A recent study, published in the journal Genetics in Medicine, have shown that artificial intelligence can be used to diagnose rare diseases more efficiently and reliably. Scientists show that using portrait photos in combination with genetic and patient data improves facial diagnoses.
Rare disorders often show up in somebody’s appearance. Persons with Noonan syndrome, a genetic condition which inhibits the body’s growth and development, can have wide-set eyes, for example.
Yearly, about half a million children are born with a rare hereditary disease worldwide. Obtaining a perfect diagnosis can be tough and time-consuming. Moreover, many patients with rare diseases go through protracted trials until they are properly diagnosed. This results in a loss of valuable time needed for early therapy to prevent progressive damage.
Now, scientists have trained artificial intelligence to identify these features, paving the way for early and cheap diagnoses. Thus, an international team of researchers has demonstrated how artificial intelligence can be used to make rather fast and reliable diagnoses in facial analysis.
A tool of artificial intelligence
The researchers of the study used data from 679 patients. These patients were with 105 different diseases caused by a single gene change. These consist of, for instance, mucopolysaccharidosis (MPS). This disorder leads to learning difficulties, bone deformation, and stunted growth. Mabry syndrome also results in cerebral disability.
All these illnesses have in common that the facial features of affected ones show different defects. This is chiefly characteristic of Kabuki syndrome. It is reminiscent of the make-up of the old Japanese form of theatre. The eyebrows are curved with wide eye-distance and long spaces between the eyelids.
The researchers used the software which can automatically detect these characteristic features from a picture. With the clinical signs of the patients and genetic data, it is potential to calculate which disease is most likely to be involved.
The AI and digital health company FDNA has developed the neural network DeepGestalt. It is used by the researchers as a tool of artificial intelligence for study. According to the researchers, PEDIA is a unique example of next-generation phenotyping technologies. Thus, mixing an advanced AI and facial analysis framework like DeepGestalt into the variant analysis workflow will bring about a new model for superior genetic testing.
Researchers train the neural network with 30,000 pictures
The researchers trained this computer program with round 30,000 portrait images of people affected by rare syndromal illnesses.
Thus, the researchers found that in combination with facial analysis, it is likely to filter out the critical genetic factors and arrange genes. Moreover, merging data in the neuronal network reduces analysis time and leads to a higher diagnosis rate.
At the UKB, the head of the Institute of Genomic Statistics and Bioinformatics has been working with FDNA for some time. This is of inordinate scientific interest to us and also permits us to find a cause in unexplained cases. Many patients are presently still looking for clarification for their symptoms.
The study is a team effort between medicine and computer science. Artificial intelligence supports physicians and scientists to provide a prompt and accurate diagnosis. This easy diagnosis also improves the quality of life of those who are affected to some extent.