The White Man's Problem
Diagnostics by photo
Dina Mingalieva, Copper News
Dr. Maximilian Muenke has a superpower: he can diagnose a disease just by looking at a person. More specifically, Munke is able to recognize certain hereditary diseases by characteristic facial features. "When you've been doing this for a number of years, you just walk into a room and say – oh, this child has Williams syndrome," says Munke, referring to a genetic disease that affects a person's heart and cognitive abilities.
This is an incredibly useful skill: firstly, thanks to such a "diagnosis" a person can be sent for a genetic test, and secondly, in many parts of the world people basically do not have access to genetic tests.
Software that analyzes the patient's face for signs of illness will help doctors better diagnose and treat people with hereditary syndromes. The article by Kate Sheridan for the publication STAT (Facial-recognition software finds a new use: diagnosing genetic disorders) is dedicated to "Facial diagnostics".
Previously, the analysis of faces was carried out on large bulky scanners, but now, in the era of smartphones, everything is different: Face2Gene, a program developed by the Boston startup FDNA, has a mobile application.
Doctors can take pictures of the patient's face and get a list of syndromes that he may have.
In general, all such algorithms operate on the basis of the same principles – the study of the size of facial features and their location in order to detect features.
Both the National Institutes of Health (NIH), with whom Muenke works, and FDNA train on photos of patients taken by doctors. NIH works with partners around the world, and FDNA uses images uploaded to Face2Gene.
However, there is a key difference: the NIH can predict if someone has a genetically determined disease, and Face2Gene does not issue diagnoses, but assumptions.
The application database has photos of people with two thousand diagnoses, and the program determines the percentage similarity of the person in the uploaded image with photos of people in their database.
And Face2Gene is a kind of "search engine" for diseases. "We are not a diagnostic tool and never will be," comments the head of FDNA, Dekel Gelbman.
The algorithm that the NIH uses is quite effective: out of 129 patients with Down syndrome, it accurately identified the syndrome in 94% of cases. For Di Georg syndrome, the number of hits on target is even higher: 98% accuracy among 156 cases.
However, both the NIH tool and the FDNA application have one big problem: a lack of data on the non-white population. "In all the textbooks I studied in Germany and the States, there were photos of people with Northern European roots," complains Munke. That is, the diagnosis of diseases by facial features is difficult in countries where most residents do not have Northern European ancestors.
For example, children with Down syndrome often have a flat bridge of the nose – the same bridge of the nose is characteristic of most African or African-American children. As for Down syndrome, there are only two reliable signs for all races and nationalities.
In fact, using a white person's face as a reference is the least representative choice. "One of the most interesting discoveries of our study is that the race that is most different from all the others in terms of facial features of Di Georg syndrome is the European race," says Muenke's colleague Marius Linguraru.
To solve this problem, more patient faces need to be uploaded to the databases of both projects. However, there are no genetic laboratories registered in the NIH Genetic Testing Registry in Africa – and there are very few such laboratories in Asia and South America. Most geneticists work in North America and Europe. In Nigeria, for example, there is not one in the whole country.
Portal "Eternal youth" http://vechnayamolodost.ru
11.04.2017