• Center on Health Equity & Access
  • Clinical
  • Health Care Cost
  • Health Care Delivery
  • Insurance
  • Policy
  • Technology
  • Value-Based Care

AI Model Interprets EEG Readings for Epilepsy With Human-Level Accuracy, Study Finds

Article

A new artificial intelligence (AI) model may improve patient diagnosis and care among specialized epilepsy centers in underserved areas.

The use of an artificial intelligence (AI) model to interpret routine clinical electroencephalograms (EEGs) was able to achieve diagnostic accuracy in detecting and identifying epilepsy readings, similar to that of human experts, according to a new study.

“These results warrant clinical implementation with a potential to improve patient care in remote and underserved areas where EEG expertise is scarce or unavailable,” wrote the researchers of the study. In addition, SCORE-AI [Standardized Computer-based Organized Reporting of EEG-Artificial Intelligence] may help improve efficiency and reduce excessive workloads for experts in tertiary care centers who regularly interpret high volumes of EEG recordings.”

This clinical validation study is published in JAMA Neurology.

Although previous AI models have been able to detect normal and abnormal epileptiform activity, the researchers believe that a fully automated AI model is needed in routine clinical practice. To the researchers knowledge, this AI model is the first of its kind to complete a fully automated and comprehensive clinically relevant assessment of routine EEGs.

The study investigators developed SCORE-AI with the aim of being able to distinguish between normal and abnormal EEG readings, as well as classify abnormal readings into clinical decision-making categories: epileptiform-focal, epileptiform-generalized, nonepileptiform-focal, and nonepiletiform-diffuse.

SCORE-AI was developed and validated using 30,493 anonymized EEGs recorded from 2014 to 2020, which were analyzed from January 17, 2022, to November 14, 2022, and annotated by 17 experts. All patients 3 months and older and not critically ill were eligible for the study. Of the patients in this dataset, 14,980 were male and the median age was 25.3 (95% CI, 1.3-76.2) years.

Additionally, SCORE-AI was validated using 3 independent datasets: a multicenter dataset of 100 patients, in which 61 were male patients and the median age was 25.8 (95% CI, 4.1-85.5) years, evaluated by 11 experts; a single-center dataset of 9785 patients, of which 5168 were male patients and the median age was 35.4 (95% CI, 0.6-87.4) years, evaluated by 14 experts; and a benchmarking dataset of 60 patients from 3 published AI models using external reference standards, with 27 male patients and median age of 23 (95% CI, 3-75) years.

The researchers found that SCORE-AI was able to achieve high accuracy, between 0.89 and 0.96 under the receiver operative characteristic curve, for identifying the different categories of abnormal EEG readings, similar to that of humans. Furthermore, benchmarking against 3 previously published AI models, SCORE-AI achieved higher accuracy (88.3%; 95% CI, 79.2%-94.9%; P < .001).

The researchers acknowledge some limitations to the study, such as that it did not include newborns or critically ill patients and the model was trained to identify biomarkers interpreted by experts.

Despite these limitations, the researchers believe the study shows how AI can be used to improve patient diagnosis and care, especially among underserved areas.

“Its application may help to provide useful clinical information in remote and underserved areas where expertise in EEG interpretation is minimal or unavailable,” wrote the researchers. “Importantly, it may also help reduce the potential for EEG misinterpretation and subsequent mistreatment, improve interrater agreement to optimize routine interpretation by neurologists, and increase efficiency by decompressing excessive workloads for human experts interpreting high volumes of EEGs.”

Reference

Tveit J, Aurlien H, Plis S, et al. Automated interpretation of clinical electroencephalograms using artificial intelligence. JAMA Neurol. Published online June 20, 2023. doi:10.1001/jamaneurol.2023.1645

Related Videos
Masanori Aikawa, MD
Dr Bonnie Qin
Dr Bonnie Qin
Glenn Balasky, executive director of the Rocky Mountain Cancer Center.
Screenshot of Stephanie Hsia, PharmD
Cesar Davila-Chapa, MD
Screenshot of an interview with Nadine Barrett, PhD
Female doctor in coat with stethoscope on blue background - Pixel-Shot - stock.adobe.com
Io Hui, PhD, researcher at The University of Edinburgh
Related Content
© 2024 MJH Life Sciences
AJMC®
All rights reserved.