AI and Facial Recognition to Diagnose Rare Genetic Disorders

The advent in technology has given birth to reverse image search, which has got matured over the past few years in facial recognition. Algorithms can spot facial features associated with rare conditions. A scan for facial recognition could be an essential part of the standard medical procedure in the future. Researchers prove that the algorithm can identify the characteristics of your face associated with genetic disorders. This new development can change your life and speed up medical diagnoses.

A study in Nature Medicine journal reveals that FDNA (a US company) has published a series of tests of DeepGestalt (their software). The company trained and updated its algorithm with dataset analysis. The dataset contained face to uses a combination of facial recognition and artificial intelligence. FDNA collected over 17,000 images and covered 200 syndromes with an app, Face2Gene.

Rare genetic syndromes are common and affect 8% of the world’s population. In the initial two tests, DeepGestalt looked for particular disorders: Angelman Syndrome and Cornelia de Lange Syndrome. These conditions are complicated and can affect your mobility and intellectual development. They have specific facial traits, such as arched eyebrows with a connection in the mid for Cornelia de Lange Syndrome. Angelman syndrome is identified by hair and fair skin.     

There was a task to distinguish photos of patients with a syndrome; DeepGestalt was 90% accurate, beating expert clinicians. Only 70% of clinicians were accurate on a similar test. After testing 502 photos, DeepGestalt recognized the target disorder.

In other challenging experiments, the algorithm identified Noonan syndrome through images. However, the accuracy of the software was low because the hit rate is 64% and the software performed over 20%.

Although, experts claim that these types of algorithm tests may not identify rare genetic disorders. Dr Bruce Gelb, Icahn School Profession of Medicine, prefers a genetic analysis over software. Gelb noted the performance of DeepGestalt that was testing fairly, but these tests were performed on a limited dataset. It was noted that the algorithm was effective on Caucasian instead of Africans. Gelb assumed that it might help physicians to save time in the diagnosis.

FDNA understands these shortcomings, and they refer to DeepGestalt as a reference tool or AI-powered software. They admit that this software can’t replace human diagnoses. However, it could assist doctors in their work. Christoffer Nellaker, an expert at Oxford University believed that the new technology could drastically cut down the time of diagnosis.