This content is only available for ASCRS Members
This content from the 2020 ASCRS Virtual Annual Meeting is only available to ASCRS members. To log in, click the teal "Login" button in the upper right-hand corner of this page.
Filed Under
Refractive
surgical planning
preoperative
keratoconus
artificial intelligence
2020 paper presentation
Purpose
To assess the efficiency of a mathematical algorithm to measure differences in the corneal tomography findings between subclinical keratoconus in one eye and healthy corneas.
Methods
In total, 92 eyes of 92 patients with very asymmetric ectasia accompanied by normal topography in one eye and very asymmetric ectasia with ectasia (VAE-E) in the fellow eye were included in the subclinical keratoconus (KC) group (VAE-NT G). Further, 323 eyes of 323 patients who underwent LVC were allocated to healthy corneas into the control group (CG). The KCG group including 129 eyes of 129 patients with KC. Artificial intelligence models were generated using Pentacam HR variables to discriminate among the 3 groups based on the preoperative data. A mathematical algorithm (MA) was built using support vector machine to identify a cutoff point to distinguish among patients from the 3 groups.
Results
The derived MA showed good accuracy with 100% sensitivity and 100% specificity for KCG (cutoff point ≤2.52). Regarding the VAE-NT G patients, MA showed an area under the curve (AUC) of 0.952 (84.8% sensitivity; 93.8% specificity; cutoff point ≤ 7.36), which was significantly greater than the AUC for the Belin–Ambrosio Display (AUC = 0.908; 76.1% sensitivity; 93.2% specificity; P = 0.0005).
Conclusion
The parameters derived from corneal tomography aided in the early detection of subclinical keratoconus. The derived MA derived showed good sensitivity, specificity, and accuracy to distinguish patients with subclinical keratoconus or KC from individuals with healthy corneas.
To assess the efficiency of a mathematical algorithm to measure differences in the corneal tomography findings between subclinical keratoconus in one eye and healthy corneas.
Methods
In total, 92 eyes of 92 patients with very asymmetric ectasia accompanied by normal topography in one eye and very asymmetric ectasia with ectasia (VAE-E) in the fellow eye were included in the subclinical keratoconus (KC) group (VAE-NT G). Further, 323 eyes of 323 patients who underwent LVC were allocated to healthy corneas into the control group (CG). The KCG group including 129 eyes of 129 patients with KC. Artificial intelligence models were generated using Pentacam HR variables to discriminate among the 3 groups based on the preoperative data. A mathematical algorithm (MA) was built using support vector machine to identify a cutoff point to distinguish among patients from the 3 groups.
Results
The derived MA showed good accuracy with 100% sensitivity and 100% specificity for KCG (cutoff point ≤2.52). Regarding the VAE-NT G patients, MA showed an area under the curve (AUC) of 0.952 (84.8% sensitivity; 93.8% specificity; cutoff point ≤ 7.36), which was significantly greater than the AUC for the Belin–Ambrosio Display (AUC = 0.908; 76.1% sensitivity; 93.2% specificity; P = 0.0005).
Conclusion
The parameters derived from corneal tomography aided in the early detection of subclinical keratoconus. The derived MA derived showed good sensitivity, specificity, and accuracy to distinguish patients with subclinical keratoconus or KC from individuals with healthy corneas.
View More Presentations from this Session
This presentation is from the session "SPS-114 Keratorefractive Surgical Planning" from the 2020 ASCRS Virtual Annual Meeting held on May 16-17, 2020.