Artificial intelligence-assisted telemedicine platform for cataract screening and management: a potential model of care for global eye health

Correspondence to Dr Daniel Shu Wei Ting, Vitreo-retinal Department, Singapore National Eye Center, Singapore 168751, Singapore; daniel.ting.s.w@singhealth.com.sg Artificial intelligence (AI) is the fourth industrial revolution.1 Deep learning is a robust machine learning technique that uses convolutional neural network to perform multilevel data abstraction without the need for manual feature engineering.2 In ophthalmology, many studies showed comparable, if not better, diagnostic performance in using AI to screen, diagnose, predict and monitor various eye conditions on fundus photographs and optical coherence tomography,3 4 including diabetic retinopathy (DR),5 age-related macular degeneration,6 glaucoma,7 retinopathy of prematurity (ROP).8 To date, many countries have reported well-established telemedicine programme to screen for DR and ROP,9–12 but limited for cataracts. The study showed that the AI algorithm had excellent diagnostic performance (area under the curve >90%) in identifying the correct capture modes (mydriatic vs non-mydriatic and diffuse vs lateral illumination), lens status (normal, cataractous or artificial lens) and the ‘referable’ cases (as defined above). [...]the authors also described a telemedicine platform to enable home monitoring (using the ocular surface images taken by family members using cell phones, visual acuity and brief clinical history), followed by referral to the community-based healthcare facilities (where the anterior segment images were captured using a slit lamp microscope, in the telemedicine platform with AI analysis), and to the tertiary settings via a fast-tract notification system for those cases that were deemed referable. In addition to its screening and diagnostic values, AI technologies have been extended to other aspects of cataract and cataract surgery, including biometry for intraocular lens (IOL) power calculation,19 corneal power evaluation following laser refractive surgery20 and identification of phases in videos of cataract surgery for training purpose.21 For instance, Sramka et al 19 reported that machine learning, using support vector machine regression model and multilayer neural network ensemble model, could achieve significantly better results in IOL calculation compared with conventional clinical method, thereby optimising the refractive outcomes following cataract surgery.19 In the future, we envisage that AI technologies may be potentially applied in other cataract-related areas such as