Table 3.

Articles on artificial intelligence application for CC screening.

Studies	Population	Objective	Device	Intervention	Results
Kudva et al., 2018 [44]	India >24 y n = 102	Develop a decision support system for cervical cancer screening with an inbuilt image processing algorithm.	Android device with a camera of 13 Mpx.	102 images Reference = expert evaluation.	Accuracy 97.9%, Se 99.0%, Sp 97.1%, AUC NR.
Bae et al., 2020 [45]	South Korea, >20 y n = 20	Develop a new cervical cancer screening technique and implement a machine-learning algorithm using images taken during VIA with a smartphone-based endoscope.	Smartphone-based endoscope.	40 images (2 per patient). Expert evaluation vs AI. Reference = histopathology.	Accuracy 78.3%, Se 75.8%, Sp 80.3%, AUC 0.805. Clinicians’ mean accuracy 77.5%, Se 62.5%, Sp 100%, AUC NR.
Xue Z. et al., 2020 [43]	Various countries >18 y n = 3221	Evaluate accuracy of automated visual evaluation (AVE) on smartphone images.	MobileODT system (smartphone with lens).	7587 images. Reference = expert evaluation	Accuracy NR, Se NR, Sp NR, AUC 0.87 (95% CI 0.81–0.92).
Viñals et al., 2021 [46]	Cameroon, Switzerland 30–49 y n = 44	Development of a smartphone-based algorithm to detect cervical precancer from the dynamic features (dynamics of aceto-whitening).	Samsung Galaxy S5	44 dynamic images; Expert evaluation vs. AI. Reference = histology	AI accuracy 89%, Se 90%, Sp 87%, AUC NR. Clinicians’ mean accuracy 71%, Se 68%, Sp 78%, AUC NR.

Abbreviations: AI (artificial intelligence), AUC (area under the curve), Mpx (megapixels), NR (not reported), Se (sensitivity), Sp (specificity), VIA (visual inspection with acetic acid).