Diagnosing genetic diseases is now made easier with a new application of artificial intelligence. A new study measured clinicians’ skills in diagnosing genetic abnormalities in newly born children. The researchers compared this to the success rate of a computer program with face recognition technology. The computer program scored better in terms of diagnosing genetic abnormalities than the clinicians did.
Here are the details. CNN gave an overview of the study.
The researchers called this technology “Deep-Gestalt”, which translates into “deep appearance”. According to the original publication in Nature three separate trials showed that the Deep-Gestalt program outperformed the recognition of clinicians.
Results of the study
The study found the following.
- About 8% of the population has genetic abnormalities. There are key genetic components that often translate into characteristic facial features, which a computer can recognize and record.
- For instance there is one genetic syndrome of the nervous system, called Angelman syndrome. These patients present with a wide mouth, teeth that have wide gaps between the teeth, a protruding tongue and strabismus. This is the condition of the eyes where each eye points in a different direction. This is one example where the Deep Gestalt program scored higher than clinicians.
- The study went through 17,000 facial images of a database with more than 200 genetic syndromes. Out of 502 images the Deep-Gestalt program identified the correct diagnosis in 92% of the top10 suggestions.
- Another genetic abnormality, the Noonan syndrome, presents with a number of clinical defects, the worst being a heart defect. Gurovich, the lead author said that the artificial intelligence program identified Noonan syndrome with a success rate of 64%. The clinicians were able to identify this genetic syndrome only in 20% of the cases.
It can be difficult with so many genetic syndromes that exist to accurately diagnose them in a particular patient. However, with an artificial intelligence technique named Deep-Gestalt it is now possible to be more accurate than clinicians would be. This type of new technology is a valuable adjunct for clinicians working in the field of genetic abnormalities. Some of the researchers suggested that improvements to this technology in the future will increase the reliability even further.