The Performance of Artificial Intelligence in Cervical Colposcopy: A Retrospective Data Analysis

Table 2

The sensitivity, specificity, positive predictive value, negative predictive value, and AUCs of the human colposcopists, AI system, and AI-assisted colposcopy.

	Sensitivity % (95% CI)	Specificity % (95% CI)	Positive predictive value % (95% CI)	Negative predictive value % (95% CI)	AUC (95% CI)
CIN2+
Human colposcopists	88.8 (80.5, 93.8)	59.5 (53.4, 65.4)	43.2 (36.2, 50.4)	93.9 (89.1, 96.6)	0.741 (0.686, 0.797)
AI system	95.5 (89.0, 98.2)	57.6 (51.5, 63.5)	43.8 (37.0, 50.9)	97.4 (93.4, 99.0)	0.765 (0.715, 0.816)
AI-assisted colposcopy	96.6 (90.6, 98.9)	38.1 (32.4, 44.2)	35.1 (29.4, 41.3)	97.0 (91.6, 99.0)	0.674 (0.616, 0.731)
CIN3+
Human colposcopists	92.6 (84.8, 96.6)	59.2 (53.2, 65.0)	41.0 (34.1, 48.2)	96.3 (92.2, 98.3)	0.759 (0.706, 0.812)
AI system	97.5 (91.4, 99.3)	56.6 (50.6, 62.4)	40.7 (34.1, 47.8)	98.7 (95.3, 99.6)	0.674 (0.616, 0.733)
AI-assisted colposcopy	97.5 (91.4, 99.3)	37.4 (31.8, 43.3)	32.2 (26.7, 38.3)	98.0 (93.1, 99.5)	0.771 (0.721, 0.820)

AI, artificial intelligence; AUC, areas under the curve; CIN, cervical intraepithelial neoplasia; 95% CI, 95% confidence intervals.