Table 1.

Summary of the AI model’s performance for motion artifact detection in CTPA exams from the three participating sites (A, B, C). (key: +, present; -, absent; CI, confidence interval).

		Training	Test	Sensitivity	Specificity	Accuracy	AUC	95% CI (AUC)
Model validation (training + validation data from all sites)	Motion +	295	126	94%	91%	93%	0.93	0.89–0.97
	Motion -	259	113
Model testing (training from A, C, and testing on B)	Motion +	196	173	85%	90%	86%	0.87	0.82–0.92
	Motion -	231	86