The Face2Gene App May Help Diagnose Rare Diseases

Selfies have never been more helpful.

Photo provided by Face2Gene

You might not recognize the name Moti Shniberg, but he could recognize you. Shniberg developed a facial recognition and tagging technology that he eventually sold to Facebook—and now, he and his team at Boston-based FDNA are taking selfies from social media to the doctor’s office.

FDNA is behind Face2Gene, an app that the company says can help doctors diagnose more than 2,000 rare diseases. The program, which is used by geneticists worldwide, uses deep learning algorithms and facial recognition technology to scan a patient’s photo for facial descriptors that can be compared with traits that signify various uncommon conditions.

About one in 10 Americans has a rare disease, and, on average, it will take a patient seven years, and seven doctors, to reach a diagnosis. FDNA is out to make life easier for these families.

“This diagnostic odyssey puts a tremendous amount of emotional and financial strain on families desperately trying to find answers,” says CEO Dekel Gelbman. “Our technology…helps doctors reduce this journey and find answers that can lead to better patient outcomes.”

The journey can be even more arduous for someone who lives near few medical resources. In an ideal scenario, a patient seeking a diagnosis would undergo whole genome sequencing, mapping his or her DNA to help doctors reach a diagnosis and treatment plan. But this process, while precise, is expensive and time consuming, and thus not available to all patients.

Programs like Face2Gene, Gelbman says, can help. It’s already made a difference for doctors in underserved parts of Mexico, he says. “The doctors are left with no real alternatives other than trying to diagnose these patients clinically, without genetic testing to support their clinical diagnosis,” he says. “They don’t really have tools to assist them, so they use our technology to capture their patient’s phenotype and match that to known phenotypes of syndromes.”

Gelbman says a doctor can list a patient’s phenotypes—behavioral and physical manifestations of a certain syndrome—in the app. The doctor then uploads a photo of the patient, and the app uses deep learning to analyze his or her facial features, eventually recommending more symptoms and phenotypes for the doctor to consider. Finally, the physician is presented with a refined list of possible diagnoses.

Face2Gene works best when used in conjunction with whole exome sequencing (WES), a process that analyzes roughly 2 percent of an individual’s DNA. WES is a cheaper alternative to whole genome sequencing, but it is only about 25 percent accurate in diagnosing rare diseases. When coupled with Face2Gene, however, the accuracy rate goes up to 85 percent, Gelbman says. This can end the diagnostic odyssey for patients, many of whom are children, and get them on the path to treatment and clinical trials.

A picture is worth a thousand words—and, sometimes, a whole lot more.