Seizure burden, outlined by a man-made intelligence (AI) algorithm utilized to point-of-care electroencephalography (POC EEG) recordings, might help predict practical outcomes.
After related cofactors had been managed for, increased seizure burden correlated with poorer practical outcomes. All the sufferers within the research had been being monitored as a part of their customary of care owing to suspicion of seizures or as a result of they had been in danger for seizures, research investigator Masoom Desai, MD, with the Division of Neurology, College of New Mexico, Albuquerque, instructed Medscape Medical Information, and the outcomes had been “compelling.”
“Our research addresses the vital want for automation in monitoring epileptic exercise and seizure burden,” Desai added throughout a press briefing on the American Academy of Neurology (AAN) 2024 Annual Assembly.
A Pivotal Shift
“A number of many years of analysis have highlighted the numerous correlation between seizure burden and unfavorable outcomes each in grownup and pediatric populations,” mentioned Desai.
Nevertheless, the normal technique of manually decoding EEGs to establish seizures and their related burden is a “complicated and time-consuming course of that may be topic to human error and variability,” she famous.
POC EEG is a rapid-access, reduced-montage EEG system that, when paired with an automatic machine studying software referred to as Readability (Ceribell, Inc; Sunnyvale, CA), can monitor and analyze seizure burden in actual time.
The algorithm incorporates a complete record of EEG options which have been related to outcomes. It analyzes EEG exercise each 10 seconds from all EEG channels and calculates a seizure burden prior to now 5 minutes for the affected person. The upper the seizure burden, the extra time the affected person has spent in seizure exercise.
Amongst 344 individuals with POC EEG (imply age, 62 years, 45% ladies) within the SAFER-EEG trial, 178 (52%) had seizure burden of zero all through the recording and 41 (12%) had suspected standing epilepticus (most seizure burden ≥ 90%).
Earlier than adjustment for scientific covariates, there was a big affiliation between excessive seizure burden and unfavorable outcomes.
Particularly, 76% of sufferers with a seizure burden ≥ 50% had an unfavorable modified Rankin Scale rating of ≥ 4 at discharge and an analogous proportion was discharged to long-term care services, she famous.
After adjustment for related scientific covariants, sufferers with a excessive seizure burden (≥ 50 or > 90%) had a fourfold improve in odds of an unfavorable modified Rankin Scale rating in contrast with these with no seizure burden.
Excessive seizure burden current within the final quarter of the recording was notably indicative of unfavorable outcomes (fivefold elevated odds), “suggesting the vital timing of seizures and its affect on affected person prognosis,” Desai famous.
‘Profound Implications’
“The implications of our analysis are profound, indicating a pivotal shift in direction of integrating AI and machine learning-guided automated EEG interpretation in administration of critically sick sufferers with seizures,” she added.
“As we transfer ahead, our analysis will focus on making use of this superior software in scientific choice making in scientific observe, analyzing the way it can steer remedy choices for sufferers, with the final word purpose of enhancing affected person care and bettering outcomes for these affected by these neurological challenges,” Desai mentioned.
Briefing moderator Paul M. George, MD, PhD, chair of the AAN science committee, famous that this summary was one among three featured on the “prime science” press briefing themed “advancing the bounds of neurologic care,” as a result of it represents an “modern technique” of utilizing new expertise to enhance understanding of neurologic situations.
George mentioned this expertise “could possibly be notably helpful in settings with few scientific specialists. It is going to be thrilling to see as this unfolds, the place it could possibly information possibly the ED physician or major care doctor to assist enhance affected person care.”
On that word, George cautioned that it is nonetheless “early within the discipline” of utilizing AI to information decision-making and it will likely be essential to collect extra info to verify that “machine studying algorithms might help information physicians in treating sufferers with neurologic situations.”
Funding for the research was supplied by the College of Wisconsin-Madison and Ceribell, Inc. Desai obtained funding from Ceribell for this challenge. George has no related disclosures.