Loading...
These investigators asked whether artificial intelligence can interpret routine electroencephalograms (EEGs) with sufficient diagnostic accuracy, compared with a consensus of human experts, to guide clinical decision making. For this industry-funded study, they developed the SCORE-AI model using a deep-learning program trained on >30,000 highly annotated EEGs using the Standardized Computer-based Organized Reporting of EEG system. SCORE-AI was trained to classify EEGs into 6 categories: normal, epileptiform-generalized, epileptiform-focal, nonepileptiform-generalized, nonepileptiform-focal, and multiple abnormalities. The model was then validated against a hold-out data set of more than 2500 EEGs not included in the development phase and 3 …