Evaluation of a deep learning system for the joint automated detection of diabetic retinopathy and age‐related macular degeneration

Screening for eye diseases has become a high‐priority healthcare service to prevent vision loss (Cunha‐Vaz 1998; Rowe et al. 2004). Due to its proven efficiency, screening programmes based on periodical examinations of the retina have been increasingly implemented worldwide (James et al. 2000; Arun et al. 2003; Jones & Edwards 2010). Established protocols rely on manual readings by highly specialized workforce (Piñero 2013), failing to meet the requirements of large‐scale screening in high‐ and low‐resource countries (Harmon & Merritt 2009; Shaw et al. 2010; Guariguata et al. 2014; Wong et al. 2014; United Nations Department of Economic and Social Affairs 2017). Furthermore, cost‐effectiveness remains to be the main burden for establishing screening programmes (Wormald 1999; Hernández et al. 2008; Karnon et al. 2008), and different protocols are followed for different diseases (AAO 2015, 2017), which translates to a larger burden to health systems and to the patient, that needs to undergo several of them. Nevertheless, exploiting the fact that examination protocols of retinal diseases rely mostly on the same principles and actions, it becomes more efficient to integrate them in one workflow (Chan et al. 2015; Chew & Schachat 2015).