Article Text
Abstract
Background Amplitude integrated electroencephalography (aEEG) is a valuable tool for evaluating neonatal encephalopathy and identifying electrographic seizures.
Objective To compare seizure activity and background pattern (BGP) between one-channel and two-channel aEEG recordings in full-term neonates.
Methods The two-channel aEEG recordings (F3-P3; F4-P4) of 34 neonates with seizures were compared with single-channel recordings (P3-P4).
Results All 34 infants with unilateral (n=14), diffuse (n=18) or without (n=2) brain injury had seizure patterns on one-channel and two-channel recordings, with 18% more seizure patterns detected with two-channel recording. In 79% of infants with unilateral injury more seizures were noted on the ipsilateral side compared to the contralateral side. In 39% of the infants with diffuse brain damage more seizures were found with two-channel recordings. A sensitivity of 65% was found when using the automatic seizure detection algorithm. In 4/14 (29%) infants with unilateral injury a more severely affected BGP was seen on the ipsilateral side compared to the BGP on one-channel recording. In infants with diffuse injury differences in BGP pattern were seen in 6–17% of the infants depending on the system used for scoring.
Conclusion Although there were no major differences found between seizure detection with one-channel or two-channel aEEG, in a subgroup of infants with a predominantly unilateral brain lesion, two-channel recording did provide additional information with identification of more seizure patterns on the affected side, sometimes also associated with a difference in BGP. To improve early diagnosis of unilateral lesions and improve seizure detection in these infants, routine use of two-channel recordings is recommended.
Statistics from Altmetric.com
Introduction
With the introduction of amplitude integrated electroencephalography (aEEG) in neonatal intensive care units (NICUs) we now have a simple clinical tool for monitoring newborn brain function. This technique enables us to monitor infants with neonatal encephalopathy (NE), to identify some, but not all seizures and to assess the response to anti-epileptic drugs.1,–,8 One of the advantages of this technique is that it is available 7 days a week, 24 hours a day and it is relatively easy to learn for doctors as well as nursing staff. It is well known that background activity is of predictive value for neurodevelopmental outcome.9 Although there is considerable concern that one-channel or two-channel recordings will miss some of the seizure activity, it has been shown that 80–90% of infants with EEG confirmed seizures were identified by aEEG.8 10,–,12 The development of new digital aEEG equipment during recent years allows us to study more channels with simultaneous display of the original EEG signal, and a novel seizure detection algorithm was developed to improve seizure detection with aEEG.11 13,–,15 The original EEG facilitates differentiation between true ictal patterns and artefacts and with the seizure detection algorithm care-givers will be alerted that seizures may occur. The use of two channels can help to determine hemispheric asymmetries, for instance, in children who have a unilateral brain lesion.16 Whether the use of two channels will improve seizure detection is still a matter of debate.
What is already known on this topic
▶. Amplitude integrated EEG is a technique that can identify some, but not all, seizures in newborn infants and that can help to assess the response to anti-epileptic drugs.
▶. The use of two-channel aEEG recording can help to determine hemispheric asymmetries, for instance, in children with unilateral brain lesions.
What this study adds
▶. When using two-channel aEEG recording in neonates with a unilateral brain lesion more seizure patterns will be detected compared to one-channel recording.
▶. Using two-channel aEEG recording in infants with convulsions can give information about the presence of a unilateral brain lesion before neuroimaging is performed.
The aim of this study was to compare assessment of seizure activity and background pattern (BGP) between one-channel and two-channel aEEG recordings in full-term neonates with NE and either bilateral or unilateral lesions on neuroimaging.
We also evaluated the sensitivity for the automatic seizure detection algorithm in a subgroup of infants.
Patients and methods
Since 1992 all infants who were admitted to our regional tertiary NICU because of NE and/or suspected seizures have been monitored using aEEG as part of routine care. The first 11 years an analogue machine was used (CFM 4640; Lectromed, Devices Ltd, UK), and since 2003 two digital machines are used (Olympic 6000 and BRM 2 Brain Monitor; Natus, Seattle, WA, USA).
For this study we selected a group of 34 infants with aEEG-detected seizure activity, born between January 2004 and March 2008. Brain damage was assessed with ultrasound and MRI. Infants were born at term or had reached term equivalent age when the aEEG recording was started.
aEEG monitoring
For recording the BRM2 Brain monitor (Natus, Seattle, WA, USA) was used. The BRM2 records a two-channel aEEG as well as a raw EEG from two needle electrodes over each hemisphere (F3-P3, F4-P4, according to the international 10–20 system of electrode placement). The monitor can also provide a single transverse ‘cross-cerebral’ channel P3-P4 recording, as conventionally used in single-channel aEEG recordings.
aEEG analyses
For analysis, the aEEG recordings were reviewed in the Analyze 2 offline review mode (BrainZ Instruments Ltd 2008, New Zealand). With this new offline analysis program there is also access to the cross-cerebral raw EEG signal. The amplitudes of the aEEG traces were analysed using Analyze research version 1.4 (BrainZ instruments Ltd, 2006), which calculates median values for the minimum and maximum amplitude of the aEEG over any chosen time period.
Seizures patterns
Seizure patterns were defined as a characteristic pattern, with a sudden increase of the lowest and highest amplitude of the recorded signal due to increased amplitude during epileptiform activity and a decrease in voltage in the postictal period. A status epilepticus (SE) pattern was defined as continuous seizure pattern for ≥30 minutes, which can be seen as a ‘saw-tooth pattern’ or as a continuous increase of the minimum and maximum amplitude.17 Both one-channel and two-channel recordings were independently assessed by two experienced observers (LVR and MCT). The traces were studied in random order and when assessing the cross-cerebral recordings the observers were not aware of the results of the assessment of the two-channel recordings. All suspected seizure patterns on aEEG were reviewed with the raw EEG to confirm seizure activity. Only those seizure patterns were selected that could be confirmed with the raw EEG. Infants recorded after October 2006 were recorded with one of the latest software versions including seizure event detector software (Recognize). When a potential epileptiform pattern in the raw EEG pattern was recognised by the algorithm a signal seen as an orange bar is shown above both the raw EEG and the aEEG tracing. This signal is given irrespective of the channel in which the event was detected. So every time the algorithm gave an alert, this period was checked even when there were no suspected epileptiform patterns on the aEEG. Once all the traces were reviewed, the number of seizure patterns in each recording (left side, right side and cross-cerebral recording) was counted. When a SE pattern was seen we noted this as one period. When there was disagreement between the two observers, a consensus was reached with a third reader (LdV) and when there was disagreement on comparison of the aEEG with the raw EEG consensus was reached with a clinical neurophysiologist (ACvH). After the assessment of all records the two-channel and cross-cerebral recordings were compared for each patient. All single seizures patterns and periods of seizure patterns were compared for time of onset in both recordings.
Figure 1 shows a diagram with a summary of the analysis.
Background pattern
The first period of 2.5–3.5 hours without seizures or artefact and impedance below 5 kΩ was chosen for comparison of BGP. For that period the BGP was scored using our criteria published previously18 and for all these periods we also calculated (using Analyze Research software) the median minimum and maximum amplitude and classified the BGP using the criteria of al Naqeeb,1 normal amplitude (N): maximum amplitude >10 µV, minimum amplitude >5 µV; moderately abnormal (MA): maximum amplitude >10 µV, minimum amplitude ≤5 µV and severely abnormal (SA): maximum amplitude <10 µV, minimum amplitude <5 µV.
Results
Table 1 provides the clinical characteristics and diagnosis of the 34 selected infants. All infants received at least one anti-epileptic drug (AED) with a median of three different drugs. An MRI was performed during the neonatal period in 31 of the 34 infants. In two infants no abnormalities were seen on MRI. One of these infants was subsequently diagnosed with benign familial convulsions (with mutation of the KCNQ2-gene). The other infant still has epilepsy of unknown aetiology.
In figures 2 and 3 detailed information about brain injury seen on MRI or ultrasound is given.
Seizure patterns identification
Unilateral lesions
Table 2 gives the number of seizure patterns for each patient detected with both recordings. For the two-channel recording we noted the number of seizure patterns on the ipsilateral as well as the contralateral side. The ipsilateral side was chosen as the side for comparison.
Comparison of ipsilateral and cross-cerebral recording
On the ipsilateral side a total of 166 seizure patterns were seen and 126 (76%) of these seizure patterns were also detected with the cross-cerebral recording. Fifteen of 141 (11%) of seizure patterns, seen on cross-cerebral recording, were not detected with the two-channel recording. When taking all seizures together on the ipsilateral side, 166 compared to 141 on the cross-cerebral recording were detected, giving 18% higher detection rate (owing to lateralisation of the epileptiform activity).
In recordings of seven infants (50%) more seizure patterns were seen on the ipsilateral side, in two (14%) infants (cases 5 and 14) more seizure patterns were identified with cross-cerebral recording and in five (36%) infants the number of detected seizure patterns was identical. In most cases the differences were restricted to a few additional short seizure patterns. In one infant (case 2) seizures were clearly identifiable in all channels, but they were more evident on the affected side. In two infants (cases 4 and 8) seizures were more evident on the ipsilateral and cross-cerebral recording compared to the contralateral side.
Comparison of ipsilateral and contralateral recording
When comparing the contralateral side with the ipsilateral side only 36 (22%) of the 166 seizure patterns were recognised. On the contralateral side six short single seizures were identified that were not recognised on the ipsilateral side. In 11 infants (79%) more seizure patterns were identified on the affected side (in case 14 the number of seizure patterns was higher on the left side, but during these repetitive seizure patterns an SE was seen on the affected right side, so effectively more seizure patterns were seen on the affected side) and in six of these 11 infants not a single seizure was detected on the contralateral side (figure 2). In one infant a period of 11 minutes of seizure discharges was seen on the contralateral side, while no discharges were seen on the affected side (case 13).
Bilateral lesions
Table 3 gives the number of seizure patterns for each patient detected with the two-channel and cross-cerebral recording. For the two-channel recording we only noted the side with the highest number of seizure patterns. With two-channel recording a total of 199 seizure patterns were identified and 167 (84%) of these patterns were seen on cross-cerebral recording. Seizure identification between two-channel and one-channel cross-cerebral recording was equal in eight of the 18 (44%) infants. In seven of the 18 (39%) infants more seizure patterns were detected with two-channel recording; in two of these seven cases an SE was seen with all recordings, but after the SE some short seizure patterns were detected on the right side, while not identified on the cross-cerebral recording and in the other five infants the differences were small. Finally in three of the 18 (17%) infants more seizure patterns were seen on the cross-cerebral recording, but only in one infant (case 32) was this difference substantial.
Comparison of left and right side of the two-channel recording
Seizure patterns were equally detected in seven of the 18 (39%) infants (table 4). In the other infants there were small differences between both sides with, in eight infants, more seizures patterns identified on the right side, but the majority of the seizure patterns were simultaneously detected on both sides.
No brain damage
In the two infants without apparent lesions on MRI the identification of seizure patterns was almost equal in all recordings (table 5). In one patient one short discharge detected on the cross-cerebral recording was not identified with both two-channel recordings.
Seizure detection algorithm
Fifteen infants were recorded after October 2006 and reviewing their recordings with Analyze 2 the seizure detection algorithm was used as well. During the cumulative 2150 hours of recording in these 15 infants 214 electrographic seizure patterns were identified by the reviewers. Of these 214 events, 140 (sensitivity 65%) were detected by the algorithm. It was noted that in ongoing seizures (>20 minutes) not the whole period was detected by the algorithm. During the recording time there were another 158 periods detected by the algorithm that could not be confirmed as ictal discharges, giving a false positive rate of 1 per 13.5 hours of recording.
Background patterns
Table 6 shows the qualitative and quantitative BGP for both recordings.
Unilateral lesions
When looking at qualitative pattern recognition18 the same BGP was seen in both two-channel and cross-cerebral recording in eight of the 14 cases (57%). In two infants (cases 2 and 9, figure 3) an abnormal pattern was seen (burst suppression (BS), continuous low voltage (CLV or flat trace (FT)) on two-channel recording on both ipsilateral and contralateral sides while on cross-cerebral recording an almost normal pattern was seen (discontinuous normal voltage (DNV)). In two infants (cases 7 and 11) an abnormal pattern was seen on the ipsilateral side recording while on the contralateral side and on the cross-cerebral recording a normal pattern was seen. In two cases (10 and 12) there were only small differences in BGP between the ipsilateral side and the contralateral side as well as the cross-cerebral recording; in one a DNV pattern was seen on the affected side, and a continuous normal voltage (CNV) pattern on both other recordings and in the other infant a CLV pattern instead of a BS pattern was seen on the affected side. In six of the 14 infants (43%) with unilateral brain injury, the crossover background pattern was better than the background pattern of the affected side.
When looking at the quantitative criteria1 no differences in BGP between both recordings were seen in 10 of the 14 infants (71%). In three infants (cases 2, 7 and 9) an SA pattern was seen on both ipsilateral and contralateral sides, while on cross-cerebral recording a normal pattern was seen. Only in one infant (case 11) was an MA pattern seen on the ipsilateral side, while the contralateral side and the cross-cerebral recording showed a normal pattern.
Bilateral lesions
With qualitative pattern recognition no differences in BGP between two-channel and crossover recordings were seen in 17 of the 18 infants with bilateral lesions. In only one infant (case 25) was a CLV pattern seen on the left side of the two-channel recording while on the right side and with cross-cerebral recording a normal BGP was seen. With quantitative criteria the same BGP was seen in 15 of the 18 infants (83%). In two infants (cases 30 and 31) an MA pattern was seen on two-channel recording while a normal pattern was seen on cross-cerebral recording, and in one infant (case 25) an SA pattern was seen on the left side and an MA pattern on the right side and cross-cerebral recording. The two infants without brain injury showed a normal BGP on both recordings.
When comparing the two scoring systems for all 34 infants there was in some cases a discrepancy between an MA pattern when using the voltage criteria compared with a DNV or BS pattern when using the pattern recognition. In six of the 10 infants where an MA pattern was seen with voltage criteria a BS pattern was seen with pattern recognition. In the other four infants a DNV pattern was seen.
Discussion
In this study we have shown that two-channel aEEG may provide additional information, especially in infants with a predominantly unilateral brain lesion. In 11/14 (79%) infants more seizure patterns were seen on the ipsilateral side compared to the contralateral side, and in six of these infants no seizure patterns, not even single ones, were seen on the contralateral side. Even in 7/18 (39%) infants with diffuse brain damage more seizure patterns were detected with two-channel recordings, although the differences were small and in all infants at least some seizure patterns were detected simultaneously on the crossover recording. There was a small difference between one-channel and two-channel aEEG monitoring with respect to background pattern, using either qualitative or quantitative pattern recognition. An abnormal BGP on the affected side, in contrast to a normal BGP pattern on the crossover recording, was seen in four infants with a unilateral lesion. When comparing both methods to score the BGP most differences were seen in infants with an MA pattern when using the voltage criteria. In 60% of these infants a BS pattern was seen with pattern recognition, which is often associated with an abnormal outcome.9 Differences in amplitude may lead to misinterpretation in BGP, which is important when aEEG is used for inclusion of infants in intervention studies, such as hypothermia.19 20 For both recordings we used an inter-electrode distance of two electrodes positions of the 10–20 system (F3(4)-P3(4) comparable to P3-P4), but the inter-electrode distance in two-channel recording in other studies is sometimes shorter than the distance used with one-channel recording, which can lead to differences in amplitude. Furthermore, one should note that the activity differences between frontal and parietal lobes leading to differences in amplitude between the longitudinal derivations P3-F3, P4-F4 and transverse derivation P3-P4, can result in differences in BGP.
Since the increased use of the continuous aEEG bedside monitoring in neonates there is concern about the accuracy in seizure detection. Previous studies in which analogue aEEGs were compared with conventional EEG showed that 80% of newborn infants with seizures were detected with aEEG and that only focal, low amplitude and very short periods of seizure patterns were missed.8 21 Other studies, looking at seizure patterns rather than infants, found poor sensitivity, specificity and predictive values for aEEG compared to (video)-EEG monitoring.10 A recent study by Shellhaas et al showed a mean seizure pattern detection rate of 40.3% of 125 records with seizures.12 They also mentioned that especially infrequent, brief seizures or seizures of low amplitude are difficult to detect. Over the last few years new digital aEEG equipment has been developed with the ability to record more than one channel, and with simultaneous display of the raw EEG signal and more recently also with access to an automatic seizure detection algorithm.13 With these improved techniques the accuracy of seizure detection would be expected to be better. A recent study by Shah et al showed that the combination of two-channel aEEG with the raw EEG signal detected the majority (78%) of electrical seizure patterns in at-risk newborn infants.11 They also compared the reading of the one-channel and two-channel aEEG without access to the raw EEG and found no substantial difference between those two recordings for seizure detection. Using two-channel aEEG alone was less accurate (44%) than using the combination of aEEG and raw EEG.
For our study we used the latest version of the offline Analyze viewer with access also to the raw single channel EEG, allowing us to confirm suspected discharges on one-channel aEEG by the crossover raw EEG.
An important observation in our study is the additional information of two-channel recording in infants with a unilateral brain lesion. In almost all infants more seizure patterns were seen on the affected side and in about 50% no seizure patterns were identified on the contralateral side. In our NICU population about 10% of full-term infants with NE or seizures had a predominantly unilateral lesion.22 When infants with hemiconvulsions were admitted to our NICU a two-channel recording is started. This study also demonstrates that when one-channel recording in these infants would have been used most of the seizures would have been identified as well. This can probably be explained by the fact that neonatal seizure patterns have a large tendency to shift from one area to another regardless of whether the pathological process is diffuse or more localised and they become widespread or even generalised.23,–,25
A limitation of this study is the restriction to the two-channel aEEG together with the raw EEG of these two channels, so no comparison was made with a complete neonatal montage conventional EEG. In view of the studies mentioned above we are aware that seizures may be missed and that we are not informed about the focal characteristics of the seizures. In all infants in this study seizure patterns that were identified on one channel and/or two channels were treated with anti-epileptic drugs. The ability of continuous aEEG monitoring for several days therefore appears to outweigh the detailed information obtained during a much shorter standard EEG.
Since the increased use of aEEG monitoring in NICUs, also by less experienced medical staff, there is also a need for a seizure detection algorithm. Several automatic seizure detection methods have been described in the literature, but most of them are not able to reliably detect neonatal seizures.13 26,–,30 The algorithm designed by Navakatikyan showed the most favourable results compared to the other methods, but the algorithm has only been tested in retrospective conventional EEG data.13 In 15 of our patients this method was used, and our sensitivity rate of 65% was similar to the 55% found by Lawrence et al31 and lower than that reported by Navakatikyan et al who found sensitivities ranging from 83% to 95% with positive predictive values of 48%–77%. This difference could be due to the short duration of many of the seizures in our infants, as it is well known that the detection is better with seizures of longer duration.13 32 Further clinical studies are needed to give more information about reliable detection with this method for clinical use with aEEG.
In our study most of the infants included were outborn and already showed clinical seizures before aEEG recording was started. A larger study is required in infants at risk for seizures to see if two channels or, ideally, more channels will give earlier information about seizure activity in these infants as neonatal seizures tend to have a focal onset even when the underlying pathological process is diffuse.
Although we can conclude that no major differences were found between seizure detection with one or two-channel aEEG, we were able to show that in the subgroup of infants with a predominantly unilateral brain lesion, a two-channel recording does provide additional information concerning lateralisation with identification of more seizure patterns on the affected side, sometimes also associated with a difference in background activity.
Routine use of a two-channel recording is therefore recommended as it may suggest the presence of a unilateral brain lesion, at a time that neuro-imaging is still awaited.
References
Footnotes
-
Competing interests None.
-
Ethics approval This study was conducted with the approval of the medical ethics committee, UMC Utrecht, The Netherlands.
-
Funding LvR was supported by the Dutch Epilepsy Foundation (NEF 3-15).
-
Provenance and peer review Not commissioned; externally peer reviewed.
Linked Articles
- Miscellanea