A comparison of quantitative EEG features for neonatal seizure detection

OBJECTIVE

This study was undertaken to identify the best performing quantitative EEG features for neonatal seizures detection from a test set of 21.

METHODS

Each feature was evaluated on 1-min, artefact-free segments of seizure and non-seizure neonatal EEG recordings. The potential utility of each feature for neonatal seizure detection was determined using receiver operating characteristic analysis and repeated measures t-tests. A performance estimate of the feature set was obtained using a cross-fold validation and combining all features together into a linear discriminant classifier model.

RESULTS

Significant differences between seizure and non-seizure segments were found in 19 features for 17 patients. The best performing features for this application were the RMS amplitude, the line length and the number of local maxima and minima. An estimate of the patient independent classifier performance yielded a sensitivity of 81.08% and specificity of 82.23%.

CONCLUSIONS

The individual performances of 21 quantitative EEG features in detecting electrographic seizure in the neonate were compared and numerically quantified. Combining all features together into a classifier model led to superior performance than that provided by any individual feature taken alone.

SIGNIFICANCE

The results documented in this study may provide a reference for the optimum quantitative EEG features to use in developing and enhancing neonatal seizure detection algorithms.

Spectral entropy as a monitor of depth of propofol induced sedation

OBJECTIVE

The aim of this prospective, observational study was to evaluate State and Response entropy (Entropy(TM) Monitor, GE Healthcare, Finland), indices as measures of moderate ("conscious") sedation in healthy adult patients receiving a low dose propofol infusion. Sedation was evaluated using: (I) the responsiveness component of the OAA/S scale (Observer's Assessment of Alertness/Sedation scale) and (II) multi-channel electroencephalogram (EEG) interpretation by a clinical expert.

METHODS

12 ASA I patients were recruited. A target-controlled infusion of propofol was administered (using Schnider's pharmacokinetic model) with an initial effect site concentration set to 0.5 microg ml(-1). A 4 minute equilibrium period was allowed. This concentration was increased at 4 minute intervals by 0.5 microg ml(-1) to a maximum of 2.0 microg ml(-1). State (SE) and Response (RE), entropy values were recorded for each 4 minute epoch together with clinical sedation scores (OAA/S) and continuous multi-channel EEG. The multi-channel EEG recorded during the final minute of each 4 minute epoch or "patient/time unit" was presented to a neurophysiologist who assigned a label "sedated/not sedated". SE/RE values were compared in patient/time units with clinical or EEG evidence of sedation versus those without.

RESULTS

Mean SE and RE values were less in patient/time units when clinical evidence of sedation was present, [mean = 86.8 (95% CI, 84.0-88.3) and 94.3 (95%CI, 92-96.1)], P = 0.002 and P = 0.001, respectively. In patient/time units assigned the label "sedated" by the clinical neurophysiologist assessing the multi-channel EEG, SE and RE values were less [mean = 87.5 (95% CI, 86.3-88.4) and 95.0 (95% CI, 93.8-96.1)] P = 0.001 and P < 0.001, respectively.

CONCLUSIONS

A statistically significant decrease in SE and RE values was demonstrated in patient/time units in which clinical or EEG evidence of sedation was present. We conclude that spectral entropy offers potential as a monitor of propofol induced sedation.

Seizure detection in neonates: Improved classification through supervised adaptation

The goal of neonatal seizure detection is the development of a patient independent system to alert staff in the neonatal intensive care unit of ongoing seizures. This study demonstrates the potential in adapting a patient independent classifier using patient specific data. Supervised adaptation is investigated using the basic gradient descent algorithm and least mean squares procedures. An increase in mean ROC area of 3% is obtained for the best performing learning algorithm, yielding an increase in mean accuracy of 7.7% compared to the patient independent algorithm.

Automated single channel seizure detection in the neonate

Neonatal seizures are the most common neurological emergency in the neonatal period and are associated with poor long-term outcome. EEG is considered the gold standard for identification of all neonatal seizures, reducing the number of EEG electrodes required would reduce patient handling and allow faster acquisition of data. A method for automated neonatal seizure detection based on two carefully chosen cerebral scalp electrodes but trained using multi-channel EEG is presented. The algorithm was developed and tested using a multi-channel EEG dataset containing 411 seizures from 251.9 hours of EEG recorded from 17 full-term neonates. Automated seizure detection using a variety of bipolar channel derivations was investigated. Channel C3-C4 yielded correct detection of 90.77% of seizures with a false detection rate of 9.43%. This compares favourably with a multi-channel seizure detection method which detected 81.03% of seizures with a false detection rate of 3.82%.

Combination of EEG and ECG for improved automatic neonatal seizure detection

OBJECTIVE

Neonatal seizures are the most common central nervous system disorder in newborn infants. A system that could automatically detect the presence of seizures in neonates would be a significant advance facilitating timely medical intervention.

METHODS

A novel method is proposed for the robust detection of neonatal seizures through the combination of simultaneously-recorded electroencephalogram (EEG) and electrocardiogram (ECG). A patient-specific and a patient-independent system are considered, employing statistical classifier models.

RESULTS

Results for the signals combined are compared to results for each signal individually. For the patient-specific system, 617 of 633 (97.52%) expert-labelled seizures were correctly detected with a false detection rate of 13.18%. For the patient-independent system, 516 of 633 (81.44%) expert-labelled seizures were correctly detected with a false detection rate of 28.57%.

CONCLUSIONS

A novel algorithm for neonatal seizure detection is proposed. The combination of an ECG-based classifier system with a novel multi-channel EEG-based classifier system has led to improved seizure detection performance. The algorithm was evaluated using a large data-set containing ECG and multi-channel EEG of realistic duration and quality.

SIGNIFICANCE

Analysis of simultaneously-recorded EEG and ECG represents a new approach in seizure detection research and the detection performance of the proposed system is a significant improvement on previous reported results for automated neonatal seizure detection.

Electrocardiogram Based Neonatal Seizure Detection

A method for the detection of seizures in the newborn using the electrocardiogram (ECG) signal is presented. Using a database of eight recordings, a method was developed for automatically annotating each 1-min epoch as "nonseizure" or "seizure". The system uses a linear discriminant classifier to process 41 heartbeat timing interval features. Performance assessment of the method showed that on a patient-specific basis an average accuracy of 70.5% was achieved in detecting seizures with associated sensitivity of 62.2% and specificity of 71.8%. On a patient-independent basis the average accuracy was 68.3% with sensitivity of 54.6% and specificity of 77.3%. Shifting the decision threshold for the patient-independent classifier allowed an increase in sensitivity to 78.4% at the expense of decreased specificity (51.6%), leading to increased false detections. The results of our ECG-based method are comparable with those reported for EEG-based neonatal seizure detection systems and offer the benefit of an easier acquisition methodology for seizure detection.

Automated estimation of sedation depth from the EEG

A method is presented for the automatic determination of a patient's level of sedation from the EEG. Six bipolar channels of EEG recorded from 12 adult patients sedated with low-dose propofol (2, 6-disopropylphenol) were used to develop a linear discriminant based system for depth of sedation monitoring using a number of quantitative EEG measures. A cross fold validation estimate of the performance of the algorithm as a patient independent system yielded a sensitivity of 74.70% and a specificity of 81.67%. It is hoped that the methodology reported here could lead to fully automated systems for depth of sedation monitoring.

Prediction of seizures in asphyxiated neonates: correlation with continuous video-electroencephalographic monitoring

BACKGROUND

After perinatal asphyxia, predicting which infants will develop significant hypoxic-ischemic encephalopathy and neonatal seizures remains a difficult task. High-risk markers (Apgar score, acidosis, nucleated red blood cells, and resuscitation) have been used to predict neonatal seizures with varying success. The "3 strikes" of Apgar score of <5 at 5 minutes, pH <7.00, and need for intubation have been cited as having a positive predictive value of 80%. We examined whether the predictive values of these markers would be increased if early continuous electroencephalographic monitoring allowed us to accurately identify all neonatal seizures and to grade the encephalopathy.

METHOD

We recruited term infants with perinatal asphyxia. Continuous video electroencephalography was commenced soon after birth and continued for 24 to 72 hours. The abilities of high-risk markers to predict electroencephalographic seizures, background electroencephalographic activity, and Sarnat grade were examined.

RESULTS

Forty-nine infants were suitable for analysis. Electrographic seizures occurred in 11 of the 49 infants. Encephalopathy was scored by using Sarnat grade (6, severe; 18, moderate; 25, mild) and electroencephalographic findings (4 inactive, 4 major abnormalities, 16 moderate abnormalities, and 25 normal/mildly abnormal). Apgar score of <5 at 5 minutes, pH <7.0, and the need for intubation had positive predictive values for neonatal seizures of 18%, 16%, and 21%, respectively. Combining these markers gave a positive predictive value of 25% and a negative predictive value of 77%. Substituting base deficit or lactate for pH in the 3-strikes model did not improve its predictive value. Apgar score of <5 at 5 minutes, nucleated red blood cells, and a base deficit less than -15 mEq/L showed some association with Sarnat grade. Only 5-minute Apgar score was significantly associated with both Sarnat grade and electroencephalographic grade.

CONCLUSION

After perinatal asphyxia, neither the condition at birth nor the degree of metabolic acidosis reliably predict neonatal seizures.

Early continuous video-EEG in acute near-total intrauterine asphyxia

The evolution of electroencephalographic changes after acute hypoxic-ischemic injury is poorly understood, as a clear time of insult is often absent and continuous electroencephalographic monitoring in the first 3 days after such injury has not been previously reported. Infants who suffer sudden profound asphyxia, often termed "acute near-total intrauterine asphyxia", have evidence of damage to the deep gray matter. In these infants it is possible to time the onset and duration of cerebral ischemia. This report describes early continuous video-electroencephalography from 3 hours after birth in an infant with the characteristic clinical and radiologic features of acute near-total intrauterine asphyxia.

The relationship between the onset of electrographic seizure activity after birth and the time of cerebral injury in utero

In the fetal lamb model of hypoxic-ischaemic injury, the insult is followed by EEG depression, after which seizures emerge at 7-13 hours. We explored the relationship between the emergence of electrographic seizures and our estimate of the time of the cerebral injury in nine babies who underwent continuous video-EEG monitoring from soon after birth. Babies with prelabour insults had their first seizures before 12 hours of age, whereas those whose insult was peripartum had seizure onset at 18-20 hours of age. EEG seizure onset time could have important clinical and medico-legal applications, and be related to the time or severity of the insult, or both.

Second-line anticonvulsant treatment of neonatal seizures: a video-EEG monitoring study

The authors conducted a randomized trial of second-line anticonvulsant treatments for neonates. The response to treatment was assessed using continuous video-EEG because the clinical diagnosis of seizure in neonates is known to be unreliable. Of 27 neonates with EEG-confirmed seizures, 5 were excluded because of protocol violations, and 11 responded to phenobarbitone in a dose of 40 mg/kg as first line. Three of five neonates treated with lignocaine responded. Six neonates were treated with benzodiazepines as second line: None responded, and their neurodevelopmental outcome was poor.

Non-expert use of the cerebral function monitor for neonatal seizure detection

BACKGROUND

The cerebral function monitor (CFM) is widely used to detect neonatal seizures, but there are very few studies comparing it with simultaneous electroencephalography (EEG).

OBJECTIVE

To determine the accuracy of non-expert use of the CFM and to assess interobserver agreement of CFM seizure detection.

PATIENTS

Babies admitted to the neonatal intensive care unit at King's College Hospital who were at high risk of seizure and had video-EEG monitoring.

METHODS

Video-EEG was used to detect seizures. Each baby had CFM recordings at speeds of 6, 15, and 30 cm/h during the EEG. Four neonatologists, trained in CFM seizure recognition, independently rated one hour CFM samples at three speeds from each baby. Interobserver agreement was quantified using Cohen's kappa.

RESULTS

CFM traces from 19 babies with EEG seizures and 21 babies without EEG seizures were analysed. Overall non-expert interpretation of the CFM performed poorly as a seizure detector compared with simultaneous EEG (sensitivities 38% at 6 cm/h; 54% at 15 cm/h; 55% at 30 cm/h). Although babies with seizures were more likely to be correctly classified at higher speeds (p = 0.02), babies without seizures were also more likely to be misclassified (p < 0.001). Agreement between observers was not good at any speed (kappa values from 0.01 to 0.39). The observers usually detected generalised seizures but often missed seizures that were focal, low amplitude, or lasted less than one minute.

CONCLUSION

Approximately half of all neonatal seizures may be missed using CFM alone. Neonatal seizures need to be diagnosed, characterised, and quantified first using EEG. The CFM may then be useful for long term monitoring.

Neonatal seizures and their treatment

PURPOSE OF REVIEW

Neonatal seizures continue to present a diagnostic and therapeutic challenge to paediatricians worldwide, and are a worrying sign for both parents and clinicians alike. The present review summarizes recent evidence regarding the diagnosis, aetiology and treatment of neonatal seizures. It is timely because there is new evidence that seizures are damaging to the neonatal brain, and because prolonged electroencephalographic recordings during treatment have provided information that challenges established treatment regimens.

RECENT FINDINGS

Neonatal seizures can permanently disrupt neuronal development, induce synaptic reorganization, alter plasticity and "prime" the brain to increased damage from seizures later in life. Phenobarbitone remains the mainstay of treatment and is effective in about one-third of cases; babies who respond tend to have a smaller seizure burden and a relatively normal background electroencephalogram. Their prognosis is better than in those who require second-line treatments. Phenytoin and lignocaine (membrane stabilizing drugs) are probably more effective than any of the benzodiazepines as second line, but very few evaluation studies have been reported. Babies who require second-line treatments are more likely to have hypoxic ischaemic encephalopathy, an abnormal background electroencephalogram and a large seizure burden, and have a worse prognosis than do those who respond to a single agent; most have significant disability at follow up.

SUMMARY

The search for an effective antiepileptic regimen in the newborn must continue. Whether better control of neonatal seizures leads to a reduction in neurodisability in childhood cannot be determined until more effective treatments are found. Meanwhile, electroencephalography remains the most useful investigation for diagnosis and prognosis.

Phenobarbitone, neonatal seizures, and video-EEG

AIMS

To evaluate the effectiveness of phenobarbitone as an anticonvulsant in neonates.

METHODS

An observational study using video-EEG telemetry. Video-EEG was obtained before treatment was started, for an hour after treatment was given, two hours after treatment was given, and again between 12 and 24 hours after treatment was given. Patients were recruited from all babies who required phenobarbitone (20-40 mg/kg intravenously over 20 minutes) for suspected clinical seizures and had EEG monitoring one hour before and up to 24 hours after the initial dose. An EEG seizure discharge was defined as a sudden repetitive stereotyped discharge lasting for at least 10 seconds. Neonatal status epilepticus was defined as continuous seizure activity for at least 30 minutes. Seizures were categorised as EEG seizure discharges only (electrographic), or as EEG seizure discharges with accompanying clinical manifestations (electroclinical). Surviving babies were assessed at one year using the Griffiths neurodevelopmental score.

RESULTS

Fourteen babies were studied. Four responded to phenobarbitone; these had normal or moderately abnormal EEG background abnormalities and outcome was good. In the other 10 babies electrographic seizures increased after treatment, whereas electroclinical seizures reduced. Three babies were treated with second line anticonvulsants, of whom two responded. One of these had a normal neurodevelopmental score at one year, but the outcome for the remainder of the whole group was poor.

CONCLUSION

Phenobarbitone is often ineffective as a first line anticonvulsant in neonates with seizures in whom the background EEG is significantly abnormal.

An unusual presentation of fifth-day fits in the newborn

Fifth-day fits are a rare type of seizures in the infant. Tonic-clonic seizures have not been reported as a manifestation of this condition. This article presents a report of one patient and the value of video-electroencephalogram recording in assessing seizures, background electroencephalogram, and treatment. We describe a full-term female infant with fifth-day fits presenting with generalized tonic-clonic seizures associated with generalized electrical discharges, with normal background electroencephalogram activity between seizures. The fits rapidly responded to phenobarbital.

Early serial EEG in hypoxic ischaemic encephalopathy

OBJECTIVES

To perform early serial EEGs in infants with hypoxic ischaemic encephalopathy (HIE) and compare the findings with neurodevelopmental outcome.

METHODS

Nine full-term neonates with HIE had simultaneous video-EEG polygraphic studies within 8 h of birth. The EEG was repeated at 12-24 h intervals. All surviving infants had a neurodevelopmental assessment at 1 year.

RESULTS

Two infants had a normal or mildly abnormal EEG within 8 h of birth and neurodevelopmental outcome was normal. Seven infants had severely depressed background activity in the first 8 h of life. In 3 infants the EEG activity recovered within 12-24 h showing continuous activity with no or only minor abnormalities. All these infants had a normal outcome. The remaining 4 infants, who also had an initially inactive recording, subsequently developed severe background abnormalities. At follow-up, two infants had died and the remainder developed major neurological sequelae.

CONCLUSIONS

Early EEG is an excellent prognostic indicator for a favourable outcome if normal within the first 8 h of life and for a poor outcome if the background activity continues to be inactive or grossly abnormal beyond 8-12 h of life. However, an inactive or very depressed EEG within the first 8 h of life can be associated with good outcome if the EEG activity recovers within 12 h.

Dynamic cerebral autoregulation in sick newborn infants

The sick newborn infant is vulnerable to brain injury and impaired cerebral autoregulation is thought to contribute to this. Coherent averaging is a method of measuring the dynamic cerebral autoregulatory response that is particularly suitable for neonates. We used this method in combination with a measure of the gradient of the cerebral blood flow velocity (CBFV) response following transient blood pressure (BP) peaks to study dynamic autoregulation in infants undergoing intensive care. Term and preterm infants at high risk of neurologic injury were compared with a control group of infants, also undergoing intensive care. Simultaneous video-EEG, CBFV (using transcranial Doppler), and arterial blood pressure measurements were obtained intermittently during a study period of at least 2 h. Cerebral autoregulatory response curves were constructed for high risk and control groups. Intact cerebral autoregulation produces a characteristic response consisting of a brief period when CBFV follows arterial blood pressure but quickly returns to baseline value. An impaired autoregulatory response shows CBFV mirroring the arterial blood pressure curve closely. Thirteen high-risk infants, who also had seizures (10 term and 3 preterm) and 12 control infants (6 term and 6 preterm) were studied. Autoregulation was absent in high-risk term and preterm infants. It was also absent in preterm control infants. Term, neurologically healthy infants undergoing intensive care have an intact autoregulatory response. The constant passive response seen in high-risk infants may reflect the severity of the underlying neurologic disease.

Outcome of electroclinical, electrographic, and clinical seizures in the newborn infant

Three seizure types have been described in the neonate: electroclinical, electrographic, and clinical only. Controversy still exists about whether the episodic abnormal movements seen in some infants, which are not accompanied by simultaneous ictal discharges on the EEG, are true seizures. Twenty-four infants with seizures were studied, 17 had purely electrographic and/or electroclinical seizures, seven had clinical-only seizures; six of these seven had clonic seizures, without facial manifestations or autonomic change. The three seizure types were investigated using video-EEG and a Griffiths neurodevelopmental assessment was performed in each seizure group. Of the seven infants with clinical-only seizures, six had clonic seizures with a normal background EEG, neuroimaging studies and neurodevelopmental follow-up assessment were normal in five. In the remaining 17 infants with electrographic and/or electroclinical seizures, seizure discharges were often associated with ocular phenomena, apnoea, or tonic posturing, and the background EEG was abnormal in all but one subject. Neurodevelopmental follow-up assessments revealed a poor outcome (14 of 17) in this group. In otherwise healthy infants, purely clonic seizures involving only the limbs may be a benign phenomenon and an EEG should be obtained to avoid unnecessary treatment. Infants with seizures superimposed on an abnormal background EEG pattern had a poor outcome.

Cerebral blood flow velocity during neonatal seizures

AIM

To determine if cerebral blood flow velocity increases during all types of neonatal seizure, and whether the effect is due solely to an increase in blood pressure, transmitted to the cerebral circulation when autoregulation is impaired.

METHODS

Seizures were diagnosed in 11 high risk neonates using cotside 16 channel video-EEG polygraphy. EEG, cerebral blood flow velocity (CBFV) using transcranial Doppler ultrasound, and arterial blood pressure (ABP) measurements were made. At least two 5-10 minute epochs of simultaneous measurements were performed on each infant. These epochs were then reviewed to eliminate artefacts, and one minute data periods containing a clear seizure onset were created. Each period contained 20 seconds before the seizure. Data periods without seizures from the same infants were also analysed and compared with seizure periods.

RESULTS

Four infants had purely electrographic seizures-without clinical manifestations. Six infants had electroclinical seizures. One infant displayed both seizure types. A random effects linear regression analysis was used to determine the effect of seizures on CBFV and ABP. A significant increase was found in mean CBFV in those periods containing seizures. The mean percentage change in velocity for all infants was 15.6%. Three infants showed a significant increase in mean ABP after seizures but the overall increase in ABP for all infants was not significant.

CONCLUSION

Electroclinical and electrographic neonatal seizures produce an increase in CBFV. In some infants the increase is not associated with an increase in blood pressure. These preliminary results suggest that electrographic seizures are associated with disturbed cerebral metabolism. Treatment of neonatal seizures until electrographic seizure activity is abolished may improve outcome for these infants.