Study Finds FDNA AI Technology Could Accelerate Diagnosis of Rare Conditions
February 11, 2022
FDNA, which has developed an AI platform using facial recognition for early detection of rare genetic diseases, said a paper published in the journal Nature Genetics, found the use of the company’s facial analysis tool to help detect rare genetic disorders could accelerate the clinical diagnosis by medical professionals of patients with ultra-rare disorders and facial dysmorphism, as well as enable the definition of new syndromes.
Nearly 30 to 40 percent of children with disabilities in the United States have an underlying undiagnosed condition. With 95 percent of rare diseases lacking an FDA-approved treatment, there is an urgent medical need to achieve early diagnosis in children to help promote a better quality of life.
FDNA’s deep learning technology matches rare disease patients’ photos with other patients’ photos around the world instantaneously, thus, helping medical professionals diagnose children at an earlier stage. The proprietary technology strengthens next-generation phenotyping (NGP)—the capture, structuring, and analysis of complex human physiological data—by allowing medical professionals to identify hundreds of additional disorders just with facial analysis.
“This is a long-awaited innovation in medical genetics that has finally come to fruition,” said Aviram Bar-Haim, chief technology officer at FDNA and first co-author of the paper. “Overcoming the limitations of needing a minimal number of photos per disorder is a breakthrough allowing us to now identify ultra-rare diseases. Moreover, by analyzing similarities among patients with previously unknown diseases, new genotype/phenotype correlations can be detected.”
The company’s GestaltMatcher is an AI technology that is used to identify the facial representations of more than 1,000 rare genetic diseases and distills facial features into a multi -dimensional space, which helps medical professionals to accelerate the matching and diagnosing of ultra-rare disorders.
GestaltMatcher achieves a comparable top-10-accuracy on all previously supported disorders and matches one third of cases with an ultra-rare or novel disorder. The study was conducted on 17,560 portrait photos from patients with 1,115 rare disorders.
“GestaltMatcher goes where previous technology has never gone before,” said Peter Krawitz, CSO at FDNA. “With this study, we transitioned from classification to clustering. By that means we can now compute the syndromic similarity between any two individuals in our database.”
Author: Rare Daily Staff
Sign up for updates straight to your inbox.