Article: AI correctly diagnoses vision-threatening condition in babies in India
Source: Oregon Health & Science University
Published: March 22, 2021
 |
| A normal retina (left) and a retina with ROP showing dilated, tortuous blood vessels (right) |
An artificial intelligence technology used in the United States to diagnose retinopathy of prematurity (ROP) was successfully applied in India, according to researchers at OHSU and Massachusetts General Hospital. Retinopathy of prematurity develops in premature infants who are given high levels of oxygen to sustain them after birth; the high amount of oxygen causes abnormal blood vessel growth in the retina, potentially leading to blindness. Retinopathy of prematurity affects 20,000 babies worldwide, especially in developing countries where there are fewer trained ophthalmologists. According to an earlier study published in JAMA Ophthalmology, the use of this technology to diagnose ROP via images showed a 91% accuracy rate compared to expertly trained ophthalmologists. The new study applied the algorithms to 1,253 eye exams from an ROP telemedicine program in India, and correctly identified 100% of severe cases requiring treatment. The researchers also report that their i-ROP DL system is less susceptible to racial bias by implementing a two-step process, first identifying blood vessel patterns in the original eye
image, then turning the image into a black-and-white blood vessel map. It is this second image in black-and-white that is analyzed for ROP, the greyscale colors reducing ethnicity-related differences in retinal image appearance. The technology has been granted breakthrough status by the FDA in 2020 and is licensed to a company called Boston AI Labs for further development. The study authors conclude, "Integration of AI into ROP screening programs may lead to improved
access to care for secondary prevention of ROP and may facilitate
assessment of disease epidemiology and NCU resources."
My rating of this study: ⭐⭐⭐
Campbell JP, Singh P, Redd TK, et al. "Applications of Artificial Intelligence for Retinopathy of Prematurity Screening." Pediatrics. 147(3):e2020016618. 1 March 2021. https://doi.org/10.1542/peds.2020-016618
No comments:
Post a Comment