Automated Interpretation of Clinical Electroencephalograms Using Artificial Intelligence

The study developed and validated an AI model called SCORE-AI for interpreting EEG recordings. The model's diagnostic accuracy, sensitivity, and specificity were compared to human experts and an external reference standard. The results showed that SCORE-AI achieved high accuracy, similar to human experts, with an area under the receiver operating characteristic curve ranging from 0.89 to 0.96. Compared to three previously published AI models, SCORE-AI outperformed them in detecting epileptiform abnormalities. The findings suggest that SCORE-AI can provide automated interpretation of routine EEGs at a level comparable to human experts.