Barcelona Esquerra Glaucoma Artificial Intelligence-based Screening Program (BEGAS)

Overview

Two primary care-based screening systems will be tested to identify subjects with referrable glaucoma to hospital care. Subjects between 45 to 64 years old living in the metropolitan area of Barcelona will be invited to participate in a one-time visit, with an optic disc examination and intraocular pressure (IOP). The criteria for referring a patient will be the detection of glaucoma but with two different approaches depending on which Integrated Practice Unit (IPU) the patients will be allocated to: one arm using an Artificial Intelligence (AI) reading software of the optic disc picture; and the other one will base their referral after an ophthalmic examination performed by an ophthalmologist. In both circuits, an optic nerve head photography will be obtained, and a masked reading center will be established to determine the ground truth for diagnosis. This screening trial will explore the level of agreement between both systems and the cost-effectiveness of each of them. Secondary analyses will include potential diagnostic composite scores (including other ancillary tests, such as optical coherence tomography images, that could maximize the screening process); the identification of population and disease characteristics (type of glaucoma, intraocular pressure) that could increase the effectivity and adherence to the screening process.

The purpose of this study is twofold: to validate in our population an Artificial Intelligence (AI) reading software of the optic disc picture, after comparing the estimated result (glaucoma/suspect/normal) to the ground truth; and to conduct a clinical trial where the level of agreement between both systems and the cost-effectiveness of each of them will be tested In the first phase, a set of patients from our reference population will be selected. A standard-of-care ophthalmic examination with the usual ancillary tests to confirm or rule out the presence of glaucoma (including an optic disc retinography), will be performed. The patient (and the test) will be examined by a glaucoma specialist who will determine the status of the patient. Then, the retinography will be analyzed by the AI software, providing the estimated result (glaucoma/suspect/normal). The level of agreement between the ground truth and the casted result will confirm the diagnostic accuracy. In the second phase, a second set of patients will be recruited. In this case, the patients will be randomly allocated to either of the two arms of the study: In arm A the ancillary tests (including the retinography) will be performed, and the software will analyze the retinography, therefore providing the glaucoma status result. In arm B, the patients (and the test) will be examined by a glaucoma specialist who will then determine the status of the patient. All the patients, irrespective of the diagnosis and the arm of the study will be then explored by another glaucoma specialist (reading center), who will be blinded to where the diagnosis comes from (AI software or glaucoma specialist), to the determine the level of agreement between the two screening systems