- Researchers at Massachusetts General Hospital have created a convolutional neural network—a form of artificial intelligence in which a computer system learns to recognize patterns in visual data—that efficiently labels huge volumes of continuous EEG data
- In this study, they used automated EEG labeling to quantify the epileptiform activity (EA) burden in 1,967 hospitalized patients across a wide variety of diagnoses
- Increasing EA burden was associated with a poor outcome (modified Rankin Scale score of 5 or 6 at hospital discharge); specifically, peak (maximum 12-hour) EA burden was associated with worse outcomes in a dose-dependent manner
- The dose-response relationship between peak EA burden and outcome was independent of the time interval between the last EEG measurement and hospital discharge
- This automated method of quantifying peak EA burden should be useful for investigating whether suppressing EA can improve neurologic outcomes and for assessing candidate therapies
Up to half of critically ill patients who undergo electroencephalography exhibit seizures and other epileptiform activity (EA). Small cohort studies have shown that the greater the burden of EA, the greater the probability of neurologic disability and mortality. However, reviewing raw EEG data is so time-consuming that how EA burden affects prognosis has never been quantified in a wide range of neurological, medical and surgical illnesses.
Subscribe to the latest updates from Neuroscience Advances in Motion
As reported in the Journal of Neuroscience Methods, researchers at Massachusetts General Hospital have developed a convolutional neural network—a form of artificial intelligence in which a computer system learns to recognize patterns in visual data—that efficiently labels huge volumes of continuous EEG data. Using that algorithm, they've shown that peak EA burden is more important than the cumulative burden in predicting the level of neurologic disability at hospital discharge.
Sahar F. Zafar, MD, associate medical director of the Neurosciences Intensive Care Unit of the Department of Neurology, Eric S. Rosenthal, MD, director of the Neurosciences Intensive Care Unit, research fellows Jin Jing, PhD, and Wendong Ge, PhD, and M. Brandon Westover, MD, PhD, neurologist and director of the Clinical Data Animation Center at Mass General, and colleagues report the findings in Annals of Neurology.
The researchers studied 1,967 adults hospitalized in the medical, surgical and neurological units at Mass General between September 2011 and February 2017 who underwent continuous EEG. EA burden was calculated in three ways:
- EA burden over the first 24 hours
- Peak burden, defined as maximum EA burden captured within any 12-hour window during the first 72 hours of recording
- Cumulative EA burden over the first 72 hours of recording
Associations with Poor Outcome
- On univariate analysis, all three measures of EA burden were significantly associated with poor outcomes (modified Rankin Scale score of 5 or 6 at hospital discharge)
- In a multivariate model, only the peak EA burden demonstrated a strong independent association with a poor outcome
- The association between peak EA burden and a poor outcome held across three strata based on the time interval between last EEG measurement and hospital discharge: <5 days, 5 to 10 days and >10 days
- Holding other factors equal, increasing the peak EA burden from 0% to 100% increased the probability of a poor neurologic outcome by approximately 35% across the three strata
The dose-dependent relationship between EA burden and poor outcomes was present across diagnostic categories: acute brain injury, hypoxic–ischemic encephalopathy, acute seizures or status epilepticus in the absence of brain injury and primary systemic illness.
Still, the effect was greater in patients who presented with clinical seizures or status epilepticus (average 65% increase in the probability of poor outcome) than in those with the other diagnoses.
This automated method of quantifying peak EA burden should be useful for:
- Large randomized controlled trials investigating whether suppressing EA can improve neurologic outcomes
- High-throughput assessment of candidate therapies
- Studies of the long-term impact of EA on functional and cognitive outcomes, and the long-term effect of antiseizure treatment
Learn more about the Department of Neurosurgery
Refer a patient to the Department of Neurosurgery