Science & Enterprise subscription

Follow us on Twitter

  • A solid majority of Americans express at least some doubts about biometrics for accessing computers, this weekend's… https://t.co/oUU1W1bX6E
    about 2 days ago
  • New post on Science and Enterprise: Infographic – Americans Leery of Biometrics https://t.co/5xslb7xRX1 #Science #Business
    about 2 days ago
  • Thanks @innerarchitect ... A local Arlington pub, but still beer. Enjoy the weekend. https://t.co/CpxP0ewUVC
    about 2 days ago
  • An electronic wearable device worn as a patch is designed to act like a personal heating or cooling system in uncom… https://t.co/wTYoZ3MZD8
    about 3 days ago
  • New post on Science and Enterprise: Wearable Patch Designed for Personal Warmth, Cooling https://t.co/dvEVKGfKat #Science #Business
    about 3 days ago

Please share Science & Enterprise

RSS
Follow by Email
Facebook
Facebook
Google+
Twitter
Visit Us
LinkedIn
INSTAGRAM

Algorithms Identify More Genetic Syndromes

Data and person graphic

(Gerd Altmann, Pixabay)

8 Jan. 2019. An analysis of facial images with computer vision and deep learning returned more accurate identifications of genetic syndromes than 3 panels of trained clinicians. Results of the analysis, conducted by the company FDNA Inc., developer of the technology being tested, appear in yesterday’s issue of the journal Nature Medicine (paid subscription required).

FDNA, in Boston and Herzliya, Israel, offers systems using artificial intelligence for genetic analysis. In this case, the company’s Face2Gene service analyzes photos of faces to identify physical traits related to genetic syndromes. These syndromes, are typically rare inherited disorders characterized by abnormal growth patterns in the face or skull, and diagnosed in children. While the syndromes are individually rare, in the aggregate they affect an estimated 8 percent of the population. Identifying these conditions requires a trained clinician, often with the help of computerized tools, but the authors say most of these analyses point to just a few of the syndromes, rather than considering the hundreds of possibilities.

The Face2Gene service uses a combination of computer vision and deep learning to identify genetic syndromes. Deep learning is a form of machine learning and artificial intelligence that makes it possible for systems to discern underlying patterns in relationships, and build those relationships into knowledge bases applied to a number of disciplines. This technique uses machine learning to form layers of neural networks, with each layer adding to the knowledge derived from previous layers.

The Face2Gene base technology, known as DeepGestalt, looks for characteristics of genetic syndromes in photographs of faces. The subsequent analysis is based on training with crowd-sourced data from 150,000 Face2Gene patients. These data include some 17,000 patient images representing 200 different genetic syndromes. The company recommends using the Face2Gene service with high-throughput genomic sequencing.

The FDNA team, led by the company’s chief technologist Yaron Gurovich, evaluated the ability of DeepGestalt to identify individual genetic syndromes. The DeepGestalt system analyzed images with facial characteristics to identify specific syndromes from those images, and asked 2 panels of trained clinicians to review the same images and make their assessments. The results show the DeepGestalt analysis correctly identified more syndromes from the images than the clinicians.

In a third evaluation, DeepGestalt analyzed 502 images to identify genetic sub-types of Noonan syndrome, a genetic disorder with characteristic facial indicators, but can also affect a person’s height and heart functions. DeepGestalt succeeded in identifying 91 percent of the top 10 sub-types, again more accurately identifying the sub-types than a panel of clinicians.

The authors conclude that DeepGestalt adds value for genetic testing, clinical genetics, and precision medicine. Gurovich notes in a company statement that the study, “demonstrates how one can successfully apply state of the art algorithms, such as deep learning, to a challenging field where the available data is small, unbalanced in terms of available patients per condition, and where the need to support a large amount of conditions is great.”

More from Science & Enterprise:

*     *     *

Please share Science & Enterprise ...
error

Comments are closed.