Oxford University researchers have developed a computer program that can diagnose rare genetic disorders in children simply by analysing regular photographs.
The program works by recognising certain characteristic facial structures that can be present with certain conditions, including Down’s syndrome, Teacher Collins, Progeria, Fragile X and Angelman syndrome. It combines computer vision and machine learning to scan pictures for similarities to a database of pictures of people with known conditions, and then returns matches ranked by likelihood.
One person in 17 has a genetic disorder, which can be difficult to diagnose. Between 30 and 40 percent of the 7,000 or so rare genetic disorders involve some change in the face and skull. Having a diagnosis can help parents understand the risks for other children and how likely a condition is to be passed on. It can also improve estimates of how the disease will progress or help identify which symptoms are caused by the genetic disorder and which relate to other conditions.
The system was developed by a team from the Department of Engineering Science including DPhil research student Quentin Ferry and Professor Andrew Zisserman. The team aimed to teach a computer to carry out the assessments of facial features that are traditionally carried out by clinical geneticists.
Computational analysis of the shape of faces using 3D imaging has already been carried out to analyse conditions such as foetal alcohol syndrome, schizophrenia and autism, but these have relied on specialised and expensive imaging equipment and patient cooperation. Previous work with 2D images has relied on controlled lighting, pose and expression to allow for consistency.
The program they developed can recognise faces in regular snaps, taking into account variable lighting conditions, image quality, facial expression and poses. It identifies the corners of the eyes, nose, mouth and other features and then compares these details against what it has learnt from other photographs fed into it. As it learns, it builds up a “Clinical Face Phenotype” for different conditions. The algorithm then clusters patients with similar conditions together. In some cases it clusters patients together with no known diagnosis, which could indicate ultra-rare genetic disorders.
The team found that the approach was able to describe and discriminate between syndromes with an accuracy comparable to the more expensive 3D studies.
The authors of the study say that they envisage their system “becoming a standard tool to support clinical genetic counselling”.
“Since any normal 2D image can be analysed, this approach is available to any clinician worldwide with access to a camera and a computer. This can also reduce the need for patient inconvenience in a clinical setting because a family photo album could provide the required image(s).”
In the future, it may be possible to use the system to identify sub-phenotypes or even work out when someone has more than one genetic disorder.
“A doctor should in future, anywhere in the world, be able to take a smartphone picture of a patient and run the computer analysis to quickly find out which genetic disorder the person might have,’ says Dr Christoffer Nellåker from the MRC Functional Genomics Unit at the University of Oxford.
The image analysis echoes another piece of work being carried out at the University of Washington, where a piece of software can automatically generate images of a young child’s face as it would age through a lifetime using a single photo.
It does this by taking the average of thousands of faces of the same age and gender to calculate the changes between groups as they age, and then applying those changes to a new image. This can be done all the way up to the age of 80.
Although the area of genomics has not been developing at an exponential rate that experts expected when the Human Genome Project was announced to be completed, more and more ways of potential use of genomic data in medicine have showed how it might transform our lives. A few months ago, it was published that so-called “genetic mugshots” can be recreated from DNA. By only using a person’s DNA, a face can be generated which sounds like pure science fiction.
Now researchers at Oxford University have developed a computer program that can diagnose rare genetic disorders in children simply by analyzing family photos.
One day we might be able to sequence the genomes of newborns immediately after birth (or even before) to tell parents what major conditions the child might have to deal with in the future. As an additional feature, children without genomic sequences made available could get an instant diagnosis…
View original post 74 more words