Now, about 216 rare diseases that are inherited can be determined with an accuracy of up to 90% from a photograph of a person using artificial intelligence .
In addition, the developers have created a mobile application that will help make the work of the doctor more simple. The face recognition system is called “DeepGestalt”.
The essence of its system is that it divides a person’s face into small fragments with dimensions of 100×100 pixels and determines the probability of occurrence of any disease for each fragment of the face. After that, based on all the data obtained, the system diagnoses the risk of developing the disease for the person as a whole.
Scientists have taught the system to distinguish hereditary diseases from others by downloading 614 images of people suffering from mental retardation and developmental disabilities (Cornelia de Lange syndrome) and thousands of photos of people who are healthy (for control). As a result, artificial intelligence correctly identified patients with the syndrome. The system correctly identified a person with this syndrome of Cornelia de Lange in 97% of cases.
Doctors did this only in 75% of cases.
Scientists have taught the system to identify patients with Angelman syndrome (this disease is characterized by such manifestations as mental retardation, sleep disturbances, seizures, chaotic movements, frequent laughter or smiles).
As a result, the system in 92% of cases correctly identified people with this diagnosis , and the doctors managed to do it in 71% of cases.
In addition, the system is able to separate different types of the same hereditary disease on the example of Noonan syndrome . In this case, the patients had minor differences, such as rare eyebrows.