Table 2. Average Accuracy of SCORE-AI and of the Human Experts With Respect to the Human Expert Majority Consensus on 100 EEGs From the Multicenter Test Data Set.

EEG recording category	Average accuracy (95% CI)		Difference (P value)
	SCORE-AI	Human experts
Normal	95.00 (89.61-97.88)	91.36 (88.04-94.10)	.09
Epileptiform-focal	84.69 (76.73-90.54)	88.4 (84.35-91.91)	.12
Epileptiform-generalized	94.9 (89.41-97.83)	95.36 (92.51-97.48)	.34
Nonepileptiform-diffuse	84.69 (76.63-90.83)	86.09 (81.99-89.66)	.33
Nonepileptiform-focal	85.71 (77.86-91.41)	85.25 (81.04-88.78)	.47
Exact match/multiple abnormalities	65.31 (54.93-73.60)	66.7 (60.56-72.41)	.33

Abbreviations: EEG, electroencephalography; SCORE-AI, Standardized Computer-based Organized Reporting of EEG–Artificial Intelligence.