Diagnosis by photo

Rare genetic diseases will be detected by family photos

Copper news based on New Scientist: Computer spots rare diseases in family photos

Experts from the University of Oxford have developed a computer program that allows you to diagnose rare genetic diseases from family photos, based on the analysis of facial features and comparing them with characteristics from the database. The authors' goal is to make the diagnosis of such diseases simple and accessible. The work of Ferry et al. Diagnostically relevant facial gestalt information from ordinary photos is published in the journal eLife.

Rare genetic diseases are thought to affect about six percent of the world's population. Genetic tests exist to identify the most common of them, such as Down syndrome, but clinical diagnosis of most of the more rare syndromes is difficult due to the lack of data on the genetic variations associated with them. For the most part, when diagnosing specialists rely on identifying the patient's characteristic facial features and skull deformities specific to about 30-40 percent of rare genetic diseases. However, the number of such specialists is limited, which leads to long delays in the diagnosis, which in the case of a child, can be critical.

The software developed by Christopher Nellaker, Andrew Zisserman and their colleagues is designed to help family doctors and general pediatricians make a preliminary diagnosis based on photos of the child's relatives. "Our idea is to provide this software to healthcare systems around the world, since all you need is a computer and a digital photo," explained Nellecker, quoted by NewScientist magazine.

The program automatically analyzes the image, highlighting 36 key positions, such as the shape of eyebrows, eyes, lips, nose, their relative position and others, while taking into account many different parameters – lighting, image quality, posture, facial expression and so on. As a result, a description of the structure of the face is made, which is then compared according to certain parameters with photographs of patients with genetic diseases from a database uploaded to a computer. After analyzing a group of photos of the patient's relatives, as a result, the program does not make an accurate diagnosis, but calculates the probability of having a particular disease for each case. At the same time, there is an automatic grouping of images of people by phenotypic signs with, presumably, the same disease, even if it has not yet been diagnosed, which makes it possible to identify new genetic diseases.

Here and below are the drawings from the article in eLife – VM.

To test the system, Nellecker and his colleagues uploaded 1,363 known images of people with 8 genetic diseases to the database, including Down syndrome, Angelman syndrome, fragile X chromosome syndrome and progeria. The accuracy of the results at the same time, the authors note, directly depends on the number of photos in the database. So, each of the eight diseases was represented in the database by 100 – 283 images, the accuracy of calculating the probability was on average 93 percent. Now the authors have already brought the number of photos in the database to 2754, which allows the program to recognize 90 diseases. For example, the system placed Marfan syndrome, which, according to some assumptions, was observed in former US President Abraham Lincoln, in seventh place out of 91 possible.

Currently, the Nellecker group is working on improving its development so that the system can analyze faces not only full-face, but also in profile. In addition, scientists plan to combine this software with programs that analyze DNA for the presence of specific mutations, which should increase the effectiveness of diagnostics.

Portal "Eternal youth" http://vechnayamolodost.ru 25.06.2014