Passive functional mapping of receptive language cortex during general anesthesia using electrocorticography

Abstract

Objective: To investigate the feasibility of passive functional mapping in the receptive language cortex during general anesthesia using electrocorticographic (ECoG) signals.

Methods: We used subdurally placed ECoG grids to record cortical responses to speech stimuli during awake and anesthesia conditions. We identified the cortical areas with significant responses to the stimuli using the spectro-temporal consistency of the brain signal in the broadband gamma (BBG) frequency band (70-170 Hz).

Results: We found that ECoG BBG responses during general anesthesia effectively identify cortical regions associated with receptive language function. Our analyses demonstrated that the ability to identify receptive language cortex varies across different states and depths of anesthesia. We confirmed these results by comparing them to receptive language areas identified during the awake condition. Quantification of these results demonstrated an average sensitivity and specificity of passive language mapping during general anesthesia to be 49±7.7% and 100%, respectively.

Conclusion: Our results demonstrate that mapping receptive language cortex in patients during general anesthesia is feasible.

Significance: Our proposed protocol could greatly expand the population of patients that can benefit from passive language mapping techniques, and could eliminate the risks associated with electrocortical stimulation during an awake craniotomy.

Keywords: Awake craniotomy; Electrocorticography (ECoG); General anesthesia; Passive functional mapping; Receptive language cortex.

Figure 1:. Anesthesia and experimental timeline.

(A) The typical profile of anesthesia (time axis not to scale) consists of three phases, i.e., induction, maintenance, and emergence. During these phases, patients cycle through different hypnotic states, starting from awake (blue) to deep anesthesia (red). While sedated, patients experience reduced awareness yet maintain some level of responsiveness to (painful) stimuli. General anesthesia is a state that is recognized by loss of consciousness (LOC) which is indicated by the horizontal dashed line in this diagram. General anesthesia can be separated into three states, i.e., light, intermediate, and deep anesthesia (deep anesthesia is characterized by burst suppression and isoelectric behavior). (B) The experimental timelines for the two cohorts: Epilepsy (top) and Tumor (bottom). Arrows along the timeline indicate the sequence of events (time axis not to scale) that patients experienced in each cohort. Similar to panel A, the gradient color below the timeline illustrates the transition between anesthesia states, and the green line demarcates the interval during which functional mapping was executed. (ECS: electrocortical stimulation, ECoG: electrocorticography)

Figure 2:. Localizing electrode locations to a patient-specific, 3D brain model.

(A) Preoperative MRI (magnetic resonance imaging) scan. (B) ECoG (electrocorticography) grid implantation. (C) Post-implantation x-ray. (D) 3D brain model with coregistered electrodes (black circles).

Figure 3:. The dynamics of ECoG signals throughout different hypnotic states.

(A) The blue, red, and green time traces illustrate a representative raw ECoG (electrocorticography) signal (recorded from the electrode indicated with a star on the brain model – all time traces are normalized to the “awake” condition) during the awake condition and increasing states of general anesthesia. The bottom diagram in panel A shows the average spectrogram across all electrodes in this example subject. The first vertical dashed line indicates loss of consciousness (LOC), separating awake and general anesthesia. The second dashed line separates the light-intermediate general anesthesia from the deep states of general anesthesia (burst suppression). During the light-intermediate states, the signal is smaller in amplitude compared to the awake condition. Also, a shift toward slower oscillations is discernible in the spectral domain, which is expected of the signal during these states of general anesthesia. At the same time, the sharp transient changes (nonstationarity) observed during deep anesthesia, which are evident in both the temporal and spectral domains, are a well-established biomarker of a deep state of anesthesia (burst suppression). (B) Normalized broadband gamma (BBG) envelope during the awake condition (blue) as well as during different states of general anesthesia (red and green). The diagram depicts the BBG envelope during three representative trials and the average BBG envelope across all trials (shaded area indicates standard error). Each trial lasts 1.7 s and is comprised of a 1 s silent period (baseline: i.e., lack of stimulus) followed by a 0.7 s auditory stimulus (shaded rectangle). The smaller BBG response during general anesthesia represents a reduced signal-to-noise ratio (SNR) of the brain signals. (C) SNR (average BBG envelope variance across all trials) during the awake, light-intermediate, and deep anesthesia conditions for the electrode indicated with a star on the brain model. The asterisks in the bar graph indicate a significant reduction in the average variance of the BBG envelope (rank-sum test; *p < 0.01 and * * p < 0.001; error bar indicates standard error). As the depth of anesthesia increases, the variance of the BBG envelope (response magnitude, i.e., effect size) and subsequently SNR decreases.

Figure 4:. Statistical analysis.

Illustration of the statistical analysis for representative trials (recorded from the electrode indicated with a star on the brain model in Figure 3). (A) The blue and cyan time traces are the z-scored broadband gamma (BBG) envelope during “baseline” and “task” periods. The number next to each time trace indicates its corresponding trial number (1 to N). The stimulus onset is at T=0 ms. The “baseline” period is a 600 ms interval preceding the stimulus onset from T=−600 to T=0 ms, and the “task” period is a 600 ms interval from T=100 to T=700 ms following the stimulus onset. (B) The correlation analysis between all the combinations (without repetition) of “baseline” and “task” pairs. Each panel presents a pair of time traces corresponding to different trials and their Pearson correlation coefficient (r-value). (C) Distribution of baseline and task r-values. The Wilcoxon rank-sum test was used to determine whether the distributions represent different populations. (D) Receiver Operating Characteristic (ROC) curve associated with the two distributions in panel C. The area under the curve (AUC) indicates the separability of the two distributions. We defined the summation of the AUC of all responsive electrodes as the brain activity index (BAI).

Figure 5:. Passive mapping results for the Epilepsy cohort.

Black circles indicate the electrode locations. Colored circles represent electrodes with a statistically significant response to the auditory stimuli during awake (blue) and anesthesia (red) conditions. The number in the lower right corner of each map indicates the brain activity index (BAI), normalized to the awake condition for each patient.

Figure 6:. Passive mapping results for the Tumor cohort.

Black circles indicate electrode locations. Colored circles represent electrodes with a statistically significant response to the auditory stimuli during awake (blue) and anesthesia (red) conditions. The number in the lower right corner of each map indicates the brain activity index (BAI), normalized to the awake condition for each patient.

Figure 7:. Within-group averaged brain activity index (BAI).

The bar graph illustrates the average BAI values during general anesthesia in each cohort (reported in Figures 5 and 6). The average BAI is 37±5.2% and 60±8.3% for the Epilepsy and Tumor cohorts, respectively (averaged across subjects ± standard error). All values are normalized to the awake condition, and the difference between the two cohorts is not statistically significant (Wilcoxon rank-sum test; p = 0.02). This analysis suggests that passive language mapping is equally feasible during the induction and emergence phases of anesthesia.