Development of a Keratoconus Detection Algorithm by Deep Learning Analysis and Its Validation on Eyestar Images

• Conditions: Cataract; Corneal Ectasia; Keratoconus; Eye Diseases
• Interventions: Device: Corneal tomography with Eyestar 900; Other: retrospective analysis, no intervention; Device: Corneal tomography with Pentacam; Device: Biometry with IOL-Master

• Registration Number: NCT04763785
• Lead Sponsor: Insel Gruppe AG, University Hospital Bern

Brief Summary

Monocentric clinical study to develop an imaging analysis algorithm for the Eyestar 900 to identify keratoconus corneas and improve biometry for intraocular lens calculations

Detailed Description

Keratoconus is a progressive corneal ectatic disorder, characterised by thinning, protrusion and irregularity. Corneal imaging is crucial in keratoconus detection and progression analysis. Detection of keratoconus in early stages is important and has therapeutic consequence, whether to plan a surgical intervention or calculating an intraocular lens, before cataract surgery, as standard lens calculation techniques may lead to wrong results in patients with a keratoconus.

The Eyestar 900 is a swept-source OCT biometer and has the potential to be used for early keratoconus identification and progression analysis.

Study Timeline

• Study First Submitted: 2021-02-14
• Study First Submitted QC: 2021-02-18
• Study First Posted: 2021-02-21
• Study Start: 2021-05-11
• Status Verified: 2021-06
• Last Update Submitted: 2021-09-29
• Last Update Posted: 2021-09-30
• Primary Completion: 2023-06-01
• Study Completion: 2023-12-01

Recruitment & Eligibility

• Status: UNKNOWN
• Sex: All
• Target Recruitment: 4800

Inclusion Criteria

1. Patients with all stages of keratoconus
2. Patients with healthy corneas

Exclusion Criteria

1. Keratoconus patients with hydrops, status following hydrops
2. Patients with degenerative corneal diseases
3. Patients after corneal surgery

Study & Design

• Study Type: OBSERVATIONAL
• Study Design: Not specified

Arm && Interventions

Group	Intervention	Description
participants with healthy corneas	Corneal tomography with Pentacam	Corneal tomography on healthy participants
Patients with keratoconus corneas	Corneal tomography with Eyestar 900	Corneal tomography on patients with keratoconus diagnosis
participants with healthy corneas	Corneal tomography with Eyestar 900	Corneal tomography on healthy participants
participants with healthy corneas	Biometry with IOL-Master	Corneal tomography on healthy participants
Patients with keratoconus corneas	Corneal tomography with Pentacam	Corneal tomography on patients with keratoconus diagnosis
Patients with keratoconus corneas	Biometry with IOL-Master	Corneal tomography on patients with keratoconus diagnosis
retrospective part	retrospective analysis, no intervention	fully anonymised Picture data of existing 4500 patients

Primary Outcome Measures

Name	Time	Method
Keratoconus identification	2.5 years	Classification accuracy of the keratoconus identification algorithm for the Eyestar device in comparison to the gold standard (Belin-Ambrosio Enhanced Extasia Deviation Index) BAD_D in Pentacam images.

Secondary Outcome Measures

Name	Time	Method
Feasibility in clinical practice	2.5 years	Evaluation of the feasibility (percentage of valid measurements without errors and/or problems in image aquisition) of cornea measurements in keratoconus and healthy eyes.

Trial Locations

Locations (1)

Universitätsklinik für Augenheilkunde, Inselspital; 🇨🇭 Bern, Switzerland