EEG and Brain MRI Conflict After Cardiac Arrest Recovery

When a comatose patient survives a cardiac arrest, one of the most consequential questions in the intensive care unit is deceptively simple: will the brain recover? A new report published in the journal Neurocritical Care describes two patients whose courses upended that question, showing that even the most ominous electrical brain patterns are not always what they seem. In both cases, continuous electroencephalography revealed findings classified by international guidelines as “highly malignant” — patterns widely regarded as reliable harbingers of devastating, irreversible neurological injury — yet both patients eventually woke up and regained meaningful neurological function. The report, authored by Carlo Morotti Colleoni, Dimitra Vardalaki, Stelios Smirnakis, Benjamin M. Scirica, and Jong Woo Lee of Brigham and Women’s Hospital and the University of Milano-Bicocca, offers a vivid, clinically grounded lesson in the dangers of prognosticating from a single data stream.

The two patients arrived in the intensive care unit under very different circumstances. The first was a 22-year-old man who suffered a pulseless electrical activity arrest following an intentional barbiturate overdose. The second was a 68-year-old man who developed ventricular fibrillation after acute respiratory distress and required 22 minutes of cardiopulmonary resuscitation before his circulation returned. Both received targeted temperature management, a protocol of controlled cooling and rewarming designed to protect the injured brain, and in the second patient, sedative medications were discontinued after 48 hours. In line with the 2023 Neurocritical Care Society guidelines and other international consensus recommendations, the clinical team deferred formal neuroprognostication until at least 72 hours after return of spontaneous circulation and the completion of rewarming. When that window arrived, both patients remained comatose and showed no motor response to pain — a finding that, on its own, already weighs heavily against a favorable outcome.

It was the electroencephalographic data, however, that painted the grimmer picture. Continuous EEG monitoring had been initiated within the first 24 hours after cardiac arrest in both patients. In the first patient, recordings obtained during the initial 72 hours showed a persistently suppressed, nonreactive background. In the second, the EEG initially demonstrated severe background suppression followed by generalized periodic discharges superimposed on a low-voltage, nonreactive background. Under current guidelines, both patterns fall into the category Westhall and colleagues originally labeled “highly malignant,” and both are treated as moderately reliable predictors of poor functional outcome when recorded at least 72 hours after return of spontaneous circulation, in the absence of sedation and other confounders. The apparent specificity of these patterns is precisely what gives them their clinical weight: in large cohort studies, patients exhibiting suppressed or burst-suppression backgrounds weeks of observation aside rarely, if ever, wake up.

But then came the twist. Brain magnetic resonance imaging, performed on day 4 for the first patient and day 3 for the second, told a strikingly different story. The first patient’s MRI was entirely unremarkable — no restricted diffusion, no cortical laminar necrosis, no evidence whatsoever of diffuse hypoxic-ischemic brain injury. The second patient’s scan showed only subtle bilateral abnormalities in the caudate nuclei, without any of the widespread cortical and deep gray matter changes that characterize catastrophic anoxic injury. According to the 2023 Neurocritical Care Society guidelines, a diffuse pattern of restricted diffusion involving both cortex and deep gray matter, obtained between 2 and 7 days after return of spontaneous circulation, is considered a moderately reliable predictor of poor outcome. The converse is also meaningful: the absence of diffusion-weighted abnormalities, or the presence of only isolated lesions, is generally associated with a greater likelihood of favorable recovery. In these two patients, the imaging did not definitively exclude hypoxic-ischemic injury, but it dramatically lowered the probability of irreversible, devastating brain damage — and that discordance between the EEG and the MRI created a genuine prognostic dilemma with immediate, life-and-death implications for decisions about withdrawing life-sustaining treatment.

The authors systematically dissect the reasons why an EEG pattern can be “falsely malignant.” Hypothermia, or incomplete rewarming, can suppress cortical activity and reproduce highly malignant features even when neuronal viability is preserved. Timing matters enormously: EEG findings are dynamic, time-dependent markers, and recordings obtained too early may not yet reflect the true extent of injury. Residual sedation, impaired drug clearance, or sedative stacking can markedly reduce EEG continuity and reactivity, mimicking suppression or burst-suppression patterns that vanish once pharmacological effects resolve — a mechanism directly illustrated by the first patient, in whom persistent barbiturate detection prompted continuous veno-venous hemofiltration to accelerate drug elimination. Metabolic derangements, too, can transiently depress cortical activity and generate malignant-appearing tracings. The lesson is that a highly malignant EEG is a conditional finding, valid only when the conditions under which it was recorded are known to be clean.

MRI interpretation demands equal caution. Even within the optimal imaging window, MRI may underestimate the extent of hypoxic-ischemic injury when damage is subtle, predominantly subcortical, or still evolving. Diffusion-weighted abnormalities may be absent early after arrest, and conversely may normalize if imaging is performed too late, reducing the sensitivity of diffusion sequences. Prognostic accuracy also depends on sequence selection, timing, and interpretive expertise. Although not routinely recommended, quantitative diffusion analysis and advanced techniques such as magnetic resonance spectroscopy may provide complementary information in selected cases. The critical principle, the authors emphasize, is that MRI findings must be interpreted within a dynamic, multimodal framework rather than in isolation — precisely because a single reassuring or alarming result can mislead if stripped of context.

The discordant prognostic picture extended well beyond imaging. The 2023 guidelines hold that bilateral absence of pupillary light responses at 72 hours or later after return of spontaneous circulation is a reliable predictor of poor outcome, whereas partial preservation of brainstem reflexes is less specific. In these patients, the reflexes told a nuanced story: the first had preserved pupillary reflexes despite absent corneal responses, and the second had sluggish pupillary reflexes with preserved corneal responses. Biomarkers added further reassurance. Neuron-specific enolase, measured within the first 72 hours after resuscitation, was low in the first patient at 7.6 nanograms per milliliter — far below the thresholds of roughly 60 to 80 nanograms per milliliter associated with poor outcome. Somatosensory evoked potentials in the second patient demonstrated preserved bilateral N20 responses; bilateral absence of the cortical N20 response at 48 hours or later is a reliable predictor of poor outcome, and its preservation, while not independently predictive of good recovery, is undeniably reassuring. Taken together, these findings reduced confidence in an invariably poor prognosis and supported continued life-sustaining treatment with serial reassessment.

What followed was a vindication of that patience. In the first patient, from day 5 onward the EEG showed progressive recovery of background continuity and reactivity. Approximately seven days after the arrest, he began to awaken, eventually returning to his premorbid neurological baseline. In the second patient, life-sustaining treatment was continued despite the malignant EEG. On day 7, an electroclinical seizure was documented, with diffuse myoclonus accompanied by central sharp waves; treatment with levetiracetam and valproic acid abolished the myoclonus and improved the electroencephalogram. Over the following days, the EEG showed recovering background organization and reactivity, with progressive reduction of the periodic discharges. The patient regained consciousness and was discharged to rehabilitation. Both patients, despite initially highly malignant EEG patterns, recovered from coma and achieved meaningful neurological recovery.

The authors distill several lessons for clinicians confronting similar scenarios. Highly malignant EEG patterns, although highly specific for unfavorable outcome in carefully controlled conditions, should never be interpreted in isolation; their significance can be distorted by timing, temperature, sedation, metabolic disturbances, and drug accumulation. A normal or only mildly abnormal brain MRI, together with preserved brainstem reflexes — particularly pupillary responses — and the presence of purposeful motor responses, should prompt caution against early pessimistic prognostication. All available indicators must be systematically integrated within a multimodal framework. Perhaps most strikingly, delayed recovery emerges as a clinically relevant phenotype: patients may follow protracted but meaningful recovery trajectories that current guideline timeframes do not fully capture.

For families and surrogates, the counseling implications are equally clear. When prognostic tests conflict, clinicians should explicitly acknowledge uncertainty and avoid anchoring decisions to any single modality. Families should understand that meaningful recovery may still occur when prognostic information is discordant, and that an extended period of observation may be warranted when it aligns with the patient’s values and goals of care. In an era when withdrawal of life-sustaining treatment decisions are often made within days of a cardiac arrest, these two recoveries stand as a sobering reminder that the brain’s capacity for resilience can exceed what its earliest signals suggest — and that in neuroprognostication, humility is not just a virtue but a clinical necessity.