Daytona

A total of 36,070 UWF images (19,684 individuals) were collected from the CMAAI, which is a union of medical organisations, computer science research groups and related enterprises in the field of AI with the aim of improving the research and translational applications of AI in medicine. The CMAAI dataset includes 15322 images obtained from Shenzhen Eye Hospital, 7387 images from Huazhong screening program, 4929 images from Eastern Guangdong Eye Study and 8432 images from Southern China Guangming Screening program. These images were from individuals who presented for retinopathy examinations, ophthalmology consultations, or routine ophthalmic health evaluations, and were obtained between June 2016 and September 2019 using an OPTOS nonmydriatic camera (OPTOS Daytona, Dunfermline, UK) and 200-degree fields of view. Participants were examined without mydriasis. All UWF images were deidentified before they were transferred to research investigators. This study was approved by the Institutional Review Board of ZOC (identifier, 2019KYPJ107) and conducted in accordance with the tenets of the Declaration of Helsinki. Informed consent was exempted due to the retrospective nature of the
data collection and the use of deidentified UWF images.

Fundus photography was done using a widefield imaging device (Optos Daytona, Optos PLC, Dunfermline, UK), and OCT-A with Cirrus 6000OCT-Angiography scans (Zeiss OCT, Carl Zeiss Meditec, Jena, Germany) with 3 × 3 mm and 6 × 6 mm foveal-centered scan patterns at an acquisition frequency of 100,000 A-scans per second, a wavelength centered on 840 nm, and axial and transverse resolution of 5 µm and 12 µm, respectively. Motion signal contrast was captured using complex-based OCT-microangiography (OMAG) technology software that coanalyses both signal phase and signal intensity. All images were produced by a single appropriately trained operator (‘TR’). The built-in device software segmented the superficial retinal plexus and deep retinal plexus, and segmentation was checked after each image acquisition and manually corrected if necessary.
Any OCT-A images presenting significant artefacts (motion, projection, doubling of the retinal vessels, stretching defect, etc.) were excluded.
Furthermore, all images with an SSI < 7 were excluded. The workflow for all images used eye-tracking software, a signal motion correction algorithm, and an integrated projection artefact correction tool.
DR staging was conducted by two skilled ophthalmologists (‘TR’ & ‘PG’) working blind on the ultra-widefield imaging and was based on the Airlie House classification system [13 (link)].

This retrospective cross-sectional study involved patients evaluated through a DR screening program at Zuckerberg San Francisco General Hospital and Trauma Center from April 2018 to September 2019. This study was approved by the Human Research Protection Program at the University of California, San Francisco and followed the tenets of the Declaration of Helsinki.
Adult (18 years or older) patients with type 1 or type 2 DM presenting for DR screening underwent ultra-widefield fundus photography (Optos Daytona, Optos PLC, Dunfermline, UK) and OCTA imaging during the visit. All eyes without detectable DR on the ultra-widefield fundus images were considered for inclusion in the study. Eyes with a history of ocular trauma or any form of ocular disease, with the exception of cataract, were excluded.

To develop the cascaded deep learning system, a total of 11,087 UWF images were retrospectively obtained from patients presenting for retinopathy examinations or undergoing a routine ophthalmic health evaluation between November 2016 and January 2019 at Shenzhen Eye Hospital using an OPTOS nonmydriatic camera (OPTOS Daytona, Dunefermline, UK) and 200° fields of view. Patients underwent this examination without mydriasis. All images were deidentified prior to transfer to research investigators. This study was approved by the Institutional Review Board of Zhongshan Ophthalmic Centre (Guangzhou, Guangdong, China) and adhered to the tenets of the Declaration of Helsinki.

Nonmydriatic stereoscopic 200° UWF pseudocolor images were acquired using the Optos Daytona (Optos Plc, Dunfermline, Scotoland, United Kingdom) in India and China and the Optos P200Tx (Optos Plc) in Los Angeles and Canada. UWF images were deidentified, exported, and sent to the Doheny Image Reading and Research Laboratory for analysis.