Digital screening offers the additional benefit of lending itself to informatics analyses, a field led by Michael Chang, MD; Peter Campbell, MD; and colleagues at the University of Oregon.
“The greatest potential of digital imaging is not how we use it today, but in what the future holds in bringing the power of machine learning to bear on the problem of ROP diagnostics,” Dr. Capone said, adding that deep learning is not a panacea.
The topic is increasingly being mentioned in retina publications, he said, primarily in imaging and diabetic retinopathy articles.
“In ROP, changes to just a few pixels in a retinal image may cause the algorithm to miss disease readily apparent to human graders, or clinician-mimicking algorithms,” he said.
Physicians will not be replaced, Dr. Capone pointed out, but they should embrace digital imaging to enable better patient care.
A 2014 report from the American Academy of Pediatrics (doi: 10.1542/peds.2014-0978) embraced telemedicine for ROP. At Dr. Capone’s facility in Michigan, digital fundus imaging options are limited to devices that require multiple images “to maximally capture fundus details” (see Figure 2).
Pan-fundus, digital-imaging technologies are lacking, Dr. Capone said. Ironically, while the first-generation devices made telemedicine for ROP a reality, the cost was more than one could imagine. That may have resulted in limiting the widespread adoption of telemedicine technology for ROP surveillance.