A Novel Approach to the Study of Hypoxia-Ischemia-Induced Clinical and Subclinical Seizures in the Neonatal Rat

Abstract

Perinatal hypoxic-ischemic encephalopathy (HIE) is a major cause of acute mortality and chronic neurologic morbidity in infants and children. HIE is the most common cause of neonatal seizures, and seizure activity in neonates can be clinical, with both EEG and behavioral symptoms, subclinical with only EEG activity, or just behavioral. The accurate detection of these different seizure manifestations and the extent to which they differ in their effects on the neonatal brain continues to be a concern in neonatal medicine. Most experimental studies of the interaction between hypoxia-ischemia (HI) and seizures have utilized a chemical induction of seizures, which may be less clinically relevant. Here, we expanded our model of unilateral cerebral HI in the immature rat to include video EEG and electromyographic recording before, during and after HI in term-equivalent postnatal-day-12 rats. We observed that immature rats display both clinical and subclinical seizures during the period of HI, and that the total number of seizures and time to first seizure correlate with the extent of tissue damage. We also tested the feasibility of developing an automated seizure detection algorithm for the unbiased detection and characterization of the different types of seizure activity observed in this model.

Introduction

Perinatal hypoxic-ischemic encephalopathy (HIE) is a major cause of acute mortality and chronic neurologic morbidity in infants and children, with an incidence of 1–2 per 1,000 live term births [1,2]. Neonatal seizures occur in 1.8–3.5 per 1,000 live term births, with HIE being the most common cause [3]. Seizures are frequently both underrecognized and overdiagnosed, creating clear challenges in the design of effective treatment. Term infants with HIE can suffer from clinical seizures, with clinical correlates of seizure-like activity on the EEG occurring in an estimated 40–60% of all infants with HIE [4,5,6]. The concern over missing a high percentage of neonatal seizures, and in particular electrographic seizures, especially in severely asphyxiated infants, has prompted the increase in continual bedside monitoring by either amplitude-integrated EEG or video EEG (VEEG) [7]. Conversely, neonates may exhibit behaviors that appear to be seizures but lack an EEG correlate, leading to frequent overestimation of seizures in infants with HIE [8]. This ‘electroclinical dissociation’ or uncoupling of symptoms from EEG changes has led to confusion over whether such episodes might be subcortically generated seizures that may damage the brain and whether they should be treated [9].

The extent to which neonatal seizures damage the developing brain and whether or not they should be treated continues to be debated. Clinical studies suggest that, in term infants, an adverse neurological outcome occurs in approximately 30% of the survivors of neonatal seizures [10,11], with a stronger association between EEG seizures and poor neurodevelopmental outcome [12,13]. Data from both clinical and experimental studies suggest that seizures in the context of HIE are most likely to be associated with a more devastating outcome [8]. Although phenobarbital and phenytoin are the first-line anticonvulsants used clinically, the response to these drugs is not uniformly consistent as neonatal seizures are often refractory to these treatments [9]. These issues gain further importance because of the documented adverse effects of barbiturates and other anticonvulsant drugs on normal cerebral development [14].

One widely used animal model for the study of hypoxic-ischemic brain damage involves a unilateral carotid ligation in the neonatal rat, as originally developed by Vannucci and colleagues [15,16]. This model was originally developed in the postnatal-day-7 (P7) rat pup as representative of a 32- to 36-week-gestation human infant [17], but has been extended to rats and mice across the developmental age span from very premature (P2–3) through adult [18,19]. The majority of experimental studies that examined the effects of seizures in the context of HIE in this model employed seizure-inducing agents such as bicuculline [20], kainic acid and fluorothyl vapor [21] in rat pups from P7 to P10. These artificial means of seizure induction potentially make the results less relevant to the clinical situation. The primary goal of this study was to develop a methodology for monitoring naturally occurring seizure activity during and following hypoxia-ischemia (HI) in term-equivalent rat pups (P10–12) without chemical seizure induction; in addition, we tested the feasibility of developing an automated seizure detection algorithm (SDA) for the unbiased detection and characterization of the different types of seizure activity observed in this model, i.e. clinical, subclinical and behavioral.

Animals and Methods

Animals

Dated pregnant Wistar rats were purchased from Charles River Laboratories at embryonic day 15, housed individually and allowed to deliver vaginally. On the day of birth (equal to P1), pups from 3 litters were randomized and reassigned to the dams in groups of 10, with approximately equal numbers of males and females when possible. All procedures described were approved by the Weill Cornell Medical College animal care and use committee.

Procedure for Attachment of VEEG Head Mount

One day prior to right common carotid ligation and HI, pups were anesthetized with isoflurane (4% induction, 1–1.5% maintenance). The scalp was sterilized with topical Betadine. An incision was made rostral to caudal on the skin to expose the skull, and the exposed skull surface was dried with ethanol. A small amount (<20 µl) of cyanoacrylate was applied to the bottom side of the head mount (Pinnacle Technology Inc., Lawrence, Kans., USA), and the head mount was placed on the dry skull 3.0–3.5 mm anterior to the bregma and allowed to dry for 1–2 min. With a sterile 22-gauge needle, 4 small holes were made in the skull via the holes in the head mount, and the supplied screws were then passed through the holes in the head mount and affixed to the head mount and skull with silver epoxy. After the head mount was securely fastened, a small pocket was made in the neck muscle with a forceps to insert the electromyography (EMG) wires. The skin was secured over the head mount by placing 1–2 sutures at the front and back ends of the scalp incision. The duration of the surgical procedure was 10–15 min. The pups were allowed to recover from anesthesia and returned to their dam overnight.

Induction of Unilateral Cerebral HI

HI was induced according to our standard protocol [15,16] with modifications to accommodate VEEG monitoring (as described below). To best approximate the full-term human infant brain, P12 rat pups of both genders were used for this study [17]. The pups were anesthetized with isoflurane, a suprasternal midline incision was made, the right common carotid artery was isolated and ligated with 2 knots of surgical silk, and the wound closed (<10 min of anesthesia exposure). Control pups underwent the above sedation and midline incision but did not undergo the right common carotid ligation. Sham-ligated pups were also exposed to the hypoxia and VEEG analysis. Upon recovery, the pups were returned to their dam for 2 h. Prior to HI, the pups were connected to the VEEG via a swivel connector and allowed to move freely within the chamber. The chamber temperature was maintained at 36°C. After a baseline VEEG was obtained, a gas mixture of 8% oxygen and 92% nitrogen was delivered to the chamber for 75 or 90 min, for moderate and severe insults, respectively. VEEG monitoring was performed during HI and for 1 h afterwards. After recovery, the pups were returned to their dams. Twenty-four hours after HI, a 1-hour spot VEEG monitoring was performed. Seventy-two hours after HI, brains were removed and frozen in −30°C isopentane and stored at −80°C. Then, 20-µm cryosections were stained with HE, and the infarct area of the ipsilateral hemisphere – relative to the contralateral hemisphere and correcting for edema [22] – was calculated using NIH ImageJ.

Data Acquisition

Two channels of EEG and 1 of EMG from the neck muscle were recorded along with time-synchronous digital video using the Pinnacle PAL-8200 data acquisition system (Pinnacle Technology). The signal was passed through a low-pass filter with a cutoff of −3 dB at 100 Hz and stored at a sampling rate of 400 Hz in European Data Format files. VEEG recordings were obtained prior to, during, immediately after and 24 h after HI.

Seizure Definition and Scoring

Using the Sirenia software package by Pinnacle Technology, a single observer analyzed the VEEG data. For this study, a seizure on EEG was defined as a recording with rhythmic or repetitive tracing with an amplitude that increases to more than 3 times the baseline level and lasts at least 3 s (fig. 1). Clinical seizures were defined as any observed repetitive movements in the pup, such as forelimb and hindlimb paddling, circling or head rearing, that correlated with a seizure on EEG. Subclinical seizures were defined as seizures observed on EEG without any visibly correlated behavioral movements. Clinical and subclinical seizures were recorded.

Fig. 1.

A 30-second interval of EEG recording demonstrating a clinical seizure (EEG 2) defined as a recording with a rhythmic or repetitive tracing with an amplitude that increases to more than 3 times that of the baseline and lasting 3 s or longer. The top EEG trace (EEG 1) measures the potential between the left posterior and right anterior cortices. The middle EEG trace (EEG 2) measures the bilateral difference in potential in the parietal cortex. The bottom trace is the EMG recording from the upper extremities. Events such as this are compared with video to categorize them as clinical or subclinical seizures. Inset Video signal recorded in synchrony with EEG/EMG.

Data Analysis

Linear regression analysis was performed for damage (infarct area) as a function of number of seizures (clinical, subclinical, total) and time to first seizure (clinical and subclinical); Pearson's correlation coefficient and p values were calculated. Unpaired t tests were used to compare damage between males and females.

Development of Automated SDA

The procedure for electrographic seizure detection was as follows. The cortical EEG signals were high-pass filtered at 5 Hz to exclude possible motion artifacts, at the same time amplifying high-frequency epileptiform spikes; then the mean square value was computed in successive 1-second windows. A combination of rank filters (i.e. max or min operations) was applied to estimate the seizure content of the signal. These filters, more commonly known as morphological operators, are good at capturing the texture or topography of a signal. Simple recursions of max and min operations in a window can produce similarly useful nonlinear filtering effects. Morphological operators offer an attractive means of capturing the structure of individual spikes in a seizure, thereby detecting seizures with greater specificity. First, a large morphological opening (i.e. a min followed by a max operation, or ‘min-max’, in a 60-second moving window) was taken to estimate the instantaneous baseline activity and subtract it from the mean square signal; this helped compensate for changes in signal strength over the course of the experiment and produced a uniform baseline against which seizures stood out as bursts of power. During a seizure, spiking activity may occur in brief bursts with periods of suppression between them. A close-open operation (i.e. a closing followed by an opening, where close = ‘max-min’) was applied to merge bursts of activity separated by less than 10 s and therefore part of the same event. The close-open operation also has the property of preserving isolated bursts of spikes. Finally, a threshold was applied to separate the resulting trace into candidate seizure and nonseizure segments.

Although the behavior associated with clinical seizures was verified by visual review of the digital video signal, the activity associated with seizure-like behavior is usually evident on the EMG. This provided the opportunity to characterize the electrical uncoupling phenomenon observed in HIE. Seizure-like behavior was detected using exactly the same procedure for EEG seizure detection, except that it was applied to the EMG signal to detect tonic-clonic activity, and a different threshold was required. A validation was performed to determine if there was a genuine correlation between observed behavior and SDA output for the EMG signal. Seizures detected on EEG that overlapped with seizure-like behavior detected on the EMG were labeled as clinical seizures; if unaccompanied by detected EMG activity, they were labeled as subclinical (electrographic) seizures. Correlations between behavioral analysis and SDA EMG output were analyzed by Kruskal-Wallis analysis of variance.

Results

Twelve pups underwent HI for 90 min with VEEG analysis (table 1). Of the 12, only 1 pup did not have any clinical or subclinical seizures and did not have brain damage on HE staining. One other pup exhibited 8 clinical seizures but no subclinical seizures and had the most severe damage (90% infarct area). The remaining 10 pups exhibited both clinical and subclinical seizures, and for 9 of these 10 the damage ranged from 48.5 to 80%; 1 of the 10 exhibited only 1 clinical and 1 subclinical seizure at the very end of hypoxic exposure, and this pup had very mild damage despite exhibiting 5 clinical seizures during the initial recovery period. Two control (sham-operated) pups were exposed to 90 min of hypoxia and VEEG analysis, and both demonstrated repetitive behavioral movements suggestive of seizure activity, but that did not correlate with any EEG alterations and – as is routinely reported for hypoxic exposure alone – did not have any damage.

Table 1.

Clinical and subclinical seizure detection in 12 P12 rats during 90 min of HI

No.	Sex	Total n	Clinical		Total n	Subclinical
			time of 1st Sz after start of HI min	duration range s		time of 1st Sz after start of HI min	duration range s	damage % area
1	female	4	51	24–95	12	47	3–90	51.7
2	male	1	81	55	8	48	35–70	48.5
3	female	8	31	10–140	10	34	8–180	51.5
4	male	0			0			0
5	male	1	90	45	1	86	47	2.8
6	female	7	37	20–60	6	40	6–110	72.0
7	female	8	31	22–110	14	35	10–108	70.7
8	female	5	56	16–120	10	67	11–90	67.4
9	female	7	6	10–55	15	37	5–87	80.3
10	male	8	40	10–96	0			90.7
11	female	10	37	10–90	4	35	13–78	57.3
12	male	10	72	6–105	5	74	5–150	49.0

Total numbers denote numbers of seizures. Sz = Seizure.

Ninety minutes of HI resulted in severe brain damage for the majority of the pups, making the interpretation of the correlation between seizure activity and extent of damage more difficult. To reduce the overall severity of the insult and to further examine possible correlations over a greater range of damage, 15 pups were subjected to HI for 75 min and VEEG analysis (table 2). In the 75-min group, the damage ranged from 0 to 83%, with 2 pups having no damage and 3 exhibiting only mild-to-moderate damage. Of the 2 with no damage, 1 did not have any seizure activity during or after HI, and the other had 8 short (10- to 40-second) clinical seizures during HI, no subclinical seizures during HI, but 7 subclinical seizures during the initial 1-hour recovery period (data not shown). All 3 pups with less damage had 5 or fewer clinical seizures during HI. The 75-min hypoxic controls did not exhibit any seizure or behavioral abnormalities and had no damage.

Table 2.

Clinical and subclinical seizure detection in 15 P12 rats during 75 min of HI

No.	Sex	Total n	Clinical		Total n	Subclinical
			time of 1st Sz after start of HI min	duration range s		time of 1st Sz after start of HI min	duration range s	damage % area
1	female	5	33	20–40	7	48	20–96	32.1
2	female	14	27	10–105	9	25	12–120	72.0
3	male	6	36	56–220	9	26	16–210	42.4
4	female	0			0			0
5	female	7	30	10–120	5	47	36–100	50.5
6	male	3	29	84–100	8	32	26–165	35.8
7	female	2	58	26–40	0			0
8	male	9	38	22–110	5	58	12–58	49.4
9	female	17	24	14–160	9	17	17–93	83.4
10	male	5	46	30–120	7	39	25–82	70.6
11	female	9	34	35–124	5	49	11–27	61.9
12	male	14	40	5–130	3	61	40–81	49.3
13	male	7	41	58–150	7	34	23–110	58.5
14	female	7	41	10–50	2	56	16–30	65.7
15	male	5	53	10–81	0			27.4

Total numbers denote numbers of seizures. Sz = Seizure.

Although these two sets of experiments (90 and 75 min of hypoxic exposure) were conducted at different times, there was no difference between the groups in the median time to first clinical or subclinical seizure (fig. 2). The median time periods to the first clinical seizure after initiation of HI were 40 (range: 6–90) min and 37 (range: 24–58) min for the 90-min and 75-min groups, respectively, and for subclinical seizures, they were 43.5 (35–86) min and 43 (17–61) min. The effect of the shorter duration of HI on the range of damage scores is depicted in figure 3. The 2 pups of the 90-min group with little-to-no damage were statistical outliers, whereas the minimally damaged pups of the 75-min group were not, generating a significantly greater extent of infarct from 90 min of HI, as predicted. The time to first clinical seizure was significantly correlated with the extent of damage for both groups (fig. 4; table 3), whereas the time to first subclinical seizure was a significant predictor of damage only for the 90-min group. Although total numbers of seizures were significantly correlated with damage for both groups, the level of significance was greater for the 75-min group, supporting our original hypothesis that the reduction in severity of insult would demonstrate the interactions more clearly.

Fig. 2.

Time to first seizure was different between the 75-min group and the 90-min group. a Clinical. b Subclinical.

Fig. 3.

Brain damage as a function of duration of HI. Box-and-whisker plots demonstrating the range of damage, expressed as the infarct area, the percentage of the ipsilateral hemisphere relative to the contralateral hemisphere. The shorter duration, 75 min, resulted in a greater range of damage, whereas the more severe 90-min insult resulted in a greater number of severely damaged animals. The 2 animals with no damage were determined to be statistical outliers (more than 2 SD from the mean) and were not included in the analysis.

Fig. 4.

Hypoxic-ischemic brain damage as a function of numbers of seizures. Pearson correlation coefficients for these analyses are presented in table 3. The 2 animals with no damage in the 90-min group are more than 2 SD from the mean and were not included in the statistical analysis. a, b 90 min of HI. c, d 75 min of HI.

Table 3.

Correlation of seizure yield statistics with anatomical brain damage reported as the Pearson correlation coefficient

	Correlation coefficient	p
Time to 1st clinical Sz–75 min	−0.599	0.024
Time to 1st subclinical Sz–75 min	−0.339	0.282
Total number of clinical Sz–75 min	0.763	0.0009*
Total number of subclinical Sz–75 min	0.645	0.009*
Total number of Sz–75 min	0.833	0.0001*
Time to 1st clinical Sz–90 min	−0.741	0.009*
Time to 1st subclinical Sz–90 min	−0.691	0.027
Total number of clinical Sz–90 min	0.549	0.08
Total number of subclinical Sz–90 min	0.702	0.024
Total number of Sz–90 min	0.701	0.011*

Correlations that were significant at the 99% confidence level (p < 0.01) are identified by an asterisk.

Development of SDA

While developing this unique method of observing seizures in the HI neonatal rat model, we wanted to develop a system that would allow easy identification of clinical and subclinical seizures. The initial analysis of VEEG in this study was performed by human observation and correlation of seizure activity, both behavioral and electroencephalographic. The observed electrographic seizure activity in the HI animals was characterized by rhythmic large amplitude spikes at seizure onset, repeating in the range of 4 units per second. Hence, after some experimentation, an SDA based on spectral properties and the morphology of this activity was proposed for quantifying seizure activity in this model. This algorithm was described earlier in the Animals and Methods section. The seizures can last tens of seconds, and may be classified as clinical or electrographic (subclinical) depending on whether or not they are accompanied by characteristic behavioral manifestations, respectively. Candidate detections of clinical and subclinical seizures made by an SDA in sample recordings were compared against human expert scoring of the VEEG to evaluate the performance of the algorithm. Figure 5 depicts the working of the algorithm for a typical HI experiment. Various statistical descriptors can be extracted from this analysis. For instance, the time of the first seizure detected on the EEG was 447 s into the recording; the animal experienced 34 subclinical seizure detections with a mean duration of 23 s. A comparison of the SDA output with human scores gave an agreement of about 70–80%; in some instances, the algorithm uncovered events that were missed during visual review, which may have contributed to the SDA errors.

Fig. 5.

Sample seizure detection analysis for a P12 rat pup subjected to 90 min of HI. a Raw EEG and EMG signals during HI. EEG and EMG traces for the duration of hypoxia, with bursts of activity seen during possible seizures. b Mean signal power in a 1-second moving window. The signals are high-pass filtered at 5 Hz to reduce motion artifacts while amplifying high-frequency epileptiform spikes, and the mean square value in a moving 1-second window is computed. c Background correction and estimation of seizure content. The output is corrected for gross variations in the signal background, and a combination of rank filters (max-min operations) is applied to estimate the ‘seizure content’ of the signal in terms of the ratio of epileptiform activity normalized to background power. d Thresholding indicates periods of seizure activity. A threshold is applied to the EEG and EMG outputs in c to identify periods of time corresponding to candidates for subclinical seizures (EEG only), behavioral ‘seizures’ (EMG only) and clinical seizures (activity on both EEG and EMG). These candidates are verified visually by a human expert, and the performance of the algorithm is assessed using standard receiver operating characteristic analysis.

Correlation of EMG with Behavior

A validation was performed to determine if there is a genuine correlation between observed behavior and SDA output for the EMG signal. This was done to determine whether the EMG is a reliable indicator of behavior-related movement for discrimination between clinical, subclinical and behavioral seizures. The behavior of the P12 rat pup shown in figure 1 was scored based on the video monitoring obtained during hypoxic exposure and without access to the corresponding EMG recording. Observed behaviors were classified by the nature of the movement and assigned to one of three categories: (1) no movement; (2) forelimb or hindlimb paddling or visible tensing, and (3) vigorous circling movements of the lower body. Brief jerks or twitches of <1 s duration were excluded due to the difficulty of aligning the visual scores with the EMG data with adequate precision. Values of the SDA output computed sequentially from 2-second windows of the EMG recording were tested for significant differences between the three categories of behavior using the nonparametric Kruskal-Wallis analysis of variance. In the event of a significant difference (p < 0.05), post hoc pairwise comparisons were made to identify interindividual differences. The Kruskal-Wallis analysis was used because it is a distribution-free test and the SDA output is likely to be nonnormal and highly skewed to the right.

The results indicate a highly significant difference among the three groups, with p < 0.0001. Categories 2 and 3 (seizure-like movements) were significantly different from category 1 (no movement). Although categories 2 and 3 were not significantly different from each other, the mean SDA output for category 3, which contained more vigorous movements, was an order of magnitude higher than for category 2. These results indicate that EMG can be used to obtain some level of discrimination between behaviors associated with seizure-like events in the neonatal HIE model. Taken together, these results suggest that the development of an SDA for characterizing the seizure yield of HI with reasonable accuracy is feasible. However, a complete performance characterization and optimization of the SDA was beyond the scope of the present study and will be reported later.

Discussion

We have developed a unique method of observing and quantifying seizures in the neonatal rat during and following unilateral cerebral HI, without chemical means of seizure induction. To the best of our knowledge, this is the first paper to publish this novel methodology in this model to date. The benefit of the VEEG technique employed in this study is its ability to identify seizure-like movements and to correlate them with the EEG. This allowed us to distinguish between movements associated with an electrographic seizure and those that merely mimic the physical signs of seizure but do not produce an EEG recording consistent with a seizure. In addition, the use of VEEG facilitated the identification of naturally occurring clinical and subclinical seizures induced by HI, which could then be correlated with the severity of damage to the ipsilateral hemisphere. This study demonstrates that immature rats exhibit both clinical and subclinical seizures during HI, as well as numerous seizure-like movements without EEG changes. The motor activity typical of clinical seizures sometimes occurs without any discernible change on the EEG: these may be referred to as behavioral seizures, although the provenance of these events is not known and may be subcortical and therefore not visible in the EEG measurements used in the present study. It is important to characterize all three seizure types – electrographic, clinical and behavioral – and to correlate the yield with experimental conditions. There is no clear treatment of purely behavioral seizure-like events as a consequence of HI described in the literature, except for a brief mention in the review by Mizrahi and Kellaway [8], which recommends that they be regarded as questionable. Since it is not clear whether such events should be treated, they warrant closer investigation. Fortunately, our SDA can identify seizure-like behavior using EMG (and, by exclusion, on EEG), which provides the unique opportunity to study their yield as a consequence of HI and/or therapeutic intervention in future investigations.

The unilateral HI model employed here was initially developed for the P7 rat pup, as being roughly comparable to the 32- to 36-week-gestation human infant in terms of the maturational state of the brain [15,17], and the most extensive characterization of the effects of HI has been done for this age. This model has now been extended to study hypoxic-ischemic injury over a range of developmental ages, in both the rat and the neonatal mouse. For this initial study, we specifically chose to study the P12 rat because we were interested in a full-term-equivalent (40-week-gestation) model [17,23]. A recent study correlating the development of amplitude-integrated EEG to correlate rat to human brain maturation supports that P7 most closely approximates 30–32 weeks of gestation, while P9–12 is representative of 40–42 weeks [23].

Other investigators have reported on the effects of seizures on the hypoxic-ischemic damage in the immature rodent brain. Wirrell et al. [24] induced seizures with kainic acid after HI in P10 rat pups. Their results demonstrated that seizures superimposed on HI exacerbated brain damage. However, in the same study, seizures induced in rats not exposed to HI did not cause any neuropathologic injury. They subsequently demonstrated that the exacerbated brain damage of HI plus kainic acid was related to the hyperthermic effect of the seizures [25]. The absence of histologic damage due to seizures alone has been confirmed by other studies exposing rats at various ages to kainic acid [26,27]. In the present study, rats were exposed to HI in temperature-controlled chambers, but we did not specifically monitor individual temperatures. Hayakawa et al. [28] studied the effect of pretreatment with the anticonvulsant zonisamide on hypoxic-ischemic damage in the P7 rat pup, and monitored EEG activity in a very small cohort of treated and control pups during hypoxia, making that study most similar to the one reported here. However, pharmacological treatment did not have any effect on the seizure activity recorded, and only brief information was presented, making any further comparisons difficult.

Most studies and clinicians have focused on the treatment of clinical seizures in the neonatal population, but subclinical seizures may account for more than 60% of seizures [29,30]. Whether or not to treat subclinical seizures, and their role in HIE brain damage, continues to be debated. In an HIE piglet model, subclinical seizures have been shown to occur in 29% of the observed animals during resuscitation and have been associated with brain damage [31]. We observed subclinical seizures in 80% of the pups during HI in this study and concur with the piglet model finding that both clinical and subclinical seizures are associated with brain damage. However, a separate analysis of the contribution of subclinical seizure activity alone was not conducted in this study.

The results from this study mimic the acute phase of HIE, where seizures are observed during and immediately after the HI period [2]. Due to the time-consuming methods required for each rat VEEG experiment, a 1-hour spot VEEG recording at 24 h after the HI insult was performed and revealed clinical or subclinical seizures (data not shown). Continuous 24-hour VEEG monitoring, at least during the initial post-24 h, would compromise feedings with the dam and would require data storage resources that were beyond the scope of the present feasibility study. It is our speculation that spot VEEG for a period of 4–6 h during the first 24 h after HI insult may be more feasible and yield a higher incidence of seizures. Continued development and validation of the SDA presented here will facilitate the review of many hours of VEEG/EMG recordings.

The methods developed in this study provide the ability to study the effects of HI and seizure activity over a wide range of developmental ages. Tucker et al. [23] studied amplitude-integrated EEG in the rat from P1 to P21. They demonstrated that as the rats matured, their raw EEG became more continuous. Our study confirms these findings with the baseline EEG recordings appearing continuous. We have performed a pilot study on the P7 rat and have observed EEG recordings that are more discontinuous when compared to the P12 rat (data not shown). It is possible that the difference in EEG patterns corresponding to maturity may also be observed with seizures and HI. Future studies will validate this seizure detection model at different developmental stages.

While this study is able to demonstrate an association between seizures and HI brain damage, it does not address the issue of whether seizures are a reflection of the ongoing brain damage during HI or whether they contribute to further dysfunction and damage. Eliminating or reducing the amount of seizures during HI will help elucidate the answer to this question. In support of this concept, Liu et al. [32] studied topiramate in combination with hypothermia after HI and found decreased brain damage. VEEG may be able to discriminate whether or not the decreased pathology is due to protective effects of topiramate or its antiseizure properties. Phenobarbital, a commonly used antiseizure medication, would be a logical drug to study with our model, but animal studies have hinted at deleterious effects on the brain [33,34]. However, a recent study demonstrated that treatment of neonatal rats with phenobarbital after HI improved the neuroprotective effects of hypothermia [35], although seizure activity was not specifically investigated in this study. In conclusion, our novel VEEG model is an exciting and useful tool for observing both clinical and subclinical seizures during HI. While it does have physical limitations with regard to the number of days that the head mount can remain in place, it has the potential for a wide variety of applications due to its unique ability to distinguish between movements associated with electrographic seizures and those that are not. With the use of VEEG, the questions of what contributing role seizures have in HI brain damage and what is the optimal anticonvulsant treatment for seizures in the setting of HI brain injury can be answered.