Detecting Anemia Using AI & Human Eyes at UA Little Rock

(Shutterstock)

A few years ago, after learning of his father’s anemia diagnosis, Nahiyan Noor began researching the blood disorder and how it is diagnosed. That led him to a study about how individuals’ hemoglobin levels can be detected via the conjunctiva (membrane) of their eyes.

People with anemia do not have enough hemoglobin, a protein in red blood cells that carries oxygen. Noor, who recently graduated with a master’s in information science from the University of Arkansas at Little Rock, realized there might be a way to detect anemia through examining the eye rather than through a painful blood test.

The 26-year-old’s project is another example of the cutting-edge research students and professors are conducting at UA Little Rock. Last month, I wrote about professors and graduate students there working on creating nanomaterials that could be applied in fields ranging from renewable energy and health care to food safety and space exploration.

Noor has been working alongside professors in UA Little Rock’s School of Engineering & Engineering Technology and the Collaboratorium for Social Media & Online Behavioral Studies, or COSMOS, a think tank focused on social media and online behavior. (Next month, in the final column of a three-part series, I’ll take a look at what they’re up to at COSMOS.)

The recent graduate has published numerous papers on social media behavior related to the coronavirus pandemic as well as his research about detecting anemia by looking at human eyes. This spring, he presented his findings about online discourse, COVID-19 and toxicity on social media platforms at an academic conference in Barcelona, Spain.

For both avenues of research, Noor has applied machine learning and artificial intelligence to help with data analysis. Since beginning his research on eyes, other scholars around the world have joined the effort, creating a global collection of data to increase the accuracy of predicting the blood disorder.

So far, Noor says he’s collected more than 400 samples. From those, he uses colors from the light spectrum to measure hemoglobin levels, and thus detect anemia. The data is placed in a spreadsheet and crunched using machine learning and code that Noor has written for its analysis.

He said his research can detect anemia in a patient with about 93% accuracy. Thousands of samples are needed to move toward 100% certainty, Noor said.

With about 3 million cases, anemia is the most common blood disorder in the U.S., according to the National Institutes of Health.

Researchers at Purdue University in West Lafayette, Indiana, are also working on eye research and anemia detection. They’ve developed a software application that would allow doctors to instantaneously detect the blood disorder rather than wait hours for lab results. The real-time results of this method of detection are one of the most important benefits.

“Blood hemoglobin tests are regularly performed for a wide range of patient needs, from screenings for general health status to assessment of blood disorders and hemorrhage detection after a traumatic injury,” Purdue said in a news release about the software.

Young Kim, a Purdue biomedical engineering professor, said that while the technology may not replace a conventional blood test, “it gives a comparable hemoglobin count right away and is non-invasive and real-time.”

Noor said his interest in collecting, processing and cleaning data using machine learning is what led him to UA Little Rock. Because of his father’s health condition, he envisioned a future in which data and artificial intelligence could be used to predict and diagnose diseases.

For now, Noor is exploring jobs in medical or public health settings in which he would work with more data analysis but says he will continue to work on the anemia detection research until test results can diagnose anemia with almost total accuracy. “If it is around 97% or 98%, then it could be commercialized,” he said.