Since rare genetic disorders are often hard to be diagnosed, scientists have found a solution by creating an AI that can detect genetic disorders by looking at a patient’s face shape.
A Boston-based digital-health firm known as FDNA came up with a new AI-powered smartphone app named ‘Face2Gene’ that will make proper diagnosis of rare genetic disorders by looking at the shape of the face.
The team trained the AI algorithm, called ‘DeepGestalt’, for identifying rare genetic disorders by first feeding it over 17,000 images of people diagnosed with one of the 216 genetic syndromes. From that data, the AI learned to look for distinctive facial features linked with specific disorder.
The technology produces a heat map visualization looking at what regions of the face contributed to the classification of diseases. The researchers then tested the AI on 502 images that it hadn’t seen before, and the app provided correct diagnosis among its list of 10 responses 91% of the time, as per the study.
The team also tested the AI’s ability to differentiate between various genetic mutations that can lead of the same syndrome. For this, they used images of people with Noonan syndrome that can result from mutations in one of the five genes. The AI perfectly identified the genetic source of the physical appearance 64% of the time.
Lead researcher Yaron Gurovich said, “It’s clearly not perfect, but it’s still much better than humans are at trying to do this.” Talking about the results, Gurovich added, “We showed that this system can be used in clinical settings.”
During clinical trials, DeepGestalt outperformed clinicians in identifying a range of syndromes in three trials. This also opens door for future research and applications, and the identification of new genetic syndromes.