Screens newborns for ROP using AI

Orbis International and Siloam Vision said on May 14 that clinicians in Ulaanbaatar, Mongolia, had begun using assistive artificial intelligence to screen premature infants for retinopathy of prematurity, or ROP, a disease that can cause blindness if it is not detected and treated in time. (orbis.org) The first screenings took place at the National Center for Maternal and Child Health neonatal intensive care unit, according to statements from Orbis and PR Newswire. The companies described the deployment as the first use of assistive AI for this purpose in a low- and middle-income country, though that claim comes from the organizations involved and was not independently verified in peer-reviewed literature reviewed for this article. The launch adds a real-world deployment to a field that already has published evidence that AI can help triage ROP cases from retinal images and videos. ### Why are doctors screening these babies in the first place? Retinopathy of prematurity affects babies born too early, when retinal blood vessels have not fully developed. The American Academy of Pediatrics said in its screening policy that severe ROP must be diagnosed in a timely fashion because treatment is effective only if the disease is found before it becomes destructive. The disease is also well suited to image-based review. A 2024 review in *International Ophthalmology Clinics* said ROP diagnosis is primarily image-based and therefore suitable for AI systems, while also noting that rising case loads in low- and middle-income countries have put pressure on health systems. (orbis.org) ### What exactly was launched in Mongolia? Ulaanbaatar’s National Center for Maternal and Child Health is the first site named by Orbis for the new program. Orbis said the screening session marked the launch of a program delivered through Siloam Vision’s proprietary iTelegen telemedicine platform and included hands-on training with local ophthalmologists as well as ongoing research at the hospital. (publications.aap.org) Among the first infants screened were premature twins, Ariunbaatar and Ariunnandin, Orbis said. The nonprofit said its partnership with Siloam Vision began in 2023 and combines Siloam’s software with Orbis’s local training and health-system support model. (pubmed.ncbi.nlm.nih.gov) ### Is the AI making the diagnosis on its own? Siloam Vision and Orbis described the system as “assistive AI” used to support clinicians, not replace them. The stated role is triage: helping clinicians identify babies who may need referral or closer review in places where pediatric ophthalmologists are scarce. (orbis.org) Published research points in the same direction. A diagnostic study published in *JAMA Network Open* in April 2025 reported that a machine-learning system using smartphone-collected retinal videos reached 93.3% sensitivity at the patient level for classifying whether ROP was likely, while specificity and accuracy were higher for ophthalmologist assessment than for the model. (orbis.org) ### How much should readers trust the “world first” claim? The May 14 announcement was a press release issued by Orbis and distributed by PR Newswire, and it used promotional language. (orbis.org) The claim that this was the first such deployment in low- and middle-income countries was repeated on Orbis’s own site, but the sources reviewed for this article did not include an independent registry or journal article establishing that no earlier comparable deployment existed. What can be verified is narrower. Orbis named the hospital, the city, the partner company and the date of the launch, and Siloam Vision has previously said its ROP system received U.S. (microsoft.com) Food and Drug Administration breakthrough device designation. Breakthrough designation signals that the FDA considers a device potentially important enough for expedited development support; it is not the same as full clearance or approval. ### Why is Mongolia the test case? Dr. Chimgee Chuluunkhuu, Orbis Mongolia’s country director, said in the organization’s statement that Mongolia’s vast geography and low population density make telemedicine and AI useful for supporting doctors outside the capital. (prnewswire.com) That is the rationale offered by the program’s organizers for starting there. The broader access problem is well documented. The 2024 review on AI in ROP care said barriers to implementation include infrastructure, regulation, cost, sustainability and scalability, even when the technology itself is promising. (orbis.org) ### What happens next? April 14, 2026, is the most recent dated milestone beyond the Mongolia launch: Siloam Vision said then that its iROP-Assist platform had produced positive results in a pivotal clinical study. Orbis said the Mongolia program includes ongoing research at the National Center for Maternal and Child Health, suggesting that additional data from the site will be part of the next phase of the rollout. (orbis.org) (siloamvision.com) (pubmed.ncbi.nlm.nih.gov)