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

Abstract

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.

Figures

Figure 1

A. 5-min of burst suppression recorded from a patient following a propofol bolus. The 5-microvolt threshold used for the detection of the binary events is shown in red. B. The corresponding binary time-series where 1 represents a suppression and 0 represents a burst. C. 1-min segment taken from A. D. The binary time-series corresponding to C.

Figure 2

A. The EEG of an isoflurane anesthetized rat administered physostigmine to assess its arousal effects. At minute 10 (red vertical arrow) normal saline is injected. At approximately minute 16 (red star), physostigmine is injected and the EEG promptly switches from a burst suppression pattern to a dominant delta oscillations. B. The binary time-series associated with A. C. The BSP smoothing algorithm estimate (black curve), the BSR estimate computed using 15-s intervals with no overlap (green curve), and the BSR estimate with 14-s overlap (red curve). D. The BSP smoothing algorithm estimate (black curve), the BSR estimate computed using 60-s intervals with no overlap (green curve), and the BSR estimate with 59-s overlap (red curve).

Figure 3

A. BSP smoothing algorithm estimate (black curve) and its associated 95% confidence (Bayesian credibility) intervals (red curves) for minutes 2–5 in figure 2(C). B. BSR estimate computed with a 60-s interval and 59-s overlap (black curve) and its approximate 95% confidence interval based on the Gaussian approximation to the binomial (red curves) for minutes 2–5 in figure 2(C). C. Point-by-point comparison matrix evaluating the Pr (pj > pi) where pj , the y-axis corresponds to the BSP at time j and pi the x-axis, corresponds to the BSP at time for i, for 1 < j ≤ i. Red (black) means that at the given x, y pair Pr(pj > pi ) > 0.975 (Pr(pj > pi ) < 0.025). Gray means that 0.025 ≤ Pr(pj > pi) ≤ 0.975. The coordinate 20, 15 is red because p15 is substantially greater than p20. The coordinate 25, 20 is black because p25 and p20 are both near 0.

Figure 4

A. The EEG recorded in a patient undergoing controlled hypothermia. B. The binary time-series associated with A. C. The BSP filter estimate (black curve), the one-sided BSR estimate computed using 15-s intervals with no overlap (green curve), and the one-sided, 15-s BSR estimate with 14-s overlap (red curve). D. The BSP filter estimate (black curve), the one-sided BSR estimate computed using 60-s intervals with no overlap (green curve), and the one-sided, 60-s BSR estimate with 59-s overlap (red curve).