Adaptive segmentation of burst-suppression pattern in isoflurane and enflurane anesthesia

Detection of EEG burst-suppression in neurocritical care patients using an unsupervised machine learning algorithm

OBJECTIVE

The burst suppression pattern in clinical electroencephalographic (EEG) recordings is an important diagnostic tool because of its association with comas of various etiologies, as with hypoxia, drug related intoxication or deep anesthesia. The detection of bursts and the calculation of burst/suppression ratio are often used to monitor the level of anesthesia during treatment of status epilepticus. However, manual counting of bursts is a laborious process open to inter-rater variation and motivates a need for automatic detection.

METHODS

We describe a novel unsupervised learning algorithm that detects bursts in EEG and generates burst-per-minute estimates for the purpose of monitoring sedation level in an intensive care unit (ICU). We validated the algorithm on 29 hours of burst annotated EEG data from 29 patients suffering from status epilepticus and hemorrhage.

RESULTS

We report competitive results in comparison to neural networks learned via supervised learning. The mean absolute error (SD) in bursts per minute was 0.93 (1.38).

CONCLUSION

We present a novel burst suppression detection algorithm that adapts to each patient individually, reports bursts-per-minute quickly, and does not require manual fine-tuning unlike previous approaches to burst-suppression pattern detection.

SIGNIFICANCE

Our algorithm for automatic burst suppression quantification can greatly reduce manual oversight in depth of sedation monitoring.

Substance-Specific Differences in Human Electroencephalographic Burst Suppression Patterns

Different anesthetic agents induce burst suppression in the electroencephalogram (EEG) at very deep levels of general anesthesia. EEG burst suppression has been identified to be a risk factor for postoperative delirium (POD). EEG based automated detection algorithms are used to detect burst suppression patterns during general anesthesia and a burst suppression ratio (BSR) is calculated. Unfortunately, applied algorithms do not give information as precisely as suggested, often resulting in an underestimation of the patients’ burst suppression level. Additional knowledge of substance-specific burst suppression patterns could be of great importance to improve the ability of EEG based monitors to detect burst suppression. In a re-analysis of EEG recordings obtained from a previous study, we analyzed EEG data of 45 patients undergoing elective surgery under general anesthesia. The patients were anesthetized with sevoflurane, isoflurane or propofol (n = 15, for each group). After skin incision, the used agent was titrated to a level when burst suppression occurred. In a visual analysis of the EEG, blinded to the used anesthetic agent, we included the first distinct burst in our analysis. To avoid bias through changing EEG dynamics throughout the burst, we only focused on the first 2 s of the burst. These episodes were analyzed using the power spectral density (PSD) and normalized PSD, the absolute burst amplitude and absolute burst slope, as well as permutation entropy (PeEn). Our results show significant substance-specific differences in the architecture of the burst. Volatile-induced bursts showed higher burst amplitudes and higher burst power. Propofol-induced bursts had significantly higher relative power in the EEG alpha-range. Further, isoflurane-induced bursts had the steepest burst slopes. We can present the first systematic comparison of substance-specific burst characteristics during anesthesia. Previous observations, mostly derived from animal studies, pointing out the substance-specific differences in bursting behavior, concur with our findings. Our findings of substance-specific EEG characteristics can provide information to help improve automated burst suppression detection in monitoring devices. More specific detection of burst suppression may be helpful to reduce excessive EEG effects of anesthesia and therefore the incidence of adverse outcomes such as POD.

Novel Burst Suppression Segmentation in the Joint Time-Frequency Domain for EEG in Treatment of Status Epilepticus

We developed a method to distinguish bursts and suppressions for EEG burst suppression from the treatments of status epilepticus, employing the joint time-frequency domain. We obtained the feature used in the proposed method from the joint use of the time and frequency domains, and we estimated the decision as to whether the measured EEG was a burst segment or suppression segment by the maximum likelihood estimation. We evaluated the performance of the proposed method in terms of its accordance with the visual scores and estimation of the burst suppression ratio. The accuracy was higher than the sole use of the time or frequency domains, as well as conventional methods conducted in the time domain. In addition, probabilistic modeling provided a more simplified optimization than conventional methods. Burst suppression quantification necessitated precise burst suppression segmentation with an easy optimization; therefore, the excellent discrimination and the easy optimization of burst suppression by the proposed method appear to be beneficial.

Monitoring burst suppression in critically ill patients: Multi-centric evaluation of a novel method

OBJECTIVE

To develop a computational method to detect and quantify burst suppression patterns (BSP) in the EEGs of critical care patients. A multi-center validation study was performed to assess the detection performance of the method.

METHODS

The fully automatic method scans the EEG for discontinuous patterns and shows detected BSP and quantitative information on a trending display in real-time. The method is designed to work without setting any patient specific parameters and to be insensitive to EEG artifacts and periodic patterns. For validation a total of 3982 h of EEG from 88 patients were analyzed from three centers. Each EEG was annotated by two reviewers to assess the detection performance and the inter-rater agreement.

RESULTS

Average inter-rater agreement between pairs of reviewers was κ=0.69. On average 22% of the review segments included BSP. An average sensitivity of 90% and a specificity of 84% were measured on the consensus annotations of two reviewers. More than 95% of the periodic patterns in the EEGs were correctly suppressed.

CONCLUSION

A fully automatic method to detect burst suppression patterns was assessed in a multi-center study. The method showed high sensitivity and specificity.

SIGNIFICANCE

Clinically applicable burst suppression detection method validated in a large multi-center study.

Quantification and automatized adaptive detection of in vivo and in vitro neuronal bursts based on signal complexity

In this paper, we propose employing entropy values to quantify action potential bursts in electrophysiological measurements from the brain and neuronal cultures. Conventionally in the electrophysiological signal analysis, bursts are quantified by means of conventional measures such as their durations, and number of spikes in bursts. Here our main aim is to device metrics for burst quantification to provide for enhanced burst characterization. Entropy is a widely employed measure to quantify regularity/complexity of time series. Specifically, we investigate the applicability and differences of spectral entropy and sample entropy in the quantification of bursts in in vivo rat hippocampal measurements and in in vitro dissociated rat cortical cell culture measurement done with microelectrode arrays. For the task, an automatized and adaptive burst detection method is also utilized. Whereas the employed metrics are known from other applications, they are rarely employed in the assessment of burst in electrophysiological field potential measurements. Our results show that the proposed metrics are potential for the task at hand.

Classification of EEG bursts in deep sevoflurane, desflurane and isoflurane anesthesia using AR-modeling and entropy measures

A study relating signal patterns of burst onsets in burst suppression EEG to the anesthetic agent or anesthesia induction protocol is presented. A dataset of 82 recordings of sevoflurane, isoflurane and desflurane anesthesia underlies the study. 3 second segments from the onset of altogether 3214 bursts are described using AR model parameters, spectral entropy and sample entropy as features. The features are clustered using the K-means algorithm. The results indicate that no clear cut distinction can be made between the burst patterns induced by the mentioned anesthetics although bursts of certain properties are more common in certain patient groups. Several directions for further investigations are proposed based on visual inspection of the recordings.

Propofol and sevoflurane induce distinct burst suppression patterns in rats

Burst suppression is an EEG pattern characterized by alternating periods of high-amplitude activity (bursts) and relatively low amplitude activity (suppressions). Burst suppression can arise from several different pathological conditions, as well as from general anesthesia. Here we review current algorithms that are used to quantify burst suppression, its various etiologies, and possible underlying mechanisms. We then review clinical applications of anesthetic-induced burst suppression. Finally, we report the results of our new study showing clear electrophysiological differences in burst suppression patterns induced by two common general anesthetics, sevoflurane and propofol. Our data suggest that the circuit mechanisms that generate burst suppression activity may differ among general anesthetics.

Burst suppression probability algorithms: state-space methods for tracking EEG burst suppression

OBJECTIVE

Burst suppression is an electroencephalogram pattern in which bursts of electrical activity alternate with an isoelectric state. This pattern is commonly seen in states of severely reduced brain activity such as profound general anesthesia, anoxic brain injuries, hypothermia and certain developmental disorders. Devising accurate, reliable ways to quantify burst suppression is an important clinical and research problem. Although thresholding and segmentation algorithms readily identify burst suppression periods, analysis algorithms require long intervals of data to characterize burst suppression at a given time and provide no framework for statistical inference.

APPROACH

We introduce the concept of the burst suppression probability (BSP) to define the brain's instantaneous propensity of being in the suppressed state. To conduct dynamic analyses of burst suppression we propose a state-space model in which the observation process is a binomial model and the state equation is a Gaussian random walk. We estimate the model using an approximate expectation maximization algorithm and illustrate its application in the analysis of rodent burst suppression recordings under general anesthesia and a patient during induction of controlled hypothermia.

MAIN RESULT

The BSP algorithms track burst suppression on a second-to-second time scale, and make possible formal statistical comparisons of burst suppression at different times.

SIGNIFICANCE

The state-space approach suggests a principled and informative way to analyze burst suppression that can be used to monitor, and eventually to control, the brain states of patients in the operating room and in the intensive care unit.

Real-time segmentation of burst suppression patterns in critical care EEG monitoring

OBJECTIVE

Develop a real-time algorithm to automatically discriminate suppressions from non-suppressions (bursts) in electroencephalograms of critically ill adult patients.

METHODS

A real-time method for segmenting adult ICU EEG data into bursts and suppressions is presented based on thresholding local voltage variance. Results are validated against manual segmentations by two experienced human electroencephalographers. We compare inter-rater agreement between manual EEG segmentations by experts with inter-rater agreement between human vs automatic segmentations, and investigate the robustness of segmentation quality to variations in algorithm parameter settings. We further compare the results of using these segmentations as input for calculating the burst suppression probability (BSP), a continuous measure of depth-of-suppression.

RESULTS

Automated segmentation was comparable to manual segmentation, i.e. algorithm-vs-human agreement was comparable to human-vs-human agreement, as judged by comparing raw EEG segmentations or the derived BSP signals. Results were robust to modest variations in algorithm parameter settings.

CONCLUSIONS

Our automated method satisfactorily segments burst suppression data across a wide range adult ICU EEG patterns. Performance is comparable to or exceeds that of manual segmentation by human electroencephalographers.

SIGNIFICANCE

Automated segmentation of burst suppression EEG patterns is an essential component of quantitative brain activity monitoring in critically ill and anesthetized adults. The segmentations produced by our algorithm provide a basis for accurate tracking of suppression depth.

An effective correlation dimension and burst suppression ratio of the EEG in rat. Correlation with sevoflurane induced anaesthetic depth

BACKGROUND AND OBJECTIVE

Anaesthesiologists need parameters that measure the depth of anaesthesia. In the context of this need, the present study investigated in rats how two variables from the electroencephalogram, the burst suppression ratio and effective correlation dimension correlated with a measure of anaesthetic depth as measured in the strength of a noxious withdrawal reflex.

METHODS

Eight rats were exposed to different inspiratory concentrations of sevoflurane, each rat in two separate experiments. In the first experiment, spontaneously breathing animals could move freely and no painful stimuli were applied. In the second experiment, in mechanically ventilated restrained anaesthetized rats, the withdrawal reflex was measured every 80 s. In both experiments the electroencephalogram was continuously recorded. The concentration in the effector compartment was estimated using a first order two compartment model. Correlation dimension was computed following the Grassberger/Procaccia/Takens approach with optimized parameter settings to achieve maximum sensitivity to anaesthetic drug effects and enable real-time computation. The Hill, equation was fitted to the data, describing the effect as a function of sevoflurane concentration.

RESULTS

Good correlations of Depth of Anaesthesia with correlation dimension as well as burst suppression ratio were established in both types of experiments. Arousal by noxious stimuli decreased burst suppression ratio and increased correlation dimension. The effective sevoflurane concentration associated with 50% of the maximum effect (C50) was higher in experiment II (stimulation) than in experiment I (no stimulation): i.e. for correlation dimension 2.18% vs. 0.60% and for burst suppression ratio 3.07% vs. 1.73%. The slope factors were: gammaCD = 4.15 vs. gammaCD = 1.73 and gammaBSR = 5.2 vs. gammaBSR = 5.4. Correlation dimension and burst suppression ratio both correlated with the strength of the withdrawal reflex with correlation coefficients of 0.46 and 0.66 respectively (P < 0.001).

CONCLUSIONS

Both correlation dimension and burst suppression ratio are related to anaesthetic depth and are affected by noxious stimuli. The relationship between anaesthetic depth and burst suppression ratio is confirmed and the potential of correlation dimension is demonstrated.

Automatic analysis and monitoring of burst suppression in anesthesia

OBJECTIVE

We studied the spectral characteristics of the EEG burst suppression patterns (BSP) of two intravenous anesthetics, propofol and thiopental. Based on the obtained results, we developed a method for automatic segmentation, classification and compact presentation of burst suppression patterns.

METHODS

The spectral analysis was performed with the short time Fourier transform and with autoregressive modeling to provide information of frequency contents of bursts. This information was used when designing appropriate filters for segmentation algorithms. The adaptive segmentation was carried out using two different nonparametric methods. The first one was based on the absolute values of amplitudes and is referred to as the ADIF method. The second method used the absolute values of the Nonlinear Energy Operator (NLEO) and is referred to as the NLEO method. Both methods have been described earlier but they were modified for the purposes of BSP detection. The signal was classified to bursts, suppressions and artifacts. Automatic classification was compared with manual classification. Results. The NLEO method was more accurate, especially in the detection of artifacts. NLEO method classified correctly 94.0% of the propofol data and 92.8% of the thiopental data. With the ADIF method, the results were 90.5% and 88.1% respectively.

CONCLUSIONS

Our results show that burst suppression caused by the different anesthetics can be reliably detected with our segmentation and classification methods. The analysis of normal and pathological EEG, however, should include information of the anesthetic used. Knowledge of the normal variation of the EEG is necessary in order to detect the abnormal BSP of, for instance, seizure patients.