Skip to main content
NCBI home page
Neurology logoLink to Neurology
. 2022 Mar 22;98(12):e1238–e1247. doi: 10.1212/WNL.0000000000013301

Regional Distribution of Brain Injury After Cardiac Arrest

Clinical and Electrographic Correlates

Samuel B Snider 1,✉, David Fischer 1, Morgan E McKeown 1, Alexander Li Cohen 1, Frederic LWVJ Schaper 1, Edilberto Amorim 1, Michael D Fox 1, Benjamin Scirica 1, Matthew B Bevers 1,*, Jong Woo Lee 1,*
PMCID: PMC8967331  PMID: 35017304

Abstract

Background and Objectives

Disorders of consciousness, EEG background suppression, and epileptic seizures are associated with poor outcome after cardiac arrest. Our objective was to identify the distribution of diffusion MRI–measured anoxic brain injury after cardiac arrest and to define the regional correlates of disorders of consciousness, EEG background suppression, and seizures.

Methods

We analyzed patients from a single-center database of unresponsive patients who underwent diffusion MRI after cardiac arrest (n = 204). We classified each patient according to recovery of consciousness (command following) before discharge, the most continuous EEG background (burst suppression vs continuous), and the presence or absence of seizures. Anoxic brain injury was measured with the apparent diffusion coefficient (ADC) signal. We identified ADC abnormalities relative to controls without cardiac arrest (n = 48) and used voxel lesion symptom mapping to identify regional associations with disorders of consciousness, EEG background suppression, and seizures. We then used a bootstrapped lasso regression procedure to identify robust, multivariate regional associations with each outcome variable. Last, using area under receiver operating characteristic curves, we then compared the classification ability of the strongest regional associations to that of brain-wide summary measures.

Results

Compared to controls, patients with cardiac arrest demonstrated ADC signal reduction that was most significant in the occipital lobes. Disorders of consciousness were associated with reduced ADC most prominently in the occipital lobes but also in deep structures. Regional injury more accurately classified patients with disorders of consciousness than whole-brain injury. Background suppression mapped to a similar set of brain regions, but regional injury could no better classify patients than whole-brain measures. Seizures were less common in patients with more severe anoxic injury, particularly in those with injury to the lateral temporal white matter.

Discussion

Anoxic brain injury was most prevalent in posterior cerebral regions, and this regional pattern of injury was a better predictor of disorders of consciousness than whole-brain injury measures. EEG background suppression lacked a specific regional association, but patients with injury to the temporal lobe were less likely to have seizures. Regional patterns of anoxic brain injury are relevant to the clinical and electrographic sequelae of cardiac arrest and may hold importance for prognosis.

Classification of Evidence

This study provides Class IV evidence that disorders of consciousness after cardiac arrest are associated with widely lower ADC values on diffusion MRI and are most strongly associated with reductions in occipital ADC.

Neuropathologic studies in animals1 and humans2,3 have identified numerous brain regions susceptible to anoxic injury, but a map of brain regions affected by cardiac arrest does not yet exist. Furthermore, although disorders of consciousness (DoC),4,5 EEG background suppression,6,7 and seizures7-10 predict poor neurologic outcomes after cardiac arrest,3,11,12 the regional patterns of anoxic brain injury (ABI) associated with each are unknown.3,13

Diffusion MRI is the most sensitive imaging modality for the clinical characterization of ABI.14-19 In animal models of ABI, reductions in the apparent diffusion coefficient (ADC) are associated with the histopathologic volume of infarcted brain tissue.20 Previous work has demonstrated that greater ADC reductions across the whole brain or cortex are associated with poorer outcomes.21

It remains unclear whether common clinical and electrographic abnormalities after cardiac arrest such as DoC, a suppressed EEG background, and postanoxic seizures reflect the overall severity of ABI, as measured with whole-brain or whole-cortex ADC values, or injury to specific structures.13 Investigating the associations between regional anoxic injury and the clinical and electrographic sequalae of cardiac arrest may help identify specific brain circuits responsible for their production. Furthermore, given an increasingly multimodal approach to prognostication after cardiac arrest,22-24 an improved understanding of the associations between modalities will facilitate interpretation of discrepant results.

Here, we use diffusion MRI from a large clinical cohort to define the anatomic distribution of ABI after cardiac arrest and to identify associations between patterns of injury and common clinical and electrographic abnormalities. Specifically, we sought to address whether global or regional diffusion MRI–measured ADC values showed a stronger association with DoC, suppression of the EEG background, and epileptic seizures.

Methods

Standard Protocol Approvals, Registrations, and Patient Consents

This study was approved by the Mass General Brigham institutional review board under protocols 2014P001623 and 2021P000633. The requirement for participant consent was waived.

Cardiac Arrest Cohort

We identified patients from a prospectively maintained registry of 573 patients with in-hospital or out-of-hospital cardiac arrest at Brigham and Women's Hospital between 2009 and 2020. Criteria for inclusion in the registry included age >18 years and the absence of command following on the first assessment after return of spontaneous circulation. Of these patients, 206 (36%) had a diffusion-weighted MRI within 14 days of cardiac arrest, with 204 completing the processing pipeline outlined below. While MRIs were not acquired on the basis of a prospectively defined set of indications, at our institution, patients who do not follow commands in the first 72 hours after arrest and are stable for transportation and lying fully supine commonly undergo MRI.

Clinical Covariates

Basic demographic information, cardiac rhythm, the use of targeted temperature management, and time between cardiac arrest and MRI acquisition were abstracted from the electronic medical record (Table 1). Targeted temperature management was performed in patients who were unresponsive after resuscitation in accordance with a hospital-wide protocol. Core temperature was measured with a bladder probe, and a target of 33°C was achieved with cooling pads for a period of 24 hours. In the presence of a standard contraindication to this lower temperature target (hemodynamic instability, hemorrhage), a target of 36°C was instead used. Patients with DoC were defined as those who never had documentation in the medical record of command following after cardiac arrest. Patients with missing data were excluded from the relevant analysis.

Table 1.

Cohort Characteristics

graphic file with name NEUROLOGY2021174355t1.jpg

Open in a new tab

EEG Background Continuity

One hundred ninety-three patients also had continuous (n = 186) or routine (n = 7) EEG data. EEG data were recorded with a Natus XLTEK system (Pleasanton, CA) according to the international 10-20 system and interpreted with the use of the American Clinical Neurophysiology Society critical care EEG terminology.25 Beginning in February 2013, background continuity of each epoch was prospectively categorized26 by a board-certified epileptologist as continuous, nearly continuous (suppression [<10 μV] <10% of record), discontinuous (suppression 10%–49% of record), burst suppression (suppression 50%–90% of record), and overall suppression (suppression >90% of record). The presence or absence of IV sedating medications during each epoch was also recorded. For records predating 2013, background continuity and sedating medications were abstracted manually from EEG reports. Records with ambiguous documentation were excluded. A single continuity value was assigned to each 24-hour EEG epoch, reflecting the highest continuity level within that epoch. Each patient was classified by the most continuous EEG background observed after cardiac arrest. For further analyses, the best-achieved continuity was then binarized on the absence or presence of EEG background suppression (a category including burst suppression and overall suppression and corresponding to the recently-defined highly-malignant EEG.6

Reactivity was reported for a subset of EEGs (n = 81) in a standardized manner. This was defined as a change in amplitude or frequency (excluding myogenic artifact or ictal discharges) in response to stimulation,27 which could include auditory, nail bed pressure, sternal rub, endotracheal suctioning, nasal tickle, or nursing care.

Seizures

The presence of seizures was determined per EEG epoch according to the modified Young criteria for nonconvulsive electrographic seizures28 or Salzburg criteria for status epilepticus.29 Patients were dichotomized as either having or not having seizures after cardiac arrest.

Image Processing and Quality Control

When multiple MRIs were performed, only the first MRI was included in the analysis. Patients' clinically acquired T1 and diffusion sequences were used. Scans were acquired on several clinical magnets over the course of the study, including Siemen's Verio (3T), TrioTrim (3T), and Aera (1.5T) (Erlangen, Germany) and GE Signa (1.5T) and Discovery (3T) (Chicago, IL). Parameters of the diffusion acquisition varied but included b = 0, 500, and 1,000 seconds/mm2, 1.5 × 1.5 × 6–mm voxel size, and diffusion encoding along the principal axes. T1-weighted acquisitions included either sagittal and axial precontrast volumes (voxel size ≈0.5 × 0.5 × 5 mm) or an magnetization-prepared rapid gradient echo volume (voxel size 1 × 1 × 1 mm).

We quantitatively analyzed the ADC maps from the diffusion-weighted acquisition. Although qualitative examination of other sequences such as fluid-attenuated inversion recovery has demonstrated value after cardiac arrest,16,30,31 there is not an established precedent for using the raw intensity values.

To register the ADC maps into a standardized space, we performed a brain extraction (FSL, bet2) and concatenated linear, nonlinear, and diffeomorphic registrations (ANTs: antsRegistrationSyNQuick) between the diffusion b0 volume, the T1-weighted MRI, and Montreal Neurological Institute (MNI) 152 T1 space. This was accomplished using code from GitHub.32 When a b0 volume was unavailable, the ADC map was used for cross-modality registration. If a T1-weighted isotropic magnetization-prepared rapid gradient echo sequence was not available, clinical axial and sagittal T1 volumes were first bias-field corrected and resliced into an isotropic T1 volume. This was then reconstructed into a higher-resolution T1 volume with NiftyMIC,33 which uses a previously described 2-step, iterative method with rigid slice to volume registration and cycles of robust super resolution reconstruction.34 All registrations were then manually inspected with slicesdir (FSL 6.0.1). Two patients' MRIs were excluded due to grossly inaccurate registrations by visual inspection.

To minimize signal contribution from artifact and CSF, we thresholded the ADC maps between 200 and 1,200 × 10−6 mm2/s, as done previously.15-17 The above transformations were then applied to the thresholded ADC maps to bring them into MNI space for group-level comparisons.

Controls Without Cardiac Arrest

To identify a group of patients unlikely to have significant brain pathology, we screened our enterprise-wide electronic medical record for MRI reports containing the text string “migraine headaches” between 2000 and 2020. From this list of >6,000 patients (seen at all hospitals within our hospital system), we randomly selected 150. Of those, 57 were from our specific institution and therefore comparable to the cohort of patients with cardiac arrest; these were processed in the same manner as above. After exclusion of 9 patients missing diffusion or T1 sequences or with grossly inaccurate registrations, 48 were included for further analysis (Table 1, controls).

Between-Group Differences in Baseline Characteristics

Differences in demographic, clinical, and radiologic characteristics between individuals with and without cardiac arrest, with and without DoC, with varying EEG background continuity, and with or without seizures were calculated with 1-way analysis of variance (continuous variable, 3 groups), 2-sample t test (continuous variable, 2 groups), or χ2 tests (binary variable).

Voxel Lesion Symptom Mapping

Voxel lesion symptom mapping was conducted using randomise (FSL 6.0.1). Given that there is no agreed-on ADC threshold to predict tissue infarction after cardiac arrest14-16 or ischemic stroke,35 we treated the ADC signal as a continuous variable. To mitigate any possible effects of registration variability and sulcal volume differences between patients with cardiac arrests and controls, we excluded, from all participants, voxels with ADC values of 0 (indicating a prethreshold ADC value of <200 or >1,200 in >20% of participants) (eFigure 1A Supplement http://links.lww.com/WNL/B748). Contrasts were generated with 2-sample t tests with or without covariates (FSL, randomise), with 2,000 permutations and threshold-free cluster enhancement (TFCE).36 Because quantitative diffusion metrics like ADC are known to vary by age,37 time from injury,38 MRI manufacturer,39 and field strength,40 we conducted additional analyses controlling for these values as nuisance covariates, mean-centering the continuous variables. To ensure that the results were not dependent on the processing strategy, we replicated the primary analyses after binarizing each participant’s ADC map at previously used15,16 ADC thresholds of 550 or 650 × 10−6 mm2/s and after normalizing each participant’s ADC map to the mean value within the brainstem, a region rarely affected by ABI.19

Region of Interest ADC Measurement

Using the previously described transformations, we transformed binary, MNI space a priori regions of interest (ROIs) onto each participant’s thresholded ADC map and computed the robust (nonzero) mean ADC value within each ROI. Gray matter ROIs included bilateral frontal, temporal, insular, parietal, and occipital lobes and cerebellum (MNI Structural Atlas); bilateral basal ganglia (combined caudate, putamen, and pallidum); bilateral thalami; and brainstem (Harvard-Oxford Subcortical Atlas). White matter ROIs included all bilateral and binarized tracts in the Johns Hopkins University White Matter Atlas and the corpus callosum (Julich Atlas).

Multivariate Regional Associations With Outcome

We observed substantial collinearity between mean ADC measurements of different brain regions (Pearson R = 0.30–0.99, eFigure 2 Supplement http://links.lww.com/WNL/B748). Combining ADC measurements from multiple ROIs into a single regression model yielded unstable effect estimates, precluding the identification of independent regional predictors. We therefore used lasso regression, a technique that penalizes models with larger numbers of coefficients and is well suited to winnowing multiple colinear predictor variables (R package glmnet). Modeling binary outcomes (DoC, background suppression, or seizures), we included each anatomic ROI as a candidate predictor variable. All variables were mean-centered and scaled to unit variance. We computed the receiver operating characteristic (ROC) curve (R package pROC) for the model, as well as for the whole-brain and whole-cortex mean ADC values. We then compared the area under the ROC curve (AUROC) of the lasso regression model to that of whole-brain and whole-cortex ADC using the Delong test.

To ensure that these results were not driven by outliers, we performed a repeated, stratified cross-validation procedure, iteratively training a lasso regression model with anatomic ROIs in 50% of the data and testing in the remaining 50%, maintaining the ratio of outcomes between samples (R package caret). For each of 1,000 iterations, we computed the AUROCs for the lasso regression, mean whole-brain ADC, and mean whole-cortex ADC in the test sample. We then reported the percentage of times that the model AUROC was larger than the whole-brain and whole-cortex AUROCs.

Data Availability

Anonymized clinical and MRI data will be made available by request from any qualified investigator.

Results

Distribution of ABI After Cardiac Arrest

Characteristics of the 204 patients who were unresponsive after cardiac arrest and underwent diffusion MRI are provided in Table 1. Controls were slightly younger and had a higher proportion of women and a higher proportion of individuals identifying as White (p < 0.05, Table 1). To determine the brain regions most affected by anoxia, we compared the voxel-wise distributions of ADC values between patients who did (n = 204 Figure 1A) and did not (n = 48, Figure 1B) sustain a cardiac arrest. Compared to controls, patients with cardiac arrest had lower ADC values across multiple cortical regions, most prominently in the occipital lobes (PTFCE < 0.05, Figure 1C). An ROI-based analysis confirmed a significant region-by-group interaction (F21,252 = 2.0, p = 0.005), with the occipital lobes showing the largest ADC difference between groups (T = 4.6). The set of identified regions remained consistent after controlling for possible sources of confounding, including demographics (eFigure 3A), MRI field strength (1.5T vs 3T, eFigure 3B), and MRI manufacturer (Siemens vs GE, eFigure 3C). The set of regions was consistent after the analysis was restricted to patients enrolled after 2013 (n = 164 patients, eFigure 4A), to those scanned between 2 and 5 days after arrest (n = 125 patients, eFigure 5A), or to those scanned on the 2 most common scanner types (n = 175 patients, 21 controls, eFigure 6A), The areas of injury identified after binarizing the ADC maps with previous thresholds15 (550, 650 × 10−6, eFigure 3, D and E) were broader but located in similar regions.

Figure 1. Distribution of Anoxic Brain Injury After Cardiac Arrest.

Figure 1

Open in a new tab

(A) Mean apparent diffusion coefficient (ADC) map for controls. (B) Mean ADC map for all patients with cardiac arrest. (C) Voxels with significantly lower ADC in patients with cardiac arrest compared to controls are localized at the gray-white junction, most prominent occipitally. Map corrected for multiple comparisons using threshold-free cluster enhancement.

DoC: Distribution of Injury

The cardiac arrest cohort contained a mix of patients with (CARecovered, 39%) and without (CADoC, 61%) recovery of consciousness, as measured by command following (eTable 1 Supplement http://links.lww.com/WNL/B748). We sought to determine which brain regions were specifically lesioned in CADoC compared with CARecovered patients. Compared to CARecovered patients (Figure 2A), CADoC patients (Figure 2B) had lower ADC values across a range of subcortical and cortical regions, most prominently in the occipital lobes (PTFCE < 0.05, Figure 2C). An ROI-based analysis confirmed this interpretation, with a region-by-group interaction (F21,203 = 4.7, p < 0.001) and the largest between-group ADC difference in the occipital lobes (T = 9.4). Although CADoC patients were older than the CARecovered patients (p < 0.001, eTable 1 Supplement http://links.lww.com/WNL/B748), controlling for age did not alter the resulting maps (eFigure 7 Supplement http://links.lww.com/WNL/B748), nor did restricting the analyses to patients enrolled after 2013 (eFigure 4B), patients scanned between 2 and 5 days after arrest (eFigure 5B), or patients scanned on the most common scanner type (n = 125, eFigure 8A). A less widely significant but similarly distributed map was seen when the analyses were restricted to patients surviving >7 days (n = 146, eFigure 9). Whole-brain ADC differed between CADoC, CARecovered, and controls (F2,250 = 20, p < 0.001, Figure 2D), with ADC higher in CARecovered patients compared to controls (T = 3.7, p < 0.001, Figure 2D and eFigure 10).

Figure 2. DoC After Cardiac Arrest Are Associated With Occipitally Predominant Injury.

Figure 2

Open in a new tab

Group-wide mean apparent diffusion coefficient (ADC) maps for individuals with cardiac arrest who did (CARecovered, n = 80; A) or did not (CADoC, n = 123; B) recover consciousness after cardiac arrest show areas of different ADC signal between groups. (C) Cortical (most prominent occipitally), putamen, globus pallidus, medial thalamus, and cerebellar voxels have significantly lower ADC in CADoC compared to CARecovered (PTFCE < 0.05). Distribution of whole-brain mean ADC across study groups (D) illustrates the high variance in CADoC and the elevated mean ADC in CARecovered relative to controls. (E) Area under receiver operating characteristic curves (AUROC) demonstrates that occipital ADC (red) better classifies individuals with disorders of consciousness (DoC) than whole-brain (dark blue) or whole-cortex (light blue) ADC. AUROC values are listed next to the legend. ***p < 0.0005, **p < 0.005, *p < 0.05.

DoC: Global vs Regional Injury

We next sought to determine which set of brain regions were independently associated with recovery of consciousness after cardiac arrest and whether global or regional injury was a stronger predictor of DoC. To identify the strongest independent regional associations with DoC, we ran a penalized (lasso) regression on the set of all anatomic ROIs, finding that only the occipital ADC had a nonzero coefficient. The AUROC for this occipital-only model (0.78 [95% confidence interval 0.71, 0.84]) was significantly greater than that of whole-brain (0.71 [0.64, 0.78], Delong Z = 3.6, p = 0.0003) or whole-cortex (0.73 [0.67, 0.80], Z = 2.7, p = 0.008, Figure 2E) ADC. Confirming the robustness of these results within this dataset, an ROI-based lasso regression model had a greater AUROC than the whole-brain and whole-cortex ADC in 99.5% of 1,000 split-half iterations. The occipital ADC was included in 100% of models, and the brainstem and anterior thalamic radiation ADCs were included as covariates in 1% and 0.1% of models, respectively.

EEG Background Suppression

Characteristics of patients with different EEG background continuities (n = 175) are shown in eTable 2 Supplement http://links.lww.com/WNL/B748, and temporal trends in EEG background continuity are displayed in eFigure 11. The prevalence of background suppression decreased by whole-brain ADC tertile (eFigure 12A). Background suppression had a positive predictive value of 65% for overt ABI (whole-brain ADC value in the lowest tertile). All the brain regions lesioned in CADoC patients (eFigure 1B) were also associated with the presence of background suppression (PTFCE < 0.05, Figure 3A), even after the elapsed time between arrest and MRI was added as a nuisance covariate (eFigure 13A). The distribution remained consistent when the analysis was restricted to patients enrolled after 2013 (eFigure 4C), patients scanned between 2 and 5 days after arrest (eFigure 5C), or patients scanned on the most common scanner type (eFigure 8B). Reactivity, another commonly used EEG feature after cardiac arrest, showed a similar voxel-wise pattern of association with ADC (n = 81, eFigure 14).

Figure 3. Anoxic Brain Injury Severity Is Associated With Background Suppression After Cardiac Arrest.

Figure 3

Open in a new tab

(A) Threshold-free cluster enhancement–corrected map of brain voxels with lower apparent diffusion coefficient (ADC) in patients with compared to without persistent background suppression (BS) appears similar to the map of voxels associated with disorders of consciousness. (B) Receiver operating characteristic (ROC) curves comparing the ability of a lasso regression model including cerebellar and occipital ADC (red) to classify patients with vs without BS to whole-brain (dark blue) and whole-cortex (light blue) ADC. Area under ROC values are listed next to the legend. Occipital/cerebellar model could no better classify patients with BS than whole-brain or whole-cortex ADC.

The AUROC for the lasso regression model for background suppression with occipital and cerebellar ADC as its only nonzero coefficients (0.81 [0.74, 0.88]) was no different from that achieved with whole-brain (0.79 [0.71, 0.86], Z = 1.7, p = 0.1) or whole-cortex (0.80 [0.72, 0.87], Z = 1.2, p = 0.2, Figure 3B) ADC. In a robustness analysis, the lasso regressions included a range of anatomic ROIs (eFigure 13B), and the model-based AUROC was greater than whole-brain and whole-cortex ADC in only 56% of 1,000 split-half iterations.

Seizures: Distribution of Injury

Baseline characteristics of patients with (n = 35) and without (n = 158) seizures are shown in eTable 3 Supplement http://links.lww.com/WNL/B748. Despite no other demographic or clinical differences, whole-brain ADC was lower in patients without (mean 783 [767, 799]) compared to with (817 [793, 842]) seizures (p < 0.05, eTable 3 Supplement http://links.lww.com/WNL/B748, Figure 4A). Seizure incidence decreased nonsignificantly with each decreasing ADC tertile (χ2[2,193] = 3.6, p = 0.2, eFigure 12B). The ADC signal within lateral temporal white matter showed the strongest (inverse) association with seizures (PTFCE < 0.05, Figure 4B). This regional distribution persisted after the analysis was restricted to patients enrolled after 2013 (eFigure 4D), to patients scanned between 2 and 5 days after arrest (eFigure 5D), or to patients scanned on the 2 most common scanner types (eFigure 6B).

Figure 4. Regional ADC Is Lower in Patients Without Seizures After Cardiac Arrest.

Figure 4

Open in a new tab

(A) Patients without seizures have a lower mean whole-brain apparent diffusion coefficient (ADC) compared to those with seizures. (B) Threshold-free cluster enhancement–corrected map of brain voxels with lower ADC in patients without compared to with seizures, illustrating peak differences in the lateral temporal white matter. (C) Temporal lobe ADC (red) is superior at classifying patients with vs without seizures compared with whole-brain (dark blue) and whole-cortex (light blue) ADC. Area under receiver operating characteristic values provided adjacent to legend. ***p < 0.0005, **p < 0.005, *p < 0.05.

Seizures: Global vs Regional

No individual ROIs were associated with seizures using lasso regression. Given that the voxel-wise seizure association maps highlighted the temporal regions, we explicitly tested whether mean temporal lobe ADC values or whole-brain measures of ADC could better distinguish patients with from those without seizures. Temporal lobe ADC had a greater AUROC (0.67 [0.57, 0.77]) for identifying patients with seizures than either whole-brain (0.59 [0.50, 0.69], Z = 3.7, p < 0.001) or whole-cortex (0.59 [0.49, 0.69], Z = 3.1, p = 0.002, Figure 4C) ADC. Temporal lobe ADC had a greater AUROC than whole-brain and whole-cortical ADC in 99.1% of 1,000 samples of 50% of the data, confirming that these findings were not driven by outlier patients.

Classification of Evidence

This study provides Class IV evidence that DoC after cardiac arrest are associated with widely lower ADC values on diffusion MRI and are most strongly associated with reductions in occipital ADC.

Discussion

Here, we demonstrate that cardiac arrest produces a posterior-predominant anoxic injury pattern. While DoC after cardiac arrest are associated with widespread ABI, it is most strongly associated with the severity of occipital injury. We also find that background suppression is a nonspecific marker of overall ABI severity. Seizures are associated with less severe ABI, and lateral temporal anoxia is specifically associated with a lower incidence of seizures.

Diminished ADC after cardiac arrest occurs across a range of brain regions. Why adjacent regions are affected to different extents (medial vs lateral thalamus, putamen vs caudate, occipital vs posterior frontal) remains unknown but may relate to regional differences in metabolic demand41 and perfusion.1,42

Primary visual cortex has been theorized to have the highest resting metabolic rate in the brain,41 which could explain the occipital sensitivity to anoxia, but recent work has disputed this.43 The posterior predominance seen here is also seen in other neurologic conditions. Posterior reversible encephalopathy syndrome (PRES) is thought to be caused by a diminished autoregulatory reserve of the posterior cerebral circulation.44 We postulate that PRES and ABI may represent failures at opposite ends of the cerebral autoregulatory curve: a failure of flow limitation at high blood pressures in PRES and a failure to maintain perfusion with low blood pressures in cardiac arrest.

Although prior literature contends that the hippocampus may be particularly vulnerable to anoxia,3 we did not identify preferential ABI in the hippocampus. However, it possible that these clinical diffusion-weighted scans, which poorly discriminate mesial temporal structures,45 had insufficient resolution to fully assess hippocampal ABI.

We found that DoC after cardiac arrest are better predicted by occipital ABI than by the global burden of injury, consistent with theories that posit a posterior predominance of structures important for consciousness.46 This posterior predominance persisted after the exclusion of individuals with early death in whom behavioral categorization may have been less accurate. Therefore, injury to posterior cortical structures may be more informative than whole-brain measures of ABI for identifying patients at risk for prolonged DoC after cardiac arrest.

The increased diffusivity (ADC) that we observed in patients who recovered consciousness after cardiac arrest has been previously reported19,31 and may reflect mild vasogenic edema.47,48 This finding may also reflect unmeasured, fundamental differences between the cardiac arrest and control populations (e.g., intubation, critical illness). Whether this increased diffusivity results in permanent white matter alterations and underlies some of the persistent cognitive complaints recorded in survivors49 should be investigated.

While occipital ABI and cerebellar ABI showed the strongest regional associations, they were not superior to whole-brain summary measures at predicting background suppression. How such injury results in background suppression remains unknown. Although the thalamus is classically thought to produce the oscillatory signals underlying the continuous EEG background,50 autopsy studies have found cortical and cerebellar injury more frequently than thalamic injury in patients with background suppression after cardiac arrest, consistent with our findings.3,51 We found background suppression to be a relatively nonspecific marker of ABI severity. Indeed, other groups have postulated that background suppression may even be neuroprotective after cardiac arrest,11 representing a functional adaptation of neurons to a metabolic crisis.

Reduced whole-brain ADC values and reduced temporal lobe ADC values in particular were associated with a decreased prevalence of seizures. Widespread anoxic injury may leave behind too few intact neuronal elements to generate the synchronized neuronal activity required for seizures. Injury to the lateral temporal lobes specifically may mimic a functional temporal lobectomy, which is known to inhibit seizures. Alternatively, uncontrolled seizures may cause vasogenic temporal lobe edema, which might elevate ADC signal. The possible association between temporal lobe lesions and seizure inhibition may have broader therapeutic implications, warranting further study.

Diffusion-weighted MRI is limited in evaluating brain regions near the skull base52; thus, our sensitivity for ABI in these regions may have been similarly limited. Furthermore, while including diffusion acquisitions across multiple scanner types and field strengths permitted a larger sample size, such heterogeneity may have added variance to our data and reduced our power to identify regional associations.

EEG continuity and seizures were not systematically classified before 2013, and variability in documentation may have degraded the quality of EEG variables. Furthermore, we did not have adequate power to classify seizure subtypes. It is possible that different types of cerebral lesions predispose to different types of seizures (e.g., myoclonic seizures).

Given that we did not assess why and how life-sustaining therapy was withdrawn, it is possible that some patients with DoC or persistent background suppression may have had an intact neural substrate for recovery but died before such a recovery could be realized. In general, misclassifications such as these would have reduced our power to identify specific associations with regional injury, and our results were robust to the exclusion of patients with early death. Future research with finer-grained behavioral classifications and stricter methods for controlling for mortality may further refine these associations.

Scan time and scanner type heterogeneity in this dataset could have introduced difficult-to-model variance. While a series of sensitivity analyses demonstrate the robustness of our findings, the time course of ADC changes after cardiac arrest14-16,19 is less established than in acute ischemic stroke. A single low ADC measurement may not always indicate irreversible tissue injury.

Given that MRI results are frequently used for clinical prognostication, a self-fulfilling prophecy bias may influence the associations with clinical outcomes: the presence of ABI on MRI may have been more likely to result in the early withdrawal of life-sustaining therapy and therefore in the failure to regain command following. The measured association between regional ADC and DoC after cardiac arrest persisted after the exclusion of patients who died in the first week, suggesting that the early withdrawal of life-sustaining therapy was not the primary driver of our findings. Nonetheless, future research with stricter control for mortality and the withdrawal of life-sustaining therapy may help confirm the associations observed here.

Finally, while we demonstrate that these findings are robust to outliers within our dataset, it is important to acknowledge the limited generalizability and possible confounds that characterize any retrospective, single-center study. Furthermore, because only a subset of our dataset underwent MRI, the findings are subject to selection bias. Patients who recover consciousness or die soon after arrest and those who remain critically ill and unstable would not have been included in our cohort and are not represented here.

In a large retrospective dataset of unresponsive patients after cardiac arrest, we identified the brain-wide distribution of ABI as measured with diffusion MRI, finding more involvement in posterior cortical regions. DoC were associated with anoxic injury to a broad range of cortical and subcortical regions but could be classified most accurately by the severity of occipital injury. Background suppression, a common EEG finding after cardiac arrest, was also a nonspecific marker of overall ABI severity and was not preferentially associated with anoxic injury to a particular region. Finally, seizures were less frequent in patients with severe ABI, especially after anoxia to the lateral temporal lobes. In total, our results suggest that the regional pattern of ABI after cardiac arrest may hold importance for prognosis and salient clinical sequelae.

Disclosure

J.W. Lee reports contract work for Bioserenity and Teladoc, serving as a consultant for Biogen), and being a cofounder of Soterya, Inc. S.B Snider, D. Fischer, M.E. McKeown, A.L. Cohen, F.L.W.J Schaper, E. Amorim, M.D. Fox, B. Scirica, and M.B. Bevers report no disclosures relevant to the content of this manuscript. Go to Neurology.org/N for full disclosures.

Glossary

ABI

anoxic brain injury

ADC

apparent diffusion coefficient

AUROC

area under the ROC curve

DoC

disorders of consciousness

MNI

Montreal Neurological Institute

PRES

posterior reversible encephalopathy syndrome

ROC

receiver operating characteristic

ROI

region of interest

TFCE

threshold-free cluster enhancement

Appendix. Authors

Appendix.

Open in a new tab

Footnotes

Class of Evidence: NPub.org/coe

Study Funding

S.B.S. was supported by the American Academy of Neurology (CRTS-2020A013392). E.A. was supported during this research by the NIH (1K23NS119794), Hellman Fellows Fund, Regents of the University of California (Resource Allocation Program), CURE Epilepsy Foundation (Taking Flight Award), Weil-Society of Critical Care Medicine Research Grant, and American Heart Association (20CDA35310297). M.B.B. was supported during this research by the NIH (K23NS112474) and American Academy of Neurology (CRTS AI18-0000000062). D.F. was supported by the NIH National Institute of Neurologic Disorders and Stroke (R25NS06574309).

References

  • 1.Busl KM, Greer DM. Hypoxic-ischemic brain injury: pathophysiology, neuropathology and mechanisms. NeuroRehabilitation. 2010;26(1):5-13. [DOI] [PubMed] [Google Scholar]
  • 2.Petito CK, Feldmann E, Pulsinelli WA, Plum F. Delayed hippocampal damage in humans following cardiorespiratory arrest. Neurology. 1987;37(8):1281-1286. [DOI] [PubMed] [Google Scholar]
  • 3.Endisch C, Westhall E, Kenda M, et al. Hypoxic-ischemic encephalopathy evaluated by brain autopsy and neuroprognostication after cardiac arrest. JAMA Neurol. 2020;77(11):1430-1439. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Levy DE, Caronna JJ, Singer BH, Lapinski RH, Frydman H, Plum F. Predicting outcome from hypoxic-ischemic coma. JAMA. 1985;253(10):1420-1426. [PubMed] [Google Scholar]
  • 5.Greer DM, Yang J, Scripko PD, et al. Clinical examination for prognostication in comatose cardiac arrest patients. Resuscitation. 2013;84(11):1546-1551. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Westhall E, Rossetti AO, van Rootselaar AF, et al. Standardized EEG interpretation accurately predicts prognosis after cardiac arrest. Neurology. 2016;86(16):1482-1490. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Barbella G, Lee JW, Alvarez V, et al. Prediction of regaining consciousness despite an early epileptiform EEG after cardiac arrest. Neurology. 2020;94(16):e1675-e1683. [DOI] [PubMed] [Google Scholar]
  • 8.Lybeck A, Friberg H, Aneman A, et al. Prognostic significance of clinical seizures after cardiac arrest and target temperature management. Resuscitation. 2017;114:146-151. [DOI] [PubMed] [Google Scholar]
  • 9.Elmer J, Rittenberger JC, Faro J, et al. Clinically distinct electroencephalographic phenotypes of early myoclonus after cardiac arrest. Ann Neurol. 2016;80(2):175-184. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Knight WA, Hart KW, Adeoye OM, et al. The incidence of seizures in patients undergoing therapeutic hypothermia after resuscitation from cardiac arrest. Epilepsy Res. 2013;106(3):396-402. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Forgacs PB, Devinsky O, Schiff ND. Independent functional outcomes after prolonged coma following cardiac arrest: a mechanistic hypothesis. Ann Neurol. 2020;87(4):618-632. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Beuchat I, Sivaraju A, Amorim E, et al. MRI-EEG correlation for outcome prediction in postanoxic myoclonus: a multicenter study. Neurology. 2020;95(4):e335-e341. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Barth R, Zubler F, Weck A, et al. Topography of MR lesions correlates with standardized EEG pattern in early comatose survivors after cardiac arrest. Resuscitation. 2020;149:217-224. [DOI] [PubMed] [Google Scholar]
  • 14.Wijman CA, Mlynash M, Caulfield AF, et al. Prognostic value of brain diffusion-weighted imaging after cardiac arrest. Ann Neurol. 2009;65(4):394-402. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Hirsch KG, Fischbein N, Mlynash M, et al. Prognostic value of diffusion-weighted MRI for post-cardiac arrest coma. Neurology. 2020;94(16):e1684-e1692. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Bevers MB, Scirica BM, Avery KR, Henderson GV, Lin AP, Lee JW. Combination of clinical exam, MRI and EEG to predict outcome following cardiac arrest and targeted temperature management. Neurocrit Care. 2018;29(3):396-403. [DOI] [PubMed] [Google Scholar]
  • 17.Wu O, Sorensen AG, Benner T, Singhal AB, Furie KL, Greer DM. Comatose patients with cardiac arrest: predicting clinical outcome with diffusion-weighted MR imaging. Radiology. 2009;252(1):173-181. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Hirsch KG, Mlynash M, Eyngorn I, et al. Multi-center study of diffusion-weighted imaging in coma after cardiac arrest. Neurocrit Care. 2016;24(1):82-89. [DOI] [PubMed] [Google Scholar]
  • 19.Mlynash M, Campbell DM, Leproust EM, et al. Temporal and spatial profile of brain diffusion-weighted MRI after cardiac arrest. Stroke. 2010;41(8):1665-1672. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Dijkhuizen RM, Knollema S, van der Worp HB, et al. Dynamics of cerebral tissue injury and perfusion after temporary hypoxia-ischemia in the rat: evidence for region-specific sensitivity and delayed damage. Stroke. 1998;29(3):695-704. [DOI] [PubMed] [Google Scholar]
  • 21.Wei R, Wang C, He F, et al. Prediction of poor outcome after hypoxic-ischemic brain injury by diffusion-weighted imaging: a systematic review and meta-analysis. PLoS One. 2019;14(12):e0226295. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Sandroni C, Cariou A, Cavallaro F, et al. Prognostication in comatose survivors of cardiac arrest: an advisory statement from the European Resuscitation Council and the European Society of Intensive Care Medicine. Resuscitation. 2014;85(12):1779-1789. [DOI] [PubMed] [Google Scholar]
  • 23.Wijdicks EF, Hijdra A, Young GB, Bassetti CL, Wiebe S, Quality Standards Subcommittee of the American Academy of Neurology. Practice parameter: prediction of outcome in comatose survivors after cardiopulmonary resuscitation (an evidence-based review): report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology. 2006;67(2):203-210. [DOI] [PubMed] [Google Scholar]
  • 24.Nolan JP, Soar J, Cariou A, et al. European Resuscitation Council and European Society of Intensive Care Medicine guidelines for post-resuscitation care 2015: section 5 of the European Resuscitation Council Guidelines for resuscitation 2015. Resuscitation. 2015;95:202-222. [DOI] [PubMed] [Google Scholar]
  • 25.Hirsch LJ, LaRoche SM, Gaspard N, et al. American Clinical Neurophysiology Society's standardized critical care EEG terminology: 2012 version. J Clin Neurophysiol. 2013;30(1):1-27. [DOI] [PubMed] [Google Scholar]
  • 26.Lee JW, LaRoche S, Choi H, et al. Development and feasibility testing of a critical care EEG monitoring database for standardized clinical reporting and multicenter collaborative research. J Clin Neurophysiol. 2016;33(2):133-140. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Admiraal MM, van Rootselaar AF, Hofmeijer J, et al. Electroencephalographic reactivity as predictor of neurological outcome in postanoxic coma: a multicenter prospective cohort study. Ann Neurol. 2019;86(1):17-27. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Young GB, Jordan KG, Doig GS. An assessment of nonconvulsive seizures in the intensive care unit using continuous EEG monitoring: an investigation of variables associated with mortality. Neurology. 1996;47(1):83-89. [DOI] [PubMed] [Google Scholar]
  • 29.Beniczky S, Hirsch LJ, Kaplan PW, et al. Unified EEG terminology and criteria for nonconvulsive status epilepticus. Epilepsia. 2013;54(suppl 6):28-29. [DOI] [PubMed] [Google Scholar]
  • 30.Hirsch KG, Mlynash M, Jansen S, et al. Prognostic value of a qualitative brain MRI scoring system after cardiac arrest. J Neuroimaging. 2015;25(3):430-437. [DOI] [PubMed] [Google Scholar]
  • 31.Velly L, Perlbarg V, Boulier T, et al. Use of brain diffusion tensor imaging for the prediction of long-term neurological outcomes in patients after cardiac arrest: a multicentre, international, prospective, observational, cohort study. Lancet Neurol. 2018;17(4):317-326. [DOI] [PubMed] [Google Scholar]
  • 32.GitHub. January 1, 2021. github.com/bchcohenlab/bids_lesion_code
  • 33.GitHub. January 1, 2021. github.com/gift-surg/NiftyMIC
  • 34.Ebner M, Wang G, Li W, et al. An automated framework for localization, segmentation and super-resolution reconstruction of fetal brain MRI. Neuroimage. 2020;206:116324. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Na DG, Thijs VN, Albers GW, Moseley ME, Marks MP. Diffusion-weighted MR imaging in acute ischemia: value of apparent diffusion coefficient and signal intensity thresholds in predicting tissue at risk and final infarct size. AJNR Am J Neuroradiol. 2004;25(8):1331-1336. [PMC free article] [PubMed] [Google Scholar]
  • 36.Smith SM, Nichols TE. Threshold-free cluster enhancement: addressing problems of smoothing, threshold dependence and localisation in cluster inference. Neuroimage. 2009;44(1):83-98. [DOI] [PubMed] [Google Scholar]
  • 37.Bennett IJ, Madden DJ, Vaidya CJ, Howard DV, Howard JH Jr. Age-related differences in multiple measures of white matter integrity: a diffusion tensor imaging study of healthy aging. Hum Brain Mapp. 2010;31(3):378-390. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Eastwood JD, Engelter ST, MacFall JF, Delong DM, Provenzale JM. Quantitative assessment of the time course of infarct signal intensity on diffusion-weighted images. AJNR Am J Neuroradiol. 2003;24(4):680-687. [PMC free article] [PubMed] [Google Scholar]
  • 39.Grech-Sollars M, Hales PW, Miyazaki K, et al. Multi-centre reproducibility of diffusion MRI parameters for clinical sequences in the brain. NMR Biomed. 2015;28(4):468-485. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Huisman TA, Loenneker T, Barta G, et al. Quantitative diffusion tensor MR imaging of the brain: field strength related variance of apparent diffusion coefficient (ADC) and fractional anisotropy (FA) scalars. Eur Radiol. 2006;16(8):1651-1658. [DOI] [PubMed] [Google Scholar]
  • 41.Ishii K, Sasaki M, Kitagaki H, Sakamoto S, Yamaji S, Maeda K. Regional difference in cerebral blood flow and oxidative metabolism in human cortex. J Nucl Med. 1996;37(7):1086-1088. [PubMed] [Google Scholar]
  • 42.Cervos-Navarro J, Diemer NH. Selective vulnerability in brain hypoxia. Crit Rev Neurobiol. 1991;6(3):149-182. [PubMed] [Google Scholar]
  • 43.Shokri-Kojori E, Tomasi D, Alipanahi B, Wiers CE, Wang GJ, Volkow ND. Correspondence between cerebral glucose metabolism and BOLD reveals relative power and cost in human brain. Nat Commun. 2019;10(1):690. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Lamy C, Oppenheim C, Meder JF, Mas JL. Neuroimaging in posterior reversible encephalopathy syndrome. J Neuroimaging. 2004;14(2):89-96. [PubMed] [Google Scholar]
  • 45.Treit S, Steve T, Gross DW, Beaulieu C. High resolution in-vivo diffusion imaging of the human hippocampus. Neuroimage. 2018;182:479-487. [DOI] [PubMed] [Google Scholar]
  • 46.Boly M, Massimini M, Tsuchiya N, Postle BR, Koch C, Tononi G. Are the neural correlates of consciousness in the front or in the back of the cerebral cortex? Clinical and neuroimaging evidence. J Neurosci. 2017;37(40):9603-9613. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Barzo P, Marmarou A, Fatouros P, Hayasaki K, Corwin F. Contribution of vasogenic and cellular edema to traumatic brain swelling measured by diffusion-weighted imaging. J Neurosurg. 1997;87(6):900-907. [DOI] [PubMed] [Google Scholar]
  • 48.Schaefer PW, Buonanno FS, Gonzalez RG, Schwamm LH. Diffusion-weighted imaging discriminates between cytotoxic and vasogenic edema in a patient with eclampsia. Stroke. 1997;28(5):1082-1085. [DOI] [PubMed] [Google Scholar]
  • 49.Steinbusch CVM, van Heugten CM, Rasquin SMC, Verbunt JA, Moulaert VRM. Cognitive impairments and subjective cognitive complaints after survival of cardiac arrest: a prospective longitudinal cohort study. Resuscitation. 2017;120:132-137. [DOI] [PubMed] [Google Scholar]
  • 50.Schreckenberger M, Lange-Asschenfeldt C, Lochmann M, et al. The thalamus as the generator and modulator of EEG alpha rhythm: a combined PET/EEG study with lorazepam challenge in humans. Neuroimage. 2004;22(2):637-644. [DOI] [PubMed] [Google Scholar]
  • 51.van Putten M, Jansen C, Tjepkema-Cloostermans MC, et al. Postmortem histopathology of electroencephalography and evoked potentials in postanoxic coma. Resuscitation. 2019;134:26-32. [DOI] [PubMed] [Google Scholar]
  • 52.Alexander AL, Tsuruda JS, Parker DL. Elimination of eddy current artifacts in diffusion-weighted echo-planar images: the use of bipolar gradients. Magn Reson Med. 1997;38(6):1016-1021. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Anonymized clinical and MRI data will be made available by request from any qualified investigator.

Articles from Neurology are provided here courtesy of American Academy of Neurology

RESOURCES