Table 1. Gwet AC1 Agreement Coefficients for the 11 Human Experts, SCORE-AI, and the Human Expert Majority Consensus.
EEG recording category | Agreement coefficient (95% CI) | |
---|---|---|
Agreement among the human experts | Agreement between SCORE-AI and majority consensus of human experts | |
Normal | 0.723 (0.649-0.796)a | 0.903 (0.820-0.987)a |
Epileptiform-focal | 0.723 (0.643-0.803) | 0.757 (0.634-0.880) |
Epileptiform-generalized | 0.901 (0.854-0.949) | 0.928 (0.865-0.991) |
Nonepileptiform-diffuse | 0.630 (0.539-0.721) | 0.738 (0.608-0.868) |
Nonepileptiform-focal | 0.587 (0.499-0.674) | 0.775 (0.657-0.893) |
Exact match/multiple abnormalities | 0.497 (0.433-0.561)a | 0.689 (0.611-0.766)a |
Abbreviations: EEG, electroencephalography; SCORE-AI, Standardized Computer-based Organized Reporting of EEG–Artificial Intelligence.
Significant difference. Statistical comparisons were based on the 95% CIs. Significance means there was no overlap between the 95% CIs.