Doctors often take into account a a patient's face when diagnosing a genetic disorder. Despite many syndromes having recognizable facial features, sometimes relying on facial features for diagnosis can result in clinicians not considering hundreds of other possible conditions.
The app is powered by FDNA's image-analyzing algorithms, DeepGestalt, which has already outperformed clinicians in experiments, according to a study published in Nature Medicine.
Face2Gene was trained on 17,000 facial images that correspond to more than 200 genetic disorders.
With the framework, DeepGestalt attempted to identify potential disorders in more than 500 photographs of people it hadn't seen before. The app created a list of 10 diagnoses, and more than 90 percent of the time, the correct diagnosis was in the list.
In addition, FDNA wanted to see if Face2Gene could distinguish between different mutations that could cause the same syndrome. For example, researchers used images of people with Noonan syndrome. This disorder can stem from a mutation in one of five genes.
Face2Gene returned a correct identification based on the image 64 percent of the time.
In the three experiments, DeepGestalt outperformed clinicians.
"It's clearly not perfect," Yaron Gurovich, the FDNA's chief technology officer told The Scientist. "But it's still much better than humans at trying to do this."
More diverse data could enhance the app's abilities. "We know this problem needs to be addressed," Gurovich told Nature. "And as we move forward we're able to have less and less bias."
FDNA said it wants to make Face2Gene free for healthcare professionals.
FOLLOW Download.com on Twitter for all the latest app news.
- The Face2Gene app to help doctors narrow down genetic disorder diagnoses to ensure the best treatment plan for patients.
- It stands to reason that further training and data for DeepGestalt will only increase the app's accuracy. In addition, the FDNA wants to make Face2Gene free for healthcare professionals.
- Consumers still trust real people over AI for app recommendations, according to a new study
- Huawei app 'Facing Emotions' uses AI to translate emotions into sound for the blind
- New apps use AI to help make the world more accessible for blind and visually impaired people
- Self-care in 2019: 10 apps to help you do better
- CES 2019: Newton's Meter is an AI-powered safety button (CNET)
- Artificial intelligence has a probability problem (ZDNet)
- The top 5 AI jobs for 2019 (TechRepublic)