NIH leverages facial recognition software to diagnose rare genetic diseases
The global fight against rare genetic diseases has a new tool in its arsenal for early diagnoses of these dreaded conditions—facial recognition software.
Researchers at the National Human Genome Research Institute have successfully leveraged the software to help diagnose DiGeorge syndrome, a rare genetic disease that afflicts African, Asian and Latin American children. The technology has also been used for diagnosing Down syndrome.
According to Maximilian Muenke, MD, chief of the Medical Genetics Branch at NHGRI, facial features are crucial to the diagnosis of such diseases. However, Muenke contends that the appearance of individuals with the diseases can vary widely in diverse populations, making it difficult for clinicians to diagnose.
However, Muenke’s colleague Marius George Linguraru, an investigator at the Sheikh Zayed Institute for Pediatric Surgical Innovation at Children’s National Health System in Washington, developed facial recognition software that is very accurate in diagnosing DiGeorge syndrome and Down syndrome using photographs of patients from different ethnic groups.
“Facial landmark” detection is the key, says Muenke. The software measures facial features such as “distances between eyes, distances between the lips, and whether the eyes slant upwards or downwards,” he adds.
Results of a study, published last week in the American Journal of Medical Genetics by Muenke and his colleagues, showed that researchers using the facial recognition software made correct diagnoses more than 96 percent of the time for DiGeorge syndrome—also known as velocardiofacial syndrome, which can cause heart defects, poor immune system function, a cleft palate, and low levels of calcium in the blood.
The effort is part of NHGRI’s Atlas of Human Malformations in Diverse Populations, a website launched last September to serve as an online resource for clinicians looking to diagnose different inherited diseases through photographs of people with syndromic disorders and the molecular diagnoses gathered from around the world, including Asia, the Indian subcontinent, the Middle East, South America and sub-Saharan Africa.
Muenke, who is co-creator of the atlas, notes that the only previously available diagnostic atlas featured photos of patients of northern European descent, which was very limiting in terms of geographic value. “We were leaving out the majority of the world’s population,” he says.
“Human malformation syndromes appear different in different parts of the world," said Paul Kruszka, MD, a medical geneticist in NHGRI’s Medical Genetics Branch. “Even experienced clinicians have difficulty diagnosing genetic syndromes in non-European populations.”
DiGeorge syndrome and Down syndrome have now been added to the electronic atlas, which shows the pattern of malformations that are consistent with these syndromes, according to Muenke. Providers with limited resources or access to genetic specialists are able to use the atlas and the facial recognition software for early diagnoses of these rare diseases that often go undiagnosed, he says.
“We take photos of every syndrome that we can get our hands on,” Muenke adds. At the same time, he relates that there is no identifying personal information on the atlas website.
Going forward, Muenke says researchers plan to study Noonan syndrome and Williams syndrome, which are also rare genetic disorders. Nonetheless, he hastens to add that “the trained clinician is always better than the software,” and in those areas of the world that have shortages of trained clinicians, then “the software is so much better” for making diagnoses in those populations.