σ̂_ij(t) denote the sample mean and standard deviation respectively of all the m differences between the STL_max values (one STL_max value is produced per 10.24 sec EEG segment) at electrodes i and j, within a moving window w(t) = [t, t−m*10.24 sec] over the available EEG recording (e.g., see Figure 1(d))

We have defined disentrainment (dynamical desynchronization) between electrode sites i and j when T_ij(t) is significantly different from zero at a significance level α. The disentrainment condition between the electrode sites i and j, as detected by the paired t-test, is T_ij(t)>t_a/2m−1, where t_a/2,m−1 is the 100×(1−a/2)% critical value of the t-distribution with m-1 degrees of freedom. If T_ij(t)<t_a/2,m−1 (which means that we do not have satisfactory statistical evidence at the α level that the differences of values of a measure between electrode sites i and j within the time window w(t) are not zero), we consider that sites i and j are entrained with respect to STL_max at time t. Using α = 0.01 and m = 60, that is, using a window w(t) of 10 minutes in duration, the threshold T_th = t_a/2,m−1 = 2.662. It should be noted that similar STL_max values does not mean that the sites interact. However, progressive convergence over time of STL_max points to a diminishing probability that the sites are unrelated [6, 53].

In accordance to the T-index between brain sites at time t we define its equivalent for a single site with respect to the evolution of its dynamics between two consecutive time windows. The single channel T-index T_i(t) is given by

$$T_i\left(t\right)=\sqrt{\frac{\text{m}}{2}}\frac{\left|{\widehat{L}}_i\left(t\right)-{\widehat{L}}_i\left(t+h+m\right)\right|}{{\widehat{\text{σ}}}_i^p\left(t\right)}$$ (2)

where L̂_i(t) is the sample mean of the STL_max values in the moving window w(t) of length m, ${\widehat{\text{σ}}}_i^p\left(t\right)$ the pooled standard deviation (average of standard deviations) of the STL_max values in the windows w(t) and w(t+h+m), and h a buffer separating these two windows which we took equal to the maximum duration of a patient's recorded seizures.

2.3. Dynamical Resetting

We have shown that the observed spatial entrainment of dynamics at critical brain sites in the pre-ictal period is changed to disentrainment in the postictal period of epileptic seizures [7-9]. We have named this phenomenon “dynamical resetting”. Those studies on intracranial EEG from epileptic patients also demonstrated that dynamical resetting is quite specific to epileptic seizures; that is, dynamical resetting occurs with a significantly higher probability following epileptic seizures compared to randomly selected time points in interictal periods (statistical significance level α = 0.05). Epileptic seizures reset the excessive pathological entrainment occurring minutes prior to their onset and appear to play a homeostatic role of restoring the balance between synchronization and desynchronization of brain dynamics [9].

One way to quantify dynamical resetting at seizures is via the number of pre-ictally entrained pairs of sites that reset in the immediate postictal period. The pairs of sites (i, j), whose T_ij(t) values is below T_th for w(t) immediately prior to seizure onset are selected as entrained pairs for that seizure. (Note: the results we report herein do not change significantly, and the final conclusions remain the same, if we select as entrained pairs the ones that remain entrained for every t within the whole interval of 10 minutes before a seizure onset.) By dividing this number of pairs of sites by the total number of available pairs of recording sites, we define the entrainment power EP(t) at time t as

$$\mathit{\text{EP}}\left(t\right)=\frac{2}{N_e\left(N_e-1\right)}\sum _{i=1}^{N_e-1}\sum _{j=i+1}^{N_e}\text{Θ}\left(T_{ij}\left(t\right)<T_{th}\right)$$ (3)

Out of those pairs of sites, a subset of pairs will get disentrained after the seizure (T(t')_ij>T_th for some time t' in the immediate after the seizure's end time interval). Moreover, only a portion of this subset of disentrained pairs of sites (i, j) will also have their corresponding individual STL_max values significantly change with respect to the values they had during the pre-ictal entrainment (T_i>T_th and T_j>T_th). Figure 1(d) illustrates the application of these resetting conditions at a typical seizure in an ES patient and an event in a PNES patient.

Considering the above conditions for resetting, and by dividing this number of pairs of sites by the total number of available pairs of recording sites, we can define the resetting power RP(t) at time t mathematically as

$$\mathit{\text{RP}}\left(t\right)=\frac{2}{N_e\left(N_e-1\right)}\sum _{i=1}^{N_e-1}\sum _{j=i+1}^{N_e}\text{Θ}\left(T_{ij}\left(t\right)<T_{th}\right)\cdot \text{Θ}\left(T_i\left(t\right)>T_{th},T_j\left(t\right)>T_{th}\right)\cdot \text{Θ}\left(\exists t^{\prime }\in \left[t,t+h+m\right]:T_{ij}\left(t^{\prime }\right)>T_{th}\right)$$ (4)

where N_e is the total number of available electrode sites, and Θ is the Heaviside function such that Θ(A) = 1 if A is true and Θ(A)=0 if A is false.

We apply the same methodology to check for resetting of dynamics at non-seizure points too. Actually, we applied this methodology for every single available time point t on the STL_max profiles per recording site, that is, for every 10.24 sec on the available EEG from every electrode and patient. For consistency in statistics, we have kept the same values for m and h for the analysis at non-seizure points too, that is, m=60 points (10 minutes) and h=maximum duration of seizures per patient. As a result, we end up with a value of our measures of entrainment (EP) and resetting (RP) for every 10.24 sec per EEG recording.

For a specific time point t₀, in a recording of length N_rec, the entrainment power is EP(t₀) and the resetting power is RP(t₀). We can then define the EP score and the RP score as

$$\mathit{\text{SEP}}\left(t_0\right)=\frac{1}{N_{\mathit{\text{rec}}}}\sum _{t=1}^{N_{\mathit{\text{rec}}}}\text{Θ}\left(\mathit{\text{EP}}\left(t\right)>\mathit{\text{EP}}\left(t_0\right)\right)\text{and}\mathit{\text{SRP}}\left(t_0\right)=\frac{1}{N_{\mathit{\text{rec}}}}\sum _{t=1}^{N_{\mathit{\text{rec}}}}\text{Θ}\left(\mathit{\text{RP}}\left(t\right)>\mathit{\text{RP}}\left(t_0\right)\right)$$ (5)

These scores quantify how unlikely it is to observe larger entrainment and resetting power at a time t anywhere in the entire EEG recording than those at time t ₀. SEP and SRP are simple counting metrics on the sets of EP and RP values respectively. Therefore, these scores are independent measures of the overall entrainment and resetting power over an EEG recording. In this sense, the SEP and SRP values in different recordings (patients) can be directly comparable. For both SEP and SRP scores, small values indicate a rare event of high entrainment or resetting respectively.

From equation (5), the SEP and SRP values at seizure points can be perceived as the p-values of testing if the respective entrainment (EP) and resetting (RP) at a seizure are significant different than in the interictal period. Therefore, by using Fisher's method for combining p-values [54] we can obtain the statistical significance of overall entrainment and resetting at seizures per patient. Given k, with k equal to a patient's number of seizures, p-values p_i, the quantity

$$X^2=-2\text{∑}_{i}log\left(p_i\right)$$ (6)

follows a χ² -distribution with 2k degrees of freedom. Substituting for p_i the SEP and SRP values at seizure points t_s and comparing the obtained X² with the appropriate χ² -distribution we can estimate the combined p-value for each patient's seizures. A small combined p-value indicates significant entrainment or resetting at seizures for each patient.

3. Results

3.1. Data analyzed

Long-term (days) scalp EEG recordings from six patients diagnosed with psychogenic non-epileptic seizures and five patients with epileptic seizures were analyzed. The patients' clinical characteristics and the findings from video-EEG monitoring are tabulated in Tables 1 and 2 respectively. EEG signals were recorded from twenty-nine channels overlaying 6 brain regions using a standard EEG montage (extended international 10-20 system) including auricular references. The analog data were low-pass filtered at 70 Hz and then digitized at 200 Hz (sampling frequency) and stored on a digital hard drive in Nihon-Kohden data format. No other (digital) filters were applied to the EEG data before the subsequent dynamical analysis. This analysis was conducted continuously (without subjective or objective rejection of any EEG data segment from the dynamical analysis for any reason, even when artifacts are present and irrespectively of the vigilance state of the patient) and sequentially for non-overlapping 10.24 sec running windows over the entire scalp EEG available over days per patient.

Table 1. Patient Clinical Characteristics.

Patient	Age	Sex	Time since First Event	Event Frequency	Physical Exam	MRI	Neuro-Psych Testing
E1	35	F	33 years	weekly	wnl	R MTS	Mild bitemporal dysfunction
E2	69	M	1 year	monthly	wnl	SVID	Mild cognitive inefficiency, depression
E3	54	F	14 years	weekly	wnl	R frontal polymicrogyria	Mild cognitive inefficiency
E4	45	F	43 years	weekly	wnl	L MTS	Dominant temporal dysfunction, depression
E5	18	F	8 years	daily	wnl	wnl
P1	23	F	1 year	weekly	wnl	wnl	Conversion, PTSD
P2	39	F	1 year	weekly	giveway weakness, atypical vision loss	wnl	Conversion
P3	49	M	2 years	weekly	wnl	wnl	Depression
P4	22	F	1 year	weekly	Left sensory loss that splits the midline	wnl	Mild cognitive inefficiency, invalid personality profile
P5	64	M	7 years	weekly	wnl	SVID	Somatization
P6	39	F	7 years	daily	wnl	wnl	Conversion

Abbreviations: MRI = Magnetic Resonance Imaging, Neuropsych = Comprehensive neuropsychological testing with personality inventory, F = female, M= male, wnl= within normal limits; R=right, L=left, MTS = mesial temporal sclerosis, SVID=small vessel ischemic disease; PTSD = post-traumatic stress disorder

Table 2. Findings from Video-EEG monitoring.

Patient	Clinical Event	# Events	Average Event Duration (sec)	EEG Recording Duration (hours)	Interictal EEG	Ictal EEG	Clinical Diagnosis
E1	Fear, unresponsive, L head deviation, +/− secondary generalization	5	84	52	RFT spikes	RT	RT epilepsy
E2	Unresponsive, oral automatisms	2	38	69.5	LT sharps + slowing	LT	LT epilepsy
E3	Arousal, asymmetric tonic posturing	2	90	142.5	RFC sharps	RF	RFC epilepsy
E4	Expressive aphasia, oral automatisms, R head deviation, +/− secondary generalization	5	94	64	LT sharps + slowing	LT	LT epilepsy
E5	Arousal, asymmetric tonic posturing with L arm extension	5	52	48	wnl	Non local ized	Epilepsy, non-localized
P1	Irregular L arm/leg jerks, whispering	4	100	106.5	wnl	wnl	PNES
P2	Irregular B arm jerks, head nodding, unresponsive	3	280	52	wnl	wnl	PNES
P3	Variable responsiveness, slowed speech, tearful, subsequently amnestic for event	2	480	70.5	wnl	wnl	PNES
P4	Dizzy, unresponsive verbally but follows motor commands, body jerks	3	1200	24	wnl	wnl	PNES
P5	Irregular R arm jerks, eye flutter, stutters	2	380	28.5	wnl	wnl	PNES
P6	Staring, unresponsive	5	25	72	wnl	wnl	PNES

Abbreviations: R= right, L= left, F= frontal, T= temporal, C= central, wnl= within normal limits; PNES = psychogenic non-epileptic seizures

3.2. Dynamical Resetting following ES and PNES

We estimated the EP and RP measures over the entire available EEG recording from each patient, and subsequently the SEP and SRP scores for each seizure / event. As an example, we demonstrate the distributions of EP and RP for one patient with epilepsy (E1) and one with PNES (P5) in Figure 2. The entrainment power (EP) observed at seizures (indicated by black vertical lines) in either patient is not statistically different from the one observed interictally. This is not the case for the resetting power (RP). The RP values for the epileptic patient's seizures fall into the higher end of the resetting distribution, while for PNES patient P5 they do not. That was a typical trend across patients from the two groups that is statistically quantified by the SEP and SRP measures.

Figure 2.

Distribution for the measures of (a) entrainment power (EP) and (b) resetting power (RP) estimated every 10.24 sec over the entire EEG recordings from patient E1 (Left panels) and patient P5 (Right panels). The black vertical lines indicate the position of the EP and RP values at seizure/events per patient (5 seizures for E1, 2 events for P5). For both patients, the power of entrainment (EP) values at events fall relative close to the center of the corresponding distribution rendering them statistically non-significant, while the power of resetting (RP) values at seizures for patient E1 are located near the tail of the distribution rendering them statistically significant.

In Table 3, we present the SEP and SRP values for all events (epileptic seizures and PNES) per patient. The SRP values for patient E1 are very low (0.022 to 0.065) indicating seizures with high resetting power. Patient P5 in contrast has high SRP values (0.389 and 0.536) indicating that the resetting observed at his PNES events is not uncommon with respect to other points in the interictal period. All epileptic seizures, except the ones of patient E3, exhibited significantly low SRP values. Even though the finding for patient E3 was an outlier in our analysis of dynamical resetting at epileptic seizures, this patient's medical history was consistent with short but frequent episodes of seizure clusters/status epilepticus, an indication of serious pathology of brain's dynamics where individual epileptic seizures (like the two ones we herein analyzed from this patient) cannot reset the dynamics [2, 55]. All PNES patients exhibited non-significant resetting at events (significantly large SRP values).

Table 3. SEP and SRP values for ES and PNES events.

Patients with ES (Seizure No.)	SEP -ES-	SRP -ES-	Patients with PNES (Seizure No.)	SEP -PNES-	SRP -PNES-
E1(1)	0.370	0.044	P1(1)	0.810	0.585
E1(2)	0.342	0.058	P1(2)	0.736	0.193
E1(3)	0.361	0.065	P1(3)	0.909	0.185
E1(4)	0.334	0.056	P1(4)	0.414	0.490
E1(5)	0.137	0.022	P2(1)	0.890	0.200
E2(1)	0.700	0.132	P2(2)	0.147	0.224
E2(2)	0.661	0.596	P2(3)	0.206	0.525
E3(1)	0.043	0.069	P3(1)	0.749	0.207
E3(2)	0.358	0.054	P3(2)	0.913	0.128
E4(1)	0.048	0.118	P4(1)	0.981	0.303
E4(2)	0.395	0.106	P4(2)	0.521	0.454
E4(3)	0.109	0.081	P4(3)	0.499	0.407
E4(4)	0.113	0.096	P5(1)	0.477	0.389
E4(5)	0.174	0.126	P5(2)	0.209	0.536
E5(1)	0.046	0.058	P6(1)	0.182	0.665
E5(2)	0.179	0.223	P6(2)	0.484	0.174
E5(3)	0.139	0.008	P6(3)	0.440	0.786
E5(4)	0.117	0.280	P6(4)	0.693	0.195
E5(5)	0.008	0.006	P6(5)	0.949	0.786

Furthermore, PNES patients exhibited less entrainment and resetting at events than the ones by ES patients. In Figure 3 we show the boxplots for SEP and SRP values at events, taken from Table 2, for the two groups of patients, along with the p-value of the performed Students t-test for equality of means between the two groups. Since these p-values are less than 0.001, the two groups of epileptic seizures and PNES are statistically different with respect to either entrainment or resetting. For the group of PNES patients, the mean values of SEP and SRP at events is near 0.5 indicating a random entrainment and resetting at events, while for the group of epileptic patients the means are quite lower, indicating high entrainment and resetting at seizures.

Figure 3.

Boxplots (box-and-whisker diagrams) for the 19 events from each of the two groups of ES and PNES patients based on the values of the score of (a) entrainment power (SEP), (b) resetting power (SRP) measures from Table 2. Each box represents the interquartile range of the mean of each measure across events, the bar within the box represents the median of the measures, and the bottom and top whiskers represent the 1.5 times below and above the lower and upper quartiles respectively. The corresponding p-value for the t-tests of equality of means between the two groups are very low (p<0.001) for either measure, with resetting having an edge over entrainment, indicating significant difference between the two groups of patients in terms of entrainment and resetting with respect to their respective interictal values. These results appear to be promising for the use of SEP and SRP measures in differential diagnosis between ES and PNES.

Finally, in Table 4, the combined p-value for each patient's events is shown. According to the combined p-values, entrainment is significant (p<0.05) for two, while resetting is significant for four out of the 5 epileptic patients. For patients with PNES, both resetting and entrainment are not significant. It is noteworthy that while p-values (i.e., SEPs or SRPs) of some individual events may not be low enough to be statistically significant, combined through Fisher's test to form the combined p-value per patient may reach statistical significance levels. That was true for patients with ES who typically had low individual p-values (∼ 0.1 to 0.2) at their seizures.

Table 4. Statistical significance of entrainment and resetting per patient via Fisher's combined p-test.

Patient	Entrainment (combined p-value)	Resetting (combined p-value)
E1	0.305	0.001
E2	0.832	0.281
E3	0.083	0.025
E4	0.028	0.012
E5	0.002	0.001
P1	0.942	0.328
P2	0.340	0.322
P3	0.952	0.112
P4	0.923	0.481
P5	0.343	0.570
P6	0.675	0.563

4. Conclusions

This study shows that analysis of brain dynamics from recorded long-term scalp EEG in patients with ES and PNES constitutes a valuable route for investigations in the genesis of ES and PNES events, and the differentiation between epileptic seizures and non-epileptic seizures of psychogenic origin. The initial basis for the produced results was the previously observed pathological gradual convergence of the rates of generation of information between critical brain sites in the order of tens of minutes before an epileptic seizure (dynamical entrainment), and their quick divergence (dynamical disentraiment) after seizure's end, thus implying a functional role of epileptic seizures for the recovery from the developed pathology of brain's dynamics. Combining both observations, we have called dynamical resetting the phenomenon of the pre-ictal entrainment and postictal disentrainment.

We herein developed new measures of entrainment (entrainment power – EP) and resetting (resetting power – RP) of brain dynamics, at any time point in the challenging, but clinically practical, condition of EEG recorded from patients with scalp electrodes. We then developed statistical measures of this entrainment (SEP) and resetting (SRP) at ES or PNES events per patient with respect to his/her interictal EP and RP values. On the basis of the SEP and SRP values, we have shown at a high level of statistical significance that: a) PNES patients exhibit less level of dynamical entrainment and resetting at events than ES patients at seizures (see Figure 3; p<0.001), and b) Events in PNES patients do not reset brain's dynamics (see Table 3; combined p-values per patient > 0.1), while they typically do (4 out of 5 patients) in ES (combined p-values < 0.05). These results, combined with prior ones from intracranial EEG in ES by Iasemidis et al. [6-9], show quantitatively that epileptic seizures typically reset the brain's dynamics. They also show that PNES events typically do not.

The results on ES and PNES from this study suggest that the proposed methodology of measuring the resetting of brain dynamics could be useful for differentiation between PNES and epileptic seizures. Future studies with a larger number of PNES and ES patients, as well as PNES patients with concomitant epilepsy, are contemplated to further investigate the sensitivity and specificity of our results and the possibility for the development of a robust diagnostic tool for PNES. In addition to shedding light on the mechanisms of generation of ES and PNES, such analytical tools could drastically reduce years of diagnostic delays, improve very early the quality of life of patients with PNES, and reduce the associated health care costs.