Anemia is properly diagnosed using a blood test that measures hemoglobin, but simply looking behind a patient’s eyelid can be a pretty good alternative if you know how red the tissue is supposed to be. Now, a team at Purdue University has developed a technology that lets a clinician use smartphone pictures of the inner eyelid to automatically obtain surprisingly accurate results of blood hemoglobin levels. This capability is expected to soon be fully integrated into an app that will perform the necessary image analysis.
Unlike blood tests, smartphone apps are incredibly easy to roll out and put to use in even some of the most remote and resource poor areas of the world. Even in countries where blood testing is readily available, a quick point-of-care test that provides pretty accurate estimates of hemoglobin levels can be incredibly useful in a variety of cases. “This technology won’t replace a conventional blood test, but it gives a comparable hemoglobin count right away and is noninvasive and real-time,” said Young Kim, an associate professor of biomedical engineering at Purdue and lead author of the study appearing in journal Optica. “Depending on the hospital setting, it can take a few hours to get results from a blood test. Some situations also may require multiple blood tests, which lead to more blood loss.”
The new technique relies on super-resolution spectroscopy to convert smartphone photos into high-resolution spectral maps. An additional algorithm then reads these spectral signals and turns them into hemoglobin counts. This algorithm was created by training it using eyelid photos and hemoglobin counts of dozens of individual patients presenting at a hospital in Kenya.
“The idea is to get a spectrum of colors using a simple photo. Even if we had several photos with very similar redness, we can’t clearly see the difference. A spectrum gives us multiple data points, which increases chances of finding meaningful information highly correlated to blood hemoglobin level,” said Sang Mok Park, a Purdue Ph.D. candidate in biomedical engineering, and another researcher involved in the study.