Gao 2022.

Study characteristics
Patient Sampling	Single‐centre, retrospective, case‐control study, in the Affiliated Eye Hospital of Wenzhou Medical University, China. A total of 208 participants (1040 corneal topography images) were evaluated and divided into the following 3 groups. Keratoconus Subclinical keratoconus Healthy controls Data were collected between 2012 and 2018 using the Pentacam system and analysed from February 2019 to December 2021. The corneal data of pachymetry and elevation were exported from the Pentacam HR system (Oculus, Optikgeräte GmbH, Wetzlar, Germany).
Patient characteristics and setting	Each image was previously assigned to the following 3 groups. Normal (70 eyes; mean age 28.7 (SD 2.6) years) Subclinical keratoconus (48 eyes; mean age 24.6 (SD 5.7) years) Keratoconus (90 eyes; mean age, 25.9 (SD 5.4) years). The data set was randomly split into 70% for training and 30% for testing.
Index tests	Neural network, called Keratoscreen. The preprocessing model separately established Zernike coefficients (ZC) data sets of 5 corneal surface maps: the anterior corneal curvature (DA), posterior corneal curvature (DP), anterior corneal elevation (DAE), posterior corneal elevation (DPE), and corneal thickness data (pachymetry DPAC). The first N (N #15) ZC terms (5 orders, when N = 15) obtained from each map were used to form ZC data sets. KeratoScreen, with L = 4 layers (1 input layer, 2 hidden layers, and 1 output layer), was used in this study. The neurons on the input layer have N nodes that take an input vector ZC and outputs through 2 hidden layers. To train the algorithm they used 20 repeats with the randomly generated training and test sets, and the results were averaged. However, it is unclear whether they reused the same data for training and testing.
Target condition and reference standard(s)	For the keratoconus group, 1 of the following on slit‐lamp biomicroscopy: Vogt's striae, stromal thinning, Fleischer's ring, or corneal scarring. The central average keratometry was > 47.0 D. Asymmetric topographical features included I‐S values >2.0 D of the vertical power gradient across the central 6‐mm region. The subjects had no history of wearing contact lenses, ocular surgery, or extensive scarring. The normal eye group was selected using the following criteria: no clinical symptoms of keratoconus or subclinical keratoconus in either eye, central average keratometry < 45.0 D, I‐S value of corneal topographical features < 1.4 D at the vertical power gradient across the 6‐mm region, the spherical equivalent was within 66.0 D, and astigmatism was within 2 D. The subjects also had no ocular disease, no prior ophthalmic surgery, and no cornea ectasia on a slit lamp. Only 1 eye was randomly selected from normal subjects for analysis. The subclinical keratoconus group was selected from the participants with undiagnosed eyes of unilateral keratoconus, based on the following criteria: 1 eye was diagnosed with keratoconus; in the contra, there were no clinical symptoms of keratoconus during a slit‐lamp examination, retinoscopy, or ophthalmoscopy; the central average keratometry in this eye group was 45.0 D; the I‐S value of the corneal topographical features was > 1.4 D at the vertical power gradient through the central 6‐mm region; the spherical equivalent was within 66.0 D; astigmatism was within 2 D; and these subjects had no history of contact lens wear or ocular surgery.
Flow and timing	All cases were included in the reference standard and index test. All data were included in a 2 × 2 table.
Comparative	Not applicable
Notes	The research received funding through 21NDJC309YBM from the Zhejiang Philosophy and Social Science Planning Project (China), LZ21F020008 from the Natural Science Foundation of Zhejiang Province (China), 2019C03045 from the National Key Project of Research and Development Program of Zhejiang Province (China), and LY21A040001 from the Natural Science Foundation of Zhejiang Province (China).
Methodological quality
Item	Authors' judgement	Risk of bias	Applicability concerns
DOMAIN 1: Patient selection
Was a consecutive or random sample of patients enrolled?	Unclear
Was a case‐control design avoided?	No
Did the study avoid inappropriate exclusions?	No
Could the selection of patients have introduced bias?		High risk
Are there concerns that the included patients and setting do not match the review question?			High
DOMAIN 2: Index test (All tests)
Were the index test results interpreted without knowledge of the results of the reference standard?	Yes
If a threshold was used, was it pre‐specified?	Unclear
Was the model designed in an appropriate manner?	Unclear
Could the conduct or interpretation of the index test have introduced bias?		Unclear risk
Are there concerns that the index test, its conduct, or interpretation differ from the review question?			Low concern
DOMAIN 3: Reference standard
Is the reference standard likely to correctly classify the target condition?	Unclear
Were the reference standard results interpreted without knowledge of the results of the index tests?	Yes
Could the reference standard, its conduct, or its interpretation have introduced bias?		Unclear risk
Are there concerns that the target condition as defined by the reference standard does not match the question?			Low concern
DOMAIN 4: Flow and timing
Did all patients receive the same reference standard?	Unclear
Were all patients included in the analysis?	Yes
Could the patient flow have introduced bias?		Unclear risk
DOMAIN 5: Comparative
Were different AI tests were developed and interpreted without knowledge of each other.
Are the proportions and reasons for missing data similar for all index tests?