Artificial intelligence for detecting keratoconus

Magali Vandevenne, Eleonora Favuzza, Mitko Veta, E. Lucenteforte, Tos Berendschot, Rita Mencucci, Rudy Nuijts, Gianni Virgil, Mor M. Dickman*

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review


This is a protocol for a Cochrane Review (diagnostic). The objectives are as follows: The primary objective is to assess the diagnostic accuracy of AI algorithms in the detection of keratoconus in patients presenting with refractive errors, especially those whose vision can no longer be corrected fully with glasses, patients seeking corneal refractive surgery or those suspected of having keratoconus. AI could help ophthalmologists, optometrists and other eye-care professionals to make decisions on referral to cornea specialists for these patients.

Secondary objectives
To compare different AI algorithms, e.g. neural networks, decision trees, support vector machines.
To assess potential causes of heterogeneity in diagnostic performance across studies, according to the following:
• index test methodology: pre-processing techniques, core AI method and postprocessing techniques;
• sources of input to train algorithms: topography and tomography images from Placido-disc system or Scheimpflug system or slitscanning system or OCT, number of training and testing cases/images, label/endpoint variable used for training;
• study setting;
• study design, retrospective or prospective studies;
• ethnicity, or geographic area as its proxy; • different index test positivity criteria provided by topography or tomography device;
• reference standard used, topography or tomography, one or two cornea-specialists;
• definition of keratoconus used;
• mean age;
• patient recruitment;
• severity of keratoconus:
◦ clinically manifest keratoconus
◦ subclinical keratoconus
Original languageEnglish
Article numberCD014911
Number of pages17
JournalCochrane Database of Systematic Reviews
Issue number12
Publication statusPublished - 2021



Cite this