Algorithm could allow screening without need for invasive, infrastructure-requiring blood tests
A picture of a person’s inner eyelid taken with a standard smartphone camera can be used to screen for anemia, according to a new study published this week in the open-access journal PLOS ONE by authors including Selim Suner, M.D., James Rayner, M.D., M.Eng., and Gregory Jay, M.D., Ph.D., all emergency physicians at Rhode Island Hospital and The Miriam Hospital, and faculty at The Warren Alpert Medical School of Brown University; emergency medicine resident Caroline Meehan, M.D. was among the authors as well.
Anemia, a low blood hemoglobin concentration, affects an estimated 5.6% of Americans and more than 25% of the global population. Severe anemia is a significant risk factor for morbidity and mortality, especially in children, the elderly and the chronically ill. There is an unmet need for inexpensive, accessible and non-invasive point-of-care tools to screen for and diagnose anemia.
Previous studies have shown that the inside of a person’s lower eyelid—called the palpebral conjunctiva—appears paler with anemia. In the new study, researchers obtained smartphone images of the palpebral conjunctiva from 142 patients with a wide range of hemoglobin levels. They zoomed into a small region of the conjunctiva in each photo and developed a new algorithm optimizing color resolution as well as a prediction model linking conjunctiva color, compared to the surrounding skin and whites of the eyes, to hemoglobin levels. Then, the team tested the new algorithms on photos collected from 202 new patients.
When analyzing the new set of photographs, the model was 72.6% accurate (CI 71.4-73.8), 72.8% sensitive (71-74.6), and 72.5% specific (70.8-74.1) at predicting anemia. Accuracy for transfusion thresholds was higher, at 94.4% accurate (93.7-95.0) for a low transfusion threshold and 86% accurate (85.0-86.9) for a higher threshold. Skin tone did not change results, but image quality had some effect. The results suggest that a smartphone app could be used to screen for anemia in a telehealth or remote setting where the infrastructure for blood tests is not available.
“Others have used photos of the creases in the palms, fingernail beds, and other parts to devise algorithms to predict anemia,” said Suner. “Because those areas can be affected more readily by temperature changes or other conditions affecting bloodflow, the eyelid gives us a more reliable way to make this prediction.”
“Images of the lower eyelid’s vascular surface obtained by a smartphone camera can be utilized to estimate blood hemoglobin concentration and predict anemia which is a serious health condition afflicting billions of people world-wide with a disproportionate effect in developing countries,” said Jay.
“Utilization of non-invasive techniques to detect anemia opens the door to widespread screening, early diagnosis and treatment, particularly in low resource settings where access to healthcare is sparse,” said Rayner.
Suner added that the next iteration of an app designed for this purpose will allow novice users to take the photo with sufficient quality for accuracy, honing focus and lighting to minimize error. That app will be validated with a new cohort of patients.