Seizure burden, defined by an artificial intelligence (AI) algorithm applied to point-of-care electroencephalography (POC EEG) recordings, can help predict functional outcomes.
After relevant cofactors were controlled for, higher seizure burden correlated with poorer functional outcomes. All of the patients in the study were being monitored as part of their standard of care owing to suspicion of seizures or because they were at risk for seizures, study investigator Masoom Desai, MD, with the Department of Neurology, University of New Mexico, Albuquerque, told Medscape Medical News, and the results were “compelling.”
“Our study addresses the critical need for automation in monitoring epileptic activity and seizure burden,” Desai added during a press briefing at the American Academy of Neurology (AAN) 2024 Annual Meeting.
A Pivotal Shift
“Several decades of research have highlighted the significant correlation between seizure burden and unfavorable outcomes both in adult and pediatric populations,” said Desai.
However, the traditional method of manually interpreting EEGs to identify seizures and their associated burden is a “complex and time-consuming process that can be subject to human error and variability,” she noted.
POC EEG is a rapid-access, reduced-montage EEG system that, when paired with an automated machine learning tool called Clarity (Ceribell, Inc; Sunnyvale, CA), can monitor and analyze seizure burden in real time.
The algorithm incorporates a comprehensive list of EEG features that have been associated with outcomes. It analyzes EEG activity every 10 seconds from all EEG channels and calculates a seizure burden in the past 5 minutes for the patient. The higher the seizure burden, the more time the patient has spent in seizure activity.
Among 344 people with POC EEG (mean age, 62 years, 45% women) in the SAFER-EEG trial, 178 (52%) had seizure burden of zero throughout the recording and 41 (12%) had suspected status epilepticus (maximum seizure burden ≥ 90%).
Before adjustment for clinical covariates, there was a significant association between high seizure burden and unfavorable outcomes.
Specifically, 76% of patients with a seizure burden ≥ 50% had an unfavorable modified Rankin Scale score of ≥ 4 at discharge and a similar proportion was discharged to long-term care facilities, she noted.
After adjustment for relevant clinical covariants, patients with a high seizure burden (≥ 50 or> 90%) had a fourfold increase in odds of an unfavorable modified Rankin Scale score compared with those with no seizure burden.
High seizure burden present in the last quarter of the recording was particularly indicative of unfavorable outcomes (fivefold increased odds), “suggesting the critical timing of seizures and its impact on patient prognosis,” Desai noted.
‘Profound Implications’
“The implications of our research are profound, indicating a pivotal shift towards integrating AI and machine learning-guided automated EEG interpretation in management of critically ill patients with seizures,” she added.
“As we move forward, our research will concentrate on applying this advanced tool in clinical decision making in clinical practice, examining how it can steer treatment decisions for patients, with the ultimate goal of enhancing patient care and improving outcomes for those affected by these neurological challenges,” Desai said.
Briefing moderator Paul M. George, MD, PhD, chair of the AAN science committee, noted that this abstract was one of three featured at the “top science” press briefing themed “advancing the limits of neurologic care,” because it represents an “innovative method” of using new technology to improve understanding of neurologic conditions.
George said this technology “could be particularly useful in settings with few clinical specialists. It will be exciting to see as this unfolds, where it can guide maybe the ED doctor or primary care physician to help improve patient care.”
On that note, George cautioned that it’s still “early in the field” of using AI to guide decision-making and it will be important to gather more information to confirm that “machine learning algorithms can help guide physicians in treating patients with neurologic conditions.”
Funding for the study was provided by the University of Wisconsin-Madison and Ceribell, Inc. Desai received funding from Ceribell for this project. George has no relevant disclosures.
>>> Read full article>>>
Copyright for syndicated content belongs to the linked Source : Medscape – https://www.medscape.com/viewarticle/compelling-results-ai-eeg-predict-functional-stroke-outcomes-2024a100075x