- Email: [email protected]
Power spectral analysis in infants with seizures: Relationship to development
Epilepsy & Behavior 20 (2011) 700–705
Contents lists available at ScienceDirect
Epilepsy & Behavior j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / ye b e h
Power spectral analysis in infants with seizures: Relationship to development Kandan Kulandaivel, Gregory L. Holmes ⁎ Department of Neurology, Neuroscience Center at Dartmouth, Dartmouth Medical School, Hanover, NH, USA
a r t i c l e
i n f o
Article history: Received 12 January 2011 Revised 12 February 2011 Accepted 15 February 2011 Keywords: Spectral analysis Electroencephalogram Seizures Epilepsy Epileptic encephalopathy Delta Theta Alpha Beta
a b s t r a c t There is increasing evidence that there is a strong relationship between brain oscillations and neurocognitive function. We used EEG power spectral analysis to determine if frequency and power provide an independent measure of developmental impairment in infants. We examined the spectral power of EEGs in 200 infants between 6 and 24 months of age who were evaluated for seizures. Infants were stratified into three age groups 6–12, 12–18, and 18–24 months, and development assessments were coded as normal, moderately delayed, and severely delayed. Compared with the normal infants, children with developmental delay had lower mean frequencies and greater delta and less theta and alpha power. Delta/theta and theta/alpha ratios were highly significant indicators of developmental status. This study demonstrates that frequency and power of brain oscillations during wakefulness is a strong predictor of development in infants. The findings support the concept that normal oscillatory activity is critical for normal cognitive function during development. © 2011 Elsevier Inc. All rights reserved.
1. Introduction During the first months of life children are at particularly high risk for seizures, with the largest number of new-onset seizure disorders occurring during this time [1]. The early onset of seizures is a high risk factor for subsequent cognitive impairment [2–6]. These cognitive deficits may be severe enough to cause difficulties in school. Thus, identifying young children with seizures who are at risk for developmental delay could lead to early therapeutic intervention. Oscillations in brain structures provide temporal windows that bind coherently supporting neuronal assemblies in the representation, processing, storage, and retrieval of information. There are data from studies in both humans and animals indicating that aberrant brain rhythms are associated with neurocognitive impairment. The EEG, a readily available recording usually obtained in children following suspected seizures, not only provides information about epileptiform activity, but also captures activity of groups of oscillation generators producing rhythmic activity in several frequency ranges. Thus, the EEG allows an assessment of brain rhythms. Spectral analysis is a representation of the signal's amplitude as a function of frequency. Power spectral analysis (PSA) is one of the standard methods used for quantification of the EEG. The PSA can be
⁎ Corresponding author at: Department of Neurology, Neuroscience Center at Dartmouth, Dartmouth Medical School, Lebanon, NH 03756, USA. Fax: + 1 603 650 7617. E-mail address: [email protected] (G.L. Holmes). 1525-5050/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.yebeh.2011.02.021
used to accurately measure the frequency content and distribution of power over these frequencies. Despite the high utilization of EEGs in children with seizures, few studies have examined the relationship between brain rhythms, as measured by routine EEGs, and development in children with earlylife seizures. In this study we hypothesized that PSA in children with seizures would distinguish between normal infants and infants with developmental impairment. 2. Methods The study population consisted of 200 consecutive children between the ages of 6 and 24 months referred to the outpatient Clinical Neurophysiology Laboratory at Dartmouth–Hitchcock Medical Center for known epilepsy or suspected seizures. Because of the discontinuous features of neonatal EEGs during quiet sleep, we assessed EEGs only in older infants [7,8]. EEGs and developmental assessments were analyzed retrospectively. The study was approved by the institutional review board of Dartmouth College. Digital EEGs were recorded during the fully awake, drowsy, and sleep states using the 10–20 system of electrode placement. An electroencephalographer (K.K.) blinded to the developmental assessment evaluated two 5-second epochs of artifact-free EEGs during both the awake and sleep states for spectral analysis. Wakefulness was assessed by evaluated eye blinks and technologist observations. If there was concern about possible drowsiness, the epoch was not evaluated. The sleep epochs were obtained during stage 2 of sleep at a time when sleep spindles and vertex sharp waves were present. EEG
K. Kulandaivel, G.L. Holmes / Epilepsy & Behavior 20 (2011) 700–705
segments were taken at a time the child was not having epileptiform activity. Power spectral analysis was conducted with Twin EEG software (Grass Telefactor) using an average reference montage. Frequencies from 0 to 25 Hz were analyzed using a fast Fourier transform (FTT) with Hamming windowing, a sampling rate of 200 Hz, and epoch size of 256 Hz with a frequency resolution 0.781 Hz. PSA for all channels combined (referred to as total PSA), as well as the means of O1 and O2, were obtained. Mean, median, edge, and peak frequencies were calculated, and percentages of total power in the delta (0 to b4 Hz), theta (4 to 7 Hz), alpha (8 to 12 Hz), and beta (13 to 25 Hz) bandwidths were calculated. Because of possible age-related changes in frequency and power of oscillations, the infants were arbitrarily stratified into three 6-month age groups: 6–12 months (n = 86), 12– 18 months (n = 65), and 18–24 months (n = 49). On the basis of review of the medical records by an author (G.L.H.) blinded to EEG data, developmental assessments were classified as normal, moderate, and profound (global delays). Assessments were made using the Denver Development Screening Test, a method for screening cognitive and motor performance in preschool children. Tasks are grouped into four categories (social contact, fine motor skills, language, and gross motor skills). Children with deficits in only one domain were classified as having moderate abnormalities, whereas children with deficits in more than one domain were classified as having profound delays. More than 90% of all the infants, including 100% of those with moderate or profound delays, were evaluated by a pediatric neurologist at Dartmouth–Hitchcock Medical Center. Examples of the PSA in the three groups are provided in a montage in Fig. 1.
701
2.1. Statistical analysis All data was assessed to determine if the data came from a gaussian distribution using the D'Agostino and Pearson omnibus normality test. Although the majority of the data had a gaussian distribution, because of unequal group size we evaluated the data using nonparametric statistics. Mean, peak, and median frequencies and relative percentages of power accounted for by delta, theta, alpha, and beta frequencies were compared between the three groups (normal, moderate delay, and severe delay) using the Kruskal–Wallis test (KWT) with Dunn's multiple comparisons to compare groups. The delta/theta and theta/alpha ratios were calculated for each age and developmental group and compared using the KWT. Correlation coefficients using Pearson's r were performed between age (in months) and mean, peak, and median frequencies and percentages of power accounted for by delta, theta, alpha, and beta frequencies. Fisher's exact test was used to compare proportions. Data are expressed as means ± SE. A P valueb 0.05 was considered statistically significant. 3. Results All 200 infants had evaluable EEG records and sufficient medical records to assess PSA and development. Spontaneous sleep was obtained in 132 of 200 (66%). None of the children had infantile spasms or hypsarrhythmia on the EEG. There were minimal developmental changes across the three age groups, with only the percentage of total power in the alpha bandwidth showing an increase in the age group N18 to 24 months (Fig. 2). Although there were no significant correlations between mean, peak, and median frequencies and delta, theta, and beta
Fig. 1. EEG power spectra of infants (A) 6 to 12, (B) N12 to 18, and (C) N18 to 24 months of age. From left to right are examples from infants with normal, moderate, and severe developmental impairments. Representative epochs of the EEG analyzed are provided on the left, and PSA for each channel of the EEG on the right. The ordinate shows power in each channel, and the abscissa shows frequencies from 0 to 25 Hz. Note that in the moderate and severely impaired children, the primary power is in the delta bandwidth.
702
K. Kulandaivel, G.L. Holmes / Epilepsy & Behavior 20 (2011) 700–705
Fig. 2. Developmental changes in the EEGs of infants from 6 to 24 months of age. (A) Mean and median frequency as a function of age. No significant differences across ages were seen. (B) Percentage of total power for delta, theta, alpha, and beta across age groups. There was a significant increase in alpha percentage with age.
frequencies, there a significant correlation between age and alpha power (Pearson's r = 0.173, P = 0.007). Significant group differences in total power spectra were noted in the 6- to 12-month-olds (n = 67) among children with moderate (n = 5) or severe (n = 13) developmental delay having lower mean (KWT= 16.28, P b 0.001) and median (KWT = 8.65, P = 0.013) frequencies and greater delta (KWT = 14.40) and lower theta (KWT = 16.54, P b 0.001) and beta (KWT= 8.35, P = 0.015) percentages compared with the normal group (n = 49) (Fig. 3A). When O1/O2 power spectra of the
three groups were compared, children with moderate or severe developmental delay had lower mean (KWT = 11.30, P = 0.004), peak (KWT = 8.45, P = 0.015), and median (KWT = 9.02, P = 0.010) frequencies and increased delta (KWT = 23.84, P b 0.001) and decreased theta (KWT = 12.17, P = 0.002) percentages (Fig. 3B). Delta/theta and theta/ alpha ratios of the three groups were compared. The total PSA delta/ theta ratio was higher (KWT= 17.74. P b 0.001) and the theta/alpha lower (KWT = 8.15, P = 0.017) in the children with the developmental disorders than in the normally developing infants (Figs. 4A and B).
Fig. 3. EEG power spectra of infants 6 to 12 months of age. (A) PSA of all channels. (B) PSA of combined O1 and O2. Mean, peak, and median frequencies are on the left, and relative frequencies of delta, theta, alpha, and beta are on the right. Note the higher mean and median frequencies and lower delta and greater theta and alpha percentages in the normal infants compared with the infants with developmental delay.
K. Kulandaivel, G.L. Holmes / Epilepsy & Behavior 20 (2011) 700–705
703
Epileptiform discharges (interictal spikes, sharp waves, and spikeand-wave discharges) were also predictive of developmental status, with 3 of 136 (2.2%) children with normal development and 38 of 64 (59.4%) children with moderate or severe delays having interictal epileptiform discharges (P b 0.001). Antiepileptic drugs were prescribed to 5 of 136 (3.7%) children with normal development, 18 of 64 (29.6%) with moderate delays, and 10 of 37 (27.0%) children with profound delays. There was a strong relationship between antiepileptic drug administration and outcome (P b 0.001). During sleep no significant difference in mean, peak, or median frequency or alpha, delta, theta, or beta power in total or O1/O2 PSA was seen in any of the three age groups. 4. Discussion
Fig. 4. Comparison of delta/theta (A) and theta/alpha (B) ratios of the three age groups. Normal, moderate, and severe developmental delay is plotted as a function of age. (A) Delta/theta ratio. Compared with normal infants, both the moderately and severely impaired children had a higher delta/theta ratio at 6 to 12 months and N12 to 18 months. (B) Theta/alpha ratio. In the 6- to 12-month-olds, the higher proportion of theta in the normal infants than in infants with developmental abnormalities resulted in a higher theta/alpha ratio in the normal group than in the other two groups of infants. In the normal children, with increasing age the amount of theta decreases and the amount of alpha increases, resulting in a theta/alpha ratio lower than that of infants with developmental impairment at N 18 to 24 months.
In the N12- to 18-month-olds (n = 65), theta (KWT = 9.044, P = 0.011), alpha (KWT = 10.44, P = 0.005), and beta (KWT = 12.84, P = 0.002) percentages were significantly lower in the groups with moderate (n = 11) and severe (n = 8) developmental delay than in the normal group (n = 46) (Fig. 5A). When O1/O2 power spectra of the three groups were compared, children with moderate and severe developmental delay had a lower theta percentage (KWT = 13.51, P = 0.001) than normal infants (Fig. 5B). The delta/theta ratio was higher (KWT = 7.04, P = 0.030) in the children with developmental disorders than in the normally developing infants (Fig. 4A). In the N18- to 24-month-olds (n = 68), alpha (KWT = 17.16, P b 0.001) and beta (KWT = 6.553, P = 0.038) percentages were significantly lower in the children with moderate (n = 11) and severe (n = 16) developmental delay than in the controls (n = 41) (Fig. 6A). When O1/O2 power spectra of the three groups were compared, children with moderate and severe developmental delay had a lower mean frequency (KWT = 6.324, P = 0.042) and lower alpha (KWT = 6.73, P = 0.034) and beta (KWT = 6.76, P = 0.034) percentages (Fig. 5B). The theta/alpha ratio was higher (KWT = 7.63, P = 0.022) in the infants with developmental disorders than in the normal infants (Fig. 3B). The EEG PSA findings were predictive of developmental status. Infants with a delta/theta ratio N3 were more likely to have moderate or severe delay than be normal in both the 6- to 12-month-olds (Fisher's exact test, P b 0.001) and N12- to 18-month-olds (P = 0.011). In the N18- to 24-month-old group a theta/alpha ratio N10 was predictive of developmental delay (P = 0.004).
In this study we found that PSA of short epochs of waking EEG in children varying in age from 6 to 24 months differentiated children with developmental delay from apparently normal infants. In general, children with developmental concerns had lower mean frequencies and greater percentages of the total power of the EEG occupied by delta activity than normally developing children. The nature of differences between the normal children and children with developmental issues varied with age. In the 6- to 12-month-olds, mean and median frequencies were lower and proportions of delta greater and of theta and beta lower in the developmentally impaired groups. In the N12- to18-month-olds, theta, alpha, and beta percentages were lower than in normal children. In the N18- to 24-month-olds, the alpha percentage was significantly lower in the groups with moderate and severe developmental delay than in the controls. Abnormalities of background frequencies were strong predictors of developmental abnormalities and suggest that PSA can serve as a useful, independent marker for children at risk for adverse neurocognitive impairment. There are now considerable data showing that brain rhythms have a major influence on brain function. For example, in rodents, theta rhythm (4–10 Hz) acts as a “significance signal” and is critically involved in mnemonic function of the hippocampus [9,10]. Information arriving in the hippocampus with theta oscillations is stored in the hippocampus, whereas information arriving in the absence of normal theta activity is not encoded or not encoded with the same degree of precision as when theta is present [10–12]. Additionally, phase of theta is critical in learning and memory: tetanic stimulation in CA1 produces long-term potentiation (LTP) when administered at the peak of theta and long-term depression (LTD) when delivered at the trough [13]. Likewise, gamma oscillations (30–100 Hz) are critical in the processing or perceiving of sensory information [14,15], consciousness [16,17], and storage of immediate memories [18–20]. Based on the relationship between oscillatory activity and cognition, it is not surprising that in this study we found that children with developmental concerns had aberrant rhythms with reduced theta at 6–18 months and reduced alpha at N18 months. Slowing on the EEG is often a major feature of epilepsy occurring in children, particularly in the epileptic encephalopathies [21]. For example, the EEG in children with infantile spasms is often marked by high-voltage slowing of the EEG as well as epileptiform activity [22]. Our study, in which none of the children had infantile spasms, carries these observations in the epileptic encephalopathies further, suggesting that altered brain rhythms in children with epilepsy or suspected seizures is more widespread than may have been thought. The EEG in children between 6 and 24 months of age is dominated by delta and theta frequencies, making it a challenge to interpret an EEG as excessively slow. Of note, only four of the EEGs used in this study were interpreted as showing excessive slowing. PSA of the EEG in infants, at the time the EEG is normally dominated by delta and theta activity, appears to be a very useful technique in assessing infants at risk for developmental impairment, even in the absence of an epileptic encephalopathy pattern. Indeed, PSA has been used to detect patients
704
K. Kulandaivel, G.L. Holmes / Epilepsy & Behavior 20 (2011) 700–705
Fig. 5. EEG power spectra of infants N12 to 18 months of age. (A) PSA of all channels. (B) PSA of combined O1 and O2. In the total power panel (A), there was a significant difference in theta percentages in children with developmental delay, whereas in the O1/O2 panel (B), there were group differences in detla and theta percentages.
at risk for neurocognitive deficits including cognitive impairment in REM sleep behavior disorder [23], early Alzheimer's disease [24], diabetes mellitus [25], attention-deficit hyperactivity [26], antiepileptic drug-induced impairment [27,28], and abnormal reading skills [29,30]. We found PSA during wakefulness was useful in detecting group differences, whereas PSA during sleep did not differentiate normal children from those with developmental delay. Drowsiness and sleep are associated with an increase in slow frequencies on the EEG. Measuring the same degree of drowsiness and sleep in infants is difficult and it is possible that the stage of drowsiness and sleep was uniform in the groups. Although our study was confined to infants, it is likely that PSA at other ages would be useful in assessing neurocognitive function. For example, in a study of 95 children with epilepsy (mean age = 10.41 years) for whom both EEGs and neuropsychological batteries were performed, the presence of slow-wave activity, but not epileptiform activity, was related to memory impairment (P b 0.01) [31]. The authors suggested that when slowing on the EEG is seen, close monitoring of cognitive and academic functioning should be recommended. In the infants with aberrant PSA, a similar admonition seems warranted. In this narrow age group studied, we found that presence of epileptiform activity and antiepileptic drug use were also related to development. Whereas children with normal development rarely had interictal spikes or sharp waves, and may not have had seizures, more than 90% of children with severe developmental abnormalities had epileptiform discharges on their EEGs. Interictal epileptiform discharges as well as abnormal brain rhythms are therefore markers of developmental impairment. In addition, there was a strong relationship between antiepileptic drug use and developmental status.
Although the finding of PSA abnormalities in infants with epileptiform discharges on the EEG who are treated with antiepileptic drugs adds little to the evaluation of the individual patient, our findings suggest that PSA, independent of epileptiform activity and antiepileptic drug use, provides an additional powerful independent measure of developmental status at the time of the EEG. This study has a number of weaknesses that must be considered. This was a retrospective study with developmental assessment gleaned from the medical record. Ideally, all the infants in the study should be studied in a longitudinal fashion and undergo serial Bayley Developmental Assessments to confirm their developmental status. In addition, examination of PSA provides only one quantifiable aspect of the relationship between brain rhythms and development. There is a dynamic interaction between brain rhythms and communication among structures differentially modulated by this interaction. For example, coherence, a measure of the state in which two signals maintain a fixed phase relationship with each other linking various waveforms, plays an important role in how information is processed in the brain. Oscillations play a major role in the way information is being processed in the brain and different oscillatory states will serve different functions. This interaction of waveforms was not measured in this study. Finally, this study demonstrated only a relationship between PSA and developmental status. Whether the alterations in PSA are a cause of the development impairment or are a result of the etiology of the seizures remains unclear. In summary, we found that in children referred for an EEG because of a possible history of seizures there were age-related differences in PSA between infants with normal development and those with developmental impairment. A challenge to clinicians is whether interventional techniques can be used to reverse these abnormalities and improve developmental outcome.
K. Kulandaivel, G.L. Holmes / Epilepsy & Behavior 20 (2011) 700–705
705
Fig. 6. EEG power spectra of infants N18 to 24 months of age. (A) PSA of all channels. (B) PSA of combined O1 and O2. Alpha percentage was lower in the developmentally impaired groups than in the normal group (A). As shown in (B), in the O1/O2 groups, percentages of alpha and beta differed among the groups.
Acknowledgment This work was supported by National Institutes of Health Grant R01NS044295 (G.H). References [1] Hauser WA. Epidemiology of epilepsy in children. Neurosurg Clin N Am 1995;6: 419–29. [2] Glosser G, Cole LC, French JA, Saykin AJ, Sperling MR. Predictors of intellectual performance in adults with intractable temporal lobe epilepsy. J Int Neuropsychol Soc 1997;3:252–9. [3] Bulteau C, Jambaque I, Viguier D, Kieffer V, Dellatolas G, Dulac O. Epileptic syndromes, cognitive assessment and school placement: a study of 251 children. Dev Med Child Neurol 2000;42:319–27. [4] Hermann BP, Seidenberg M, Bell B. The neurodevelopmental impact of childhood onset temporal lobe epilepsy on brain structure and function and the risk of progressive cognitive effects. Prog Brain Res 2002;135:429–38. [5] Cormack F, Helen CJ, Isaacs E, et al. The development of intellectual abilities in pediatric temporal lobe epilepsy. Epilepsia 2007;48:201–4. [6] Farwell JR, Dodrill CB, Batzel LW. Neuropsychological abilities of children with epilepsy. Epilepsia 1985;26:395–400. [7] Holmes GL, Lombroso CT. Prognostic value of background patterns in the neonatal EEG. J Clin Neurophysiol 1993;10:323–52. [8] Lombroso CT, Holmes GL. Value of the EEG in neonatal seizures. J Epilepsy 1993;6: 39–70. [9] Senior TJ, Huxter JR, Allen K, O'Neill J, Csicsvari J. Gamma oscillatory firing reveals distinct populations of pyramidal cells in the CA1 region of the hippocampus. J Neurosci 2008;28:2274–86. [10] Vertes RP, Kocsis B. Brainstem-diencephalo-septohippocampal systems controlling the theta rhythm of the hippocampus. Neuroscience 1997;81:893–926. [11] Buzsaki G. Theta oscillations in the hippocampus. Neuron 2002;33:325–40. [12] Itskov V, Pastalkova E, Mizuseki K, Buzsaki G, Harris KD. Theta-mediated dynamics of spatial information in hippocampus. J Neurosci 2008;28:5959–64. [13] Hyman JM, Wyble BP, Goyal V, Rossi CA, Hasselmo ME. Stimulation in hippocampal region CA1 in behaving rats yields long-term potentiation when delivered to the peak of theta and long-term depression when delivered to the trough. J Neurosci 2003;23:11725–31. [14] Gray CM, Viana Di PG. Stimulus-dependent neuronal oscillations and local synchronization in striate cortex of the alert cat. J Neurosci 1997;17:3239–53.
[15] Gray CM. Synchronous oscillations in neuronal systems: mechanisms and functions. J Comput Neurosci 1994;1:11–38. [16] Vanderwolf CH. Are neocortical gamma waves related to consciousness? Brain Res 2000;855:217–24. [17] Llinas R, Ribary U. Coherent 40-Hz oscillation characterizes dream state in humans. Proc Natl Acad Sci USA 1993;90:2078–81. [18] Chrobak JJ, Buzsaki G. High-frequency oscillations in the output networks of the hippocampal–entorhinal axis of the freely behaving rat. J Neurosci 1996;16:3056–66. [19] Hasselmo ME, Wyble BP, Wallenstein GV. Encoding and retrieval of episodic memories: role of cholinergic and GABAergic modulation in the hippocampus. Hippocampus 1996;6:693–708. [20] Lisman J. The theta/gamma discrete phase code occuring during the hippocampal phase precession may be a more general brain coding scheme. Hippocampus 2005;15:913–22. [21] Holmes GL, Lenck-Santini PP. Role of interictal epileptiform abnormalities in cognitive impairment. Epilepsy Behav 2006;8:504–15. [22] Hrachovy RA, Frost JDJr, Kellaway P. Hypsarrhythmia: variations on the theme. Epilepsia 1984;25:317–25. [23] Iranzo A, Isetta V, Molinuevo JL, et al. Electroencephalographic slowing heralds mild cognitive impairment in idiopathic REM sleep behavior disorder. Sleep Med 2010;11:534–9. [24] Stevens A, Kircher T, Nickola M, Bartels M, Rosellen N, Wormstall H. Dynamic regulation of EEG power and coherence is lost early and globally in probable DAT. Eur Arch Psychiatry Clin Neurosci 2001;251:199–204. [25] Brismar T, Hyllienmark L, Ekberg K, Johansson BL. Loss of temporal lobe beta power in young adults with type 1 diabetes mellitus. Neuroreport 2002;13:2469–73. [26] Van Dongen-Boomsma M, Lansbergen MM, et al. Relation between resting EEG to cognitive performance and clinical symptoms in adults with attention-deficit/ hyperactivity disorder. Neurosci Lett 2010;469:102–6. [27] Chung SS, McEvoy LK, Smith ME, Gevins A, Meador K, Laxer KD. Task-related EEG and ERP changes without performance impairment following a single dose of phenytoin. Clin Neurophysiol 2002;113:806–14. [28] Marciani MG, Maschio MC, Spanedda F, Gigli GL, Bassetti MA, Bernardi G. Sodium valproate and mental processes in newly referred epileptic patients: a computerized EEG study. Neuropsychobiology 1995;31:210–5. [29] Ackerman PT, Dykman RA, Oglesby DM, Newton JE. EEG power spectra of dysphonetic and nondysphonetic poor readers. Brain Lang 1995;49:140–52. [30] Ackerman PT, Dykman RA, Oglesby DM, Newton JE. EEG power spectra of children with dyslexia, slow learners, and normally reading children with ADD during verbal processing. J Learn Disabil 1994;27:619–30. [31] Koop JI, Fastenau PS, Dunn DW, Austin JK. Neuropsychological correlates of electroencephalograms in children with epilepsy. Epilepsy Res 2005;64:49–62.
Contents lists available at ScienceDirect
Epilepsy & Behavior j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / ye b e h
Power spectral analysis in infants with seizures: Relationship to development Kandan Kulandaivel, Gregory L. Holmes ⁎ Department of Neurology, Neuroscience Center at Dartmouth, Dartmouth Medical School, Hanover, NH, USA
a r t i c l e
i n f o
Article history: Received 12 January 2011 Revised 12 February 2011 Accepted 15 February 2011 Keywords: Spectral analysis Electroencephalogram Seizures Epilepsy Epileptic encephalopathy Delta Theta Alpha Beta
a b s t r a c t There is increasing evidence that there is a strong relationship between brain oscillations and neurocognitive function. We used EEG power spectral analysis to determine if frequency and power provide an independent measure of developmental impairment in infants. We examined the spectral power of EEGs in 200 infants between 6 and 24 months of age who were evaluated for seizures. Infants were stratified into three age groups 6–12, 12–18, and 18–24 months, and development assessments were coded as normal, moderately delayed, and severely delayed. Compared with the normal infants, children with developmental delay had lower mean frequencies and greater delta and less theta and alpha power. Delta/theta and theta/alpha ratios were highly significant indicators of developmental status. This study demonstrates that frequency and power of brain oscillations during wakefulness is a strong predictor of development in infants. The findings support the concept that normal oscillatory activity is critical for normal cognitive function during development. © 2011 Elsevier Inc. All rights reserved.
1. Introduction During the first months of life children are at particularly high risk for seizures, with the largest number of new-onset seizure disorders occurring during this time [1]. The early onset of seizures is a high risk factor for subsequent cognitive impairment [2–6]. These cognitive deficits may be severe enough to cause difficulties in school. Thus, identifying young children with seizures who are at risk for developmental delay could lead to early therapeutic intervention. Oscillations in brain structures provide temporal windows that bind coherently supporting neuronal assemblies in the representation, processing, storage, and retrieval of information. There are data from studies in both humans and animals indicating that aberrant brain rhythms are associated with neurocognitive impairment. The EEG, a readily available recording usually obtained in children following suspected seizures, not only provides information about epileptiform activity, but also captures activity of groups of oscillation generators producing rhythmic activity in several frequency ranges. Thus, the EEG allows an assessment of brain rhythms. Spectral analysis is a representation of the signal's amplitude as a function of frequency. Power spectral analysis (PSA) is one of the standard methods used for quantification of the EEG. The PSA can be
⁎ Corresponding author at: Department of Neurology, Neuroscience Center at Dartmouth, Dartmouth Medical School, Lebanon, NH 03756, USA. Fax: + 1 603 650 7617. E-mail address: [email protected] (G.L. Holmes). 1525-5050/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.yebeh.2011.02.021
used to accurately measure the frequency content and distribution of power over these frequencies. Despite the high utilization of EEGs in children with seizures, few studies have examined the relationship between brain rhythms, as measured by routine EEGs, and development in children with earlylife seizures. In this study we hypothesized that PSA in children with seizures would distinguish between normal infants and infants with developmental impairment. 2. Methods The study population consisted of 200 consecutive children between the ages of 6 and 24 months referred to the outpatient Clinical Neurophysiology Laboratory at Dartmouth–Hitchcock Medical Center for known epilepsy or suspected seizures. Because of the discontinuous features of neonatal EEGs during quiet sleep, we assessed EEGs only in older infants [7,8]. EEGs and developmental assessments were analyzed retrospectively. The study was approved by the institutional review board of Dartmouth College. Digital EEGs were recorded during the fully awake, drowsy, and sleep states using the 10–20 system of electrode placement. An electroencephalographer (K.K.) blinded to the developmental assessment evaluated two 5-second epochs of artifact-free EEGs during both the awake and sleep states for spectral analysis. Wakefulness was assessed by evaluated eye blinks and technologist observations. If there was concern about possible drowsiness, the epoch was not evaluated. The sleep epochs were obtained during stage 2 of sleep at a time when sleep spindles and vertex sharp waves were present. EEG
K. Kulandaivel, G.L. Holmes / Epilepsy & Behavior 20 (2011) 700–705
segments were taken at a time the child was not having epileptiform activity. Power spectral analysis was conducted with Twin EEG software (Grass Telefactor) using an average reference montage. Frequencies from 0 to 25 Hz were analyzed using a fast Fourier transform (FTT) with Hamming windowing, a sampling rate of 200 Hz, and epoch size of 256 Hz with a frequency resolution 0.781 Hz. PSA for all channels combined (referred to as total PSA), as well as the means of O1 and O2, were obtained. Mean, median, edge, and peak frequencies were calculated, and percentages of total power in the delta (0 to b4 Hz), theta (4 to 7 Hz), alpha (8 to 12 Hz), and beta (13 to 25 Hz) bandwidths were calculated. Because of possible age-related changes in frequency and power of oscillations, the infants were arbitrarily stratified into three 6-month age groups: 6–12 months (n = 86), 12– 18 months (n = 65), and 18–24 months (n = 49). On the basis of review of the medical records by an author (G.L.H.) blinded to EEG data, developmental assessments were classified as normal, moderate, and profound (global delays). Assessments were made using the Denver Development Screening Test, a method for screening cognitive and motor performance in preschool children. Tasks are grouped into four categories (social contact, fine motor skills, language, and gross motor skills). Children with deficits in only one domain were classified as having moderate abnormalities, whereas children with deficits in more than one domain were classified as having profound delays. More than 90% of all the infants, including 100% of those with moderate or profound delays, were evaluated by a pediatric neurologist at Dartmouth–Hitchcock Medical Center. Examples of the PSA in the three groups are provided in a montage in Fig. 1.
701
2.1. Statistical analysis All data was assessed to determine if the data came from a gaussian distribution using the D'Agostino and Pearson omnibus normality test. Although the majority of the data had a gaussian distribution, because of unequal group size we evaluated the data using nonparametric statistics. Mean, peak, and median frequencies and relative percentages of power accounted for by delta, theta, alpha, and beta frequencies were compared between the three groups (normal, moderate delay, and severe delay) using the Kruskal–Wallis test (KWT) with Dunn's multiple comparisons to compare groups. The delta/theta and theta/alpha ratios were calculated for each age and developmental group and compared using the KWT. Correlation coefficients using Pearson's r were performed between age (in months) and mean, peak, and median frequencies and percentages of power accounted for by delta, theta, alpha, and beta frequencies. Fisher's exact test was used to compare proportions. Data are expressed as means ± SE. A P valueb 0.05 was considered statistically significant. 3. Results All 200 infants had evaluable EEG records and sufficient medical records to assess PSA and development. Spontaneous sleep was obtained in 132 of 200 (66%). None of the children had infantile spasms or hypsarrhythmia on the EEG. There were minimal developmental changes across the three age groups, with only the percentage of total power in the alpha bandwidth showing an increase in the age group N18 to 24 months (Fig. 2). Although there were no significant correlations between mean, peak, and median frequencies and delta, theta, and beta
Fig. 1. EEG power spectra of infants (A) 6 to 12, (B) N12 to 18, and (C) N18 to 24 months of age. From left to right are examples from infants with normal, moderate, and severe developmental impairments. Representative epochs of the EEG analyzed are provided on the left, and PSA for each channel of the EEG on the right. The ordinate shows power in each channel, and the abscissa shows frequencies from 0 to 25 Hz. Note that in the moderate and severely impaired children, the primary power is in the delta bandwidth.
702
K. Kulandaivel, G.L. Holmes / Epilepsy & Behavior 20 (2011) 700–705
Fig. 2. Developmental changes in the EEGs of infants from 6 to 24 months of age. (A) Mean and median frequency as a function of age. No significant differences across ages were seen. (B) Percentage of total power for delta, theta, alpha, and beta across age groups. There was a significant increase in alpha percentage with age.
frequencies, there a significant correlation between age and alpha power (Pearson's r = 0.173, P = 0.007). Significant group differences in total power spectra were noted in the 6- to 12-month-olds (n = 67) among children with moderate (n = 5) or severe (n = 13) developmental delay having lower mean (KWT= 16.28, P b 0.001) and median (KWT = 8.65, P = 0.013) frequencies and greater delta (KWT = 14.40) and lower theta (KWT = 16.54, P b 0.001) and beta (KWT= 8.35, P = 0.015) percentages compared with the normal group (n = 49) (Fig. 3A). When O1/O2 power spectra of the
three groups were compared, children with moderate or severe developmental delay had lower mean (KWT = 11.30, P = 0.004), peak (KWT = 8.45, P = 0.015), and median (KWT = 9.02, P = 0.010) frequencies and increased delta (KWT = 23.84, P b 0.001) and decreased theta (KWT = 12.17, P = 0.002) percentages (Fig. 3B). Delta/theta and theta/ alpha ratios of the three groups were compared. The total PSA delta/ theta ratio was higher (KWT= 17.74. P b 0.001) and the theta/alpha lower (KWT = 8.15, P = 0.017) in the children with the developmental disorders than in the normally developing infants (Figs. 4A and B).
Fig. 3. EEG power spectra of infants 6 to 12 months of age. (A) PSA of all channels. (B) PSA of combined O1 and O2. Mean, peak, and median frequencies are on the left, and relative frequencies of delta, theta, alpha, and beta are on the right. Note the higher mean and median frequencies and lower delta and greater theta and alpha percentages in the normal infants compared with the infants with developmental delay.
K. Kulandaivel, G.L. Holmes / Epilepsy & Behavior 20 (2011) 700–705
703
Epileptiform discharges (interictal spikes, sharp waves, and spikeand-wave discharges) were also predictive of developmental status, with 3 of 136 (2.2%) children with normal development and 38 of 64 (59.4%) children with moderate or severe delays having interictal epileptiform discharges (P b 0.001). Antiepileptic drugs were prescribed to 5 of 136 (3.7%) children with normal development, 18 of 64 (29.6%) with moderate delays, and 10 of 37 (27.0%) children with profound delays. There was a strong relationship between antiepileptic drug administration and outcome (P b 0.001). During sleep no significant difference in mean, peak, or median frequency or alpha, delta, theta, or beta power in total or O1/O2 PSA was seen in any of the three age groups. 4. Discussion
Fig. 4. Comparison of delta/theta (A) and theta/alpha (B) ratios of the three age groups. Normal, moderate, and severe developmental delay is plotted as a function of age. (A) Delta/theta ratio. Compared with normal infants, both the moderately and severely impaired children had a higher delta/theta ratio at 6 to 12 months and N12 to 18 months. (B) Theta/alpha ratio. In the 6- to 12-month-olds, the higher proportion of theta in the normal infants than in infants with developmental abnormalities resulted in a higher theta/alpha ratio in the normal group than in the other two groups of infants. In the normal children, with increasing age the amount of theta decreases and the amount of alpha increases, resulting in a theta/alpha ratio lower than that of infants with developmental impairment at N 18 to 24 months.
In the N12- to 18-month-olds (n = 65), theta (KWT = 9.044, P = 0.011), alpha (KWT = 10.44, P = 0.005), and beta (KWT = 12.84, P = 0.002) percentages were significantly lower in the groups with moderate (n = 11) and severe (n = 8) developmental delay than in the normal group (n = 46) (Fig. 5A). When O1/O2 power spectra of the three groups were compared, children with moderate and severe developmental delay had a lower theta percentage (KWT = 13.51, P = 0.001) than normal infants (Fig. 5B). The delta/theta ratio was higher (KWT = 7.04, P = 0.030) in the children with developmental disorders than in the normally developing infants (Fig. 4A). In the N18- to 24-month-olds (n = 68), alpha (KWT = 17.16, P b 0.001) and beta (KWT = 6.553, P = 0.038) percentages were significantly lower in the children with moderate (n = 11) and severe (n = 16) developmental delay than in the controls (n = 41) (Fig. 6A). When O1/O2 power spectra of the three groups were compared, children with moderate and severe developmental delay had a lower mean frequency (KWT = 6.324, P = 0.042) and lower alpha (KWT = 6.73, P = 0.034) and beta (KWT = 6.76, P = 0.034) percentages (Fig. 5B). The theta/alpha ratio was higher (KWT = 7.63, P = 0.022) in the infants with developmental disorders than in the normal infants (Fig. 3B). The EEG PSA findings were predictive of developmental status. Infants with a delta/theta ratio N3 were more likely to have moderate or severe delay than be normal in both the 6- to 12-month-olds (Fisher's exact test, P b 0.001) and N12- to 18-month-olds (P = 0.011). In the N18- to 24-month-old group a theta/alpha ratio N10 was predictive of developmental delay (P = 0.004).
In this study we found that PSA of short epochs of waking EEG in children varying in age from 6 to 24 months differentiated children with developmental delay from apparently normal infants. In general, children with developmental concerns had lower mean frequencies and greater percentages of the total power of the EEG occupied by delta activity than normally developing children. The nature of differences between the normal children and children with developmental issues varied with age. In the 6- to 12-month-olds, mean and median frequencies were lower and proportions of delta greater and of theta and beta lower in the developmentally impaired groups. In the N12- to18-month-olds, theta, alpha, and beta percentages were lower than in normal children. In the N18- to 24-month-olds, the alpha percentage was significantly lower in the groups with moderate and severe developmental delay than in the controls. Abnormalities of background frequencies were strong predictors of developmental abnormalities and suggest that PSA can serve as a useful, independent marker for children at risk for adverse neurocognitive impairment. There are now considerable data showing that brain rhythms have a major influence on brain function. For example, in rodents, theta rhythm (4–10 Hz) acts as a “significance signal” and is critically involved in mnemonic function of the hippocampus [9,10]. Information arriving in the hippocampus with theta oscillations is stored in the hippocampus, whereas information arriving in the absence of normal theta activity is not encoded or not encoded with the same degree of precision as when theta is present [10–12]. Additionally, phase of theta is critical in learning and memory: tetanic stimulation in CA1 produces long-term potentiation (LTP) when administered at the peak of theta and long-term depression (LTD) when delivered at the trough [13]. Likewise, gamma oscillations (30–100 Hz) are critical in the processing or perceiving of sensory information [14,15], consciousness [16,17], and storage of immediate memories [18–20]. Based on the relationship between oscillatory activity and cognition, it is not surprising that in this study we found that children with developmental concerns had aberrant rhythms with reduced theta at 6–18 months and reduced alpha at N18 months. Slowing on the EEG is often a major feature of epilepsy occurring in children, particularly in the epileptic encephalopathies [21]. For example, the EEG in children with infantile spasms is often marked by high-voltage slowing of the EEG as well as epileptiform activity [22]. Our study, in which none of the children had infantile spasms, carries these observations in the epileptic encephalopathies further, suggesting that altered brain rhythms in children with epilepsy or suspected seizures is more widespread than may have been thought. The EEG in children between 6 and 24 months of age is dominated by delta and theta frequencies, making it a challenge to interpret an EEG as excessively slow. Of note, only four of the EEGs used in this study were interpreted as showing excessive slowing. PSA of the EEG in infants, at the time the EEG is normally dominated by delta and theta activity, appears to be a very useful technique in assessing infants at risk for developmental impairment, even in the absence of an epileptic encephalopathy pattern. Indeed, PSA has been used to detect patients
704
K. Kulandaivel, G.L. Holmes / Epilepsy & Behavior 20 (2011) 700–705
Fig. 5. EEG power spectra of infants N12 to 18 months of age. (A) PSA of all channels. (B) PSA of combined O1 and O2. In the total power panel (A), there was a significant difference in theta percentages in children with developmental delay, whereas in the O1/O2 panel (B), there were group differences in detla and theta percentages.
at risk for neurocognitive deficits including cognitive impairment in REM sleep behavior disorder [23], early Alzheimer's disease [24], diabetes mellitus [25], attention-deficit hyperactivity [26], antiepileptic drug-induced impairment [27,28], and abnormal reading skills [29,30]. We found PSA during wakefulness was useful in detecting group differences, whereas PSA during sleep did not differentiate normal children from those with developmental delay. Drowsiness and sleep are associated with an increase in slow frequencies on the EEG. Measuring the same degree of drowsiness and sleep in infants is difficult and it is possible that the stage of drowsiness and sleep was uniform in the groups. Although our study was confined to infants, it is likely that PSA at other ages would be useful in assessing neurocognitive function. For example, in a study of 95 children with epilepsy (mean age = 10.41 years) for whom both EEGs and neuropsychological batteries were performed, the presence of slow-wave activity, but not epileptiform activity, was related to memory impairment (P b 0.01) [31]. The authors suggested that when slowing on the EEG is seen, close monitoring of cognitive and academic functioning should be recommended. In the infants with aberrant PSA, a similar admonition seems warranted. In this narrow age group studied, we found that presence of epileptiform activity and antiepileptic drug use were also related to development. Whereas children with normal development rarely had interictal spikes or sharp waves, and may not have had seizures, more than 90% of children with severe developmental abnormalities had epileptiform discharges on their EEGs. Interictal epileptiform discharges as well as abnormal brain rhythms are therefore markers of developmental impairment. In addition, there was a strong relationship between antiepileptic drug use and developmental status.
Although the finding of PSA abnormalities in infants with epileptiform discharges on the EEG who are treated with antiepileptic drugs adds little to the evaluation of the individual patient, our findings suggest that PSA, independent of epileptiform activity and antiepileptic drug use, provides an additional powerful independent measure of developmental status at the time of the EEG. This study has a number of weaknesses that must be considered. This was a retrospective study with developmental assessment gleaned from the medical record. Ideally, all the infants in the study should be studied in a longitudinal fashion and undergo serial Bayley Developmental Assessments to confirm their developmental status. In addition, examination of PSA provides only one quantifiable aspect of the relationship between brain rhythms and development. There is a dynamic interaction between brain rhythms and communication among structures differentially modulated by this interaction. For example, coherence, a measure of the state in which two signals maintain a fixed phase relationship with each other linking various waveforms, plays an important role in how information is processed in the brain. Oscillations play a major role in the way information is being processed in the brain and different oscillatory states will serve different functions. This interaction of waveforms was not measured in this study. Finally, this study demonstrated only a relationship between PSA and developmental status. Whether the alterations in PSA are a cause of the development impairment or are a result of the etiology of the seizures remains unclear. In summary, we found that in children referred for an EEG because of a possible history of seizures there were age-related differences in PSA between infants with normal development and those with developmental impairment. A challenge to clinicians is whether interventional techniques can be used to reverse these abnormalities and improve developmental outcome.
K. Kulandaivel, G.L. Holmes / Epilepsy & Behavior 20 (2011) 700–705
705
Fig. 6. EEG power spectra of infants N18 to 24 months of age. (A) PSA of all channels. (B) PSA of combined O1 and O2. Alpha percentage was lower in the developmentally impaired groups than in the normal group (A). As shown in (B), in the O1/O2 groups, percentages of alpha and beta differed among the groups.
Acknowledgment This work was supported by National Institutes of Health Grant R01NS044295 (G.H). References [1] Hauser WA. Epidemiology of epilepsy in children. Neurosurg Clin N Am 1995;6: 419–29. [2] Glosser G, Cole LC, French JA, Saykin AJ, Sperling MR. Predictors of intellectual performance in adults with intractable temporal lobe epilepsy. J Int Neuropsychol Soc 1997;3:252–9. [3] Bulteau C, Jambaque I, Viguier D, Kieffer V, Dellatolas G, Dulac O. Epileptic syndromes, cognitive assessment and school placement: a study of 251 children. Dev Med Child Neurol 2000;42:319–27. [4] Hermann BP, Seidenberg M, Bell B. The neurodevelopmental impact of childhood onset temporal lobe epilepsy on brain structure and function and the risk of progressive cognitive effects. Prog Brain Res 2002;135:429–38. [5] Cormack F, Helen CJ, Isaacs E, et al. The development of intellectual abilities in pediatric temporal lobe epilepsy. Epilepsia 2007;48:201–4. [6] Farwell JR, Dodrill CB, Batzel LW. Neuropsychological abilities of children with epilepsy. Epilepsia 1985;26:395–400. [7] Holmes GL, Lombroso CT. Prognostic value of background patterns in the neonatal EEG. J Clin Neurophysiol 1993;10:323–52. [8] Lombroso CT, Holmes GL. Value of the EEG in neonatal seizures. J Epilepsy 1993;6: 39–70. [9] Senior TJ, Huxter JR, Allen K, O'Neill J, Csicsvari J. Gamma oscillatory firing reveals distinct populations of pyramidal cells in the CA1 region of the hippocampus. J Neurosci 2008;28:2274–86. [10] Vertes RP, Kocsis B. Brainstem-diencephalo-septohippocampal systems controlling the theta rhythm of the hippocampus. Neuroscience 1997;81:893–926. [11] Buzsaki G. Theta oscillations in the hippocampus. Neuron 2002;33:325–40. [12] Itskov V, Pastalkova E, Mizuseki K, Buzsaki G, Harris KD. Theta-mediated dynamics of spatial information in hippocampus. J Neurosci 2008;28:5959–64. [13] Hyman JM, Wyble BP, Goyal V, Rossi CA, Hasselmo ME. Stimulation in hippocampal region CA1 in behaving rats yields long-term potentiation when delivered to the peak of theta and long-term depression when delivered to the trough. J Neurosci 2003;23:11725–31. [14] Gray CM, Viana Di PG. Stimulus-dependent neuronal oscillations and local synchronization in striate cortex of the alert cat. J Neurosci 1997;17:3239–53.
[15] Gray CM. Synchronous oscillations in neuronal systems: mechanisms and functions. J Comput Neurosci 1994;1:11–38. [16] Vanderwolf CH. Are neocortical gamma waves related to consciousness? Brain Res 2000;855:217–24. [17] Llinas R, Ribary U. Coherent 40-Hz oscillation characterizes dream state in humans. Proc Natl Acad Sci USA 1993;90:2078–81. [18] Chrobak JJ, Buzsaki G. High-frequency oscillations in the output networks of the hippocampal–entorhinal axis of the freely behaving rat. J Neurosci 1996;16:3056–66. [19] Hasselmo ME, Wyble BP, Wallenstein GV. Encoding and retrieval of episodic memories: role of cholinergic and GABAergic modulation in the hippocampus. Hippocampus 1996;6:693–708. [20] Lisman J. The theta/gamma discrete phase code occuring during the hippocampal phase precession may be a more general brain coding scheme. Hippocampus 2005;15:913–22. [21] Holmes GL, Lenck-Santini PP. Role of interictal epileptiform abnormalities in cognitive impairment. Epilepsy Behav 2006;8:504–15. [22] Hrachovy RA, Frost JDJr, Kellaway P. Hypsarrhythmia: variations on the theme. Epilepsia 1984;25:317–25. [23] Iranzo A, Isetta V, Molinuevo JL, et al. Electroencephalographic slowing heralds mild cognitive impairment in idiopathic REM sleep behavior disorder. Sleep Med 2010;11:534–9. [24] Stevens A, Kircher T, Nickola M, Bartels M, Rosellen N, Wormstall H. Dynamic regulation of EEG power and coherence is lost early and globally in probable DAT. Eur Arch Psychiatry Clin Neurosci 2001;251:199–204. [25] Brismar T, Hyllienmark L, Ekberg K, Johansson BL. Loss of temporal lobe beta power in young adults with type 1 diabetes mellitus. Neuroreport 2002;13:2469–73. [26] Van Dongen-Boomsma M, Lansbergen MM, et al. Relation between resting EEG to cognitive performance and clinical symptoms in adults with attention-deficit/ hyperactivity disorder. Neurosci Lett 2010;469:102–6. [27] Chung SS, McEvoy LK, Smith ME, Gevins A, Meador K, Laxer KD. Task-related EEG and ERP changes without performance impairment following a single dose of phenytoin. Clin Neurophysiol 2002;113:806–14. [28] Marciani MG, Maschio MC, Spanedda F, Gigli GL, Bassetti MA, Bernardi G. Sodium valproate and mental processes in newly referred epileptic patients: a computerized EEG study. Neuropsychobiology 1995;31:210–5. [29] Ackerman PT, Dykman RA, Oglesby DM, Newton JE. EEG power spectra of dysphonetic and nondysphonetic poor readers. Brain Lang 1995;49:140–52. [30] Ackerman PT, Dykman RA, Oglesby DM, Newton JE. EEG power spectra of children with dyslexia, slow learners, and normally reading children with ADD during verbal processing. J Learn Disabil 1994;27:619–30. [31] Koop JI, Fastenau PS, Dunn DW, Austin JK. Neuropsychological correlates of electroencephalograms in children with epilepsy. Epilepsy Res 2005;64:49–62.