Purdue technology explored as a potential tool to identify women at higher risk of preeclampsia

A researcher in Purdue University's Weldon School of Biomedical Engineering is participating in a two-year research study evaluating approaches to monitor the health of pregnant women in Africa and inform future efforts to reduce maternal mortality.

Young Kim, professor of biomedical engineering, University Faculty Scholar and Showalter Faculty Scholar, has developed an innovation being explored as a potential tool to identify women at higher risk of developing preeclampsia.

Preeclampsia is a major cause of maternal mortality, premature birth, stillbirth and neonatal death worldwide. The two-year study is funded as part of the Gates Foundation's Grand Challenge awards to reduce the burden of preeclampsia.

The solution uses a patented, noninvasive computer-vision method called mHealth conjunctiva AI imaging to analyze smartphone photographs of the eyeball to explore early prediction of preeclampsia. The method extracts microvascular patterns from photos of the conjunctiva, which is a thin and transparent membrane covering the inner eyelids and the white part of the eyeball. It will be used in collaboration with AMPATH in Kenya.

Kim disclosed the computer- and color-vision methods to the Purdue Innovates Office of Technology Commercialization, which has applied for patents to protect the intellectual property.

The work is part of Purdue's presidential One Health initiative, which involves research at the intersection of human, animal and plant health and well-being.

About preeclampsia

Preeclampsia is one of the most common pregnancy complications where persistent high blood pressure develops during pregnancy, usually after the 20th week; it also can develop during the postpartum period. Early diagnosis and treatment are crucial to prevent serious complications for both mother and baby.

The World Health Organization reports preeclampsia affects up to 8% of pregnancies around the world. There are around 46,000 maternal deaths and around half a million fetal or newborn deaths annually due to preeclampsia.

Preeclampsia is insidious in onset, and diagnosis can be either missed or made too late. If left untreated, preeclampsia could be fatal for both the mother and baby. A woman with preeclampsia may have high blood pressure, high levels of protein in her urine that signal kidney damage or other indications of organ damage.

About the photo analysis innovation

During a two-year study of ongoing clinical work, researchers will use computer- and color-vision methods developed at Purdue to analyze smartphone photos. They will recruit 1,600 pregnant women in western Kenya for the study at Moi University via a partnership called AMPATH Kenya, which is deeply engaged with the local communities.

By combining radiomics with supervised learning, we extract microvascular patterns that may be clinically relevant from unmodified photos of the conjunctiva rather than directly imaging the retina. Our team is among the first to identify the conjunctiva as a promising imaging site that offers an alternative window into health conditions and diseases, as reported in the peer-reviewed journals npj Digital Medicine, Science Advances and IEEE Transactions on Image Processing."

Young Kim, professor of biomedical engineering, Purdue University

Kim said numerous studies support the link between microvascular abnormalities like narrowing and construction with high blood pressure.

"Several previous studies found the alterations preceded the clinical onset of preeclampsia," he said. "Microvascular changes in the retina were observed during the early weeks of gestation and were linked to increased peripheral resistance before blood pressure rose."

Kim said traditional computer-vision analysis is limited because of the need for specialized devices like retinal fundus imaging systems. But the Purdue mobile health method to analyze smartphone photos removes this significant barrier.

"Our noninvasive solution eliminates the need for specialized equipment," he said. "Smartphones have recently transformed health care in resource-limited settings where community health workers are often equipped with mobile health apps to connect with health care professionals even from remote areas."

Kim said several milestones are in place to advance the work.

"Our first step will be to refine the prediction model to specifically target preeclampsia rather than general maternal hypertension," he said. "Next, we'll develop a minimally viable mobile app to support scalable validation."