Automated Interpretation of Clinical Electroencephalograms Using Artificial Intelligence

Jesper Tveit ¹, Harald Aurlien ^1,2, Sergey Plis ³

¹Holberg EEG, Bergen, Norway.
²Department of Clinical Neurophysiology, Haukeland University Hospital, Bergen, Norway.
³Tri-institutional Center for Translational Research in Neuroimaging and Data Science (TReNDS), Georgia State University, Georgia Institute of Technology, Emory University, Atlanta.
⁴Department of Neurology, Mayo Clinic, Jacksonville, Florida.
⁵Department of Neurology, Beth Israel Deaconess Medical Center, Boston, Massachusetts.
⁶Department of Neurology and Clinical Neurophysiology, St Olavs Hospital, Trondheim University Hospital, Norway.
⁷Department of Clinical Neurophysiology, Stichting Epilepsie Instellingen Nederland (SEIN), Heemstede, the Netherlands.
⁸Department of Neurology, Tel Aviv Sourasky Medical Center and Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel.
⁹Department of Neurology and Pediatrics, Stanford University Lucile Packard Children's Hospital, Palo Alto, California.
¹⁰Department of Clinical Neurophysiology, Danish Epilepsy Centre, Dianalund, Denmark.
¹¹Faculty of Health Sciences, University of Southern Denmark, Odense, Denmark.
¹²Division of Neurology, The Hospital for Sick Children, Toronto, Canada.
¹³Department of Paediatrics, University of Toronto, Toronto, Canada.
¹⁴Department of Neurology, Duke University Medical Center, Durham, North Carolina.
¹⁵Neurodiagnostic Center, Veterans Affairs Medical Center, Durham, North Carolina.
¹⁶Division of Neurology, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania.
¹⁷Department of Pediatrics, Perelman School of Medicine, University of Pennsylvania, Philadelphia.
¹⁸Department of Neurology, Perelman School of Medicine, University of Pennsylvania, Philadelphia.
¹⁹Department of Clinical Neurophysiology, Aarhus University Hospital, Aarhus, Denmark.
²⁰Department of Clinical Medicine, Aarhus University, Aarhus, Denmark.
²¹Department of Neurology, Massachusetts General Hospital, Boston, Massachusetts.
²²Department of Neurology, Oslo University Hospital, Norway.
²³Division of Neurology, Department of Medicine, Krembil Brain Institute, University Health Network, Toronto Western Hospital, University of Toronto, Toronto, Canada.

⁰Holberg EEG, Bergen, Norway.

Jama Neurology +

|

June 20, 2023

View abstract on PubMed

Summary

A new artificial intelligence (AI) model, Standardized Computer-based Organized Reporting of EEG-Artificial Intelligence (SCORE-AI), achieved human expert performance in interpreting electroencephalograms (EEGs). This AI tool can improve diagnosis and patient care, especially in underserved regions.

Area Of Science

Neurology
Artificial Intelligence
Medical Diagnostics

Background

Electroencephalograms (EEGs) are crucial for neurological diagnosis but require specialized expertise.
Existing AI models offer limited EEG interpretation capabilities.
A comprehensive, automated EEG interpretation tool is needed for clinical practice.

Purpose Of The Study

To develop and validate the SCORE-AI (Standardized Computer-based Organized Reporting of EEG-Artificial Intelligence) model.
To enable AI-driven classification of EEG recordings into normal/abnormal and specific abnormality categories.

Main Methods

A convolutional neural network (SCORE-AI) was developed using 30,493 routine EEG recordings.
The model was validated on three independent datasets totaling 10,145 EEGs.
Performance was assessed against expert interpretations and external benchmarks.

Main Results

SCORE-AI demonstrated high accuracy (AUC 0.89-0.96) in classifying EEG abnormalities.
The AI model's performance was comparable to that of human experts.
SCORE-AI significantly outperformed three previous AI models in detecting epileptiform abnormalities.

Conclusions

SCORE-AI achieves human-level performance for fully automated routine EEG interpretation.
This AI application can enhance diagnostic accuracy and patient care, particularly in resource-limited settings.
SCORE-AI has the potential to improve efficiency and consistency in epilepsy centers.