All you need to know-uneed2know.eu
Go Up

Medicine

Back

Facial Recognition Software – A New Diagnosis Tool
Add to favourite

Sponsored links

Sponsored links


         One can guess quite a lot of things about a person by looking at their face, but diagnosing rare genetic disease? That seems a bit far-fetched.Not at all, according to a team from the University of Oxford.

 

Facial Recognition Software,Diagnosis Tool,face-recognition algorithms,detect genetic disease,Three-dimensional imagery


        A revolutionary computer program developed by the team lead by Christoffer Nellåker and Andrew Zisserman is about to make the difficult, expensive and often unreliable genetic disease diagnosis a much easier task with the help of face-recognition algorithms similar to the ones used in modern handheld cameras.
It is considered that roughly six per cent of the population is affected by rare genetic disorders. Genetic tests have been developed for the more common ones, such as Down’s syndrome. However, a large number of sufferers never get the chance to benefit from a proper clinical diagnosis.The reason is simple – for many rare conditions, the gene variants that cause them have not been identified yet, so doctors often find themselves relying on nothing but the facial clues. 30 to 40 per cent of rare genetic disease involve a certain amount of change to the face or skull, and sufferers show pronounced facial features, but few professionals are properly trained to identify them.

                Alastair Kent, director of the charity Genetic Alliance UK, states that this is the reason why families typically experience long delays, sometimes years, before they can receive a clear diagnosis for their child.Guesswork based on facial features is not new to medical diagnosis. For example, Abraham Lincoln's large, pronounced features determined historians to conclude, from photographs alone, that he may have suffered from a rare genetic disease called Marfan syndrome.

Similar tell-tale signs on patients’ faces can also indicate other genetically inherited diseases, such as Down's syndrome or Apart syndrome, a congenital condition characterized by malformations of the skull, or even the face, hands and feet, due to the premature fusion of certain bones.Nellåker and Zisserman’s innovation however takes it all to a whole new digital-age level, and parents of a suffering child may soon be asked to simply email the child’s photograph to the doctor. The new software has the capacity to detect up to 90 disorders, and correct diagnosis is 30 times more likely than in the case of other methods.

                Despite being still far from perfect, tests with real patients have shown a great success, indicating the correct diagnosis 93 per cent of the time. And with so many regions where genetic testing is out of reach, the team hopes to provide a potentially life-saving tool in these areas once the software is fully developed.Nellåker admits the program is not "sufficiently accurate to provide a rock-solid diagnosis”, but he is confident that ”it helps narrow down the possibilities" as he declared for New Scientist, but there is a high chance it could at least help medical staff make a preliminary diagnosis. "The idea is to offer it to health systems right across the world because all you need is a computer and a digital photo," says Nellåker.

To ”train” the software for the task, the Oxford team used 1363 public pictures of known sufferers of eight genetic disorders, such as Down's syndrome, fragile X syndrome and progeria , each of them affecting the human face in its different manner.Using the latest technology in computer vision and machine learning, the computer was then “taught” how to identify each of the conditions using a pattern of 36 facial features like the shapes of eyes, brows, lips and noses, in each photograph, and graduallyy learned what to pay attention to and what to ignore.

                As Nellåker explains, the software analyses a picture automatically, marks key feature anchor points and then compares them to those of patients with confirmed genetic diseases, thus suggesting a prediction or possible diagnosis for each face it ”reads”.Much like Google, Picasa or many other photo software can identify human faces by recognising lighting variations, identity, expression or pose, the software developed by the Oxford team builds a description of the face structure by identifying key points like the corners of eyes, nose, mouth and other features, and then compares it against its previously fed into the system database.Basically, the more number of photos of faces affected by a specific disorder it sees, the better the accuracy of the system. For example, in the preliminary tests for the eight training diseases, each of them was represented by 100 to 283 images, and the average result was 93 per cent correct.During the test, an image of Abraham Lincoln was also used as an example. The American President was long thought to have suffered from a rare genetic disease called Marfan syndrome, typically characterised by long limbs and fingers. Out of the list of 90 possible disorders, the software ranked Marfan syndrome among the 10 most likely diagnoses based on Lincoln’s photograph.

               After the success of the tests, the team expanded their software to a range of 90 genetic diseases. Exact diagnosis is still a distant goal, but the 2754 faces currently in the database seem to make correct diagnosis almost 30 times more likely.

               A crucial feature the team aims to develop is the software’s ability to group together patients based on similar clusters, and therefore to “learn from its own experience”. Moreover, it will be able to group patients with unknown disorders if they have similar facial features and skull structures, hoping it could one day aid doctors in identifying new disorders and the DNA variations behind them.Researchers also hope to train the system to analyse faces in profile just as well as frontal images. Another goal is coupling the algorithm with DNA-analysis programs that can detect  distinctive mutations, enabling the facial and genetic features of any potential newly discovered disorder to be identified and explored simultaneously.

Next step? Three-dimensional imagery. Even though the research team has already built a similar database in 3D, their aim is to continue using the regular 2D photographs to make the technique more accessible. The 3D process is much more expensive and complicated, and more importantly, it actually requires the patient to visit the hospital.As Nellåker states, this system could “in theory” be used for rare genetic disease screening in newborns, but predicts it will mainly be used by genetic counsellors working with parents who are worried about abnormalities in their children. However, the biggest benefit of the software will be its accessibility, offering a proper, reliable diagnosis tool worldwide, especially in countries where access to clinical genetics barely exists.

                The new system has been extremely well received by organisations and charities representing rare genetic disease patients, which have welcomed it as "potentially a tremendous step forward in shortening the diagnostic journey that families embark on following birth of a child with dysmorphic features", providing key access to professional expertise, advice and guidance for families more efficiently than ever.Diagnosis is a crucial step in any kind of disease. However, with all the fear, uncertainty and limited knowledge of the general public surrounding rare genetic disorders, having a diagnosis as early as possible can greatly improve the lives of sufferers and their families.

               A correct diagnosis made at the right time can not only bring some certainty, hope, and proper counselling, but it also offers reliable estimates of how the disorder might evolve, or how likely it is to be passed on to the patient’s future children. More importantly, a correct diagnosis will clearly differentiate between the symptoms caused by the genetic disorder and the ones caused by various other clinical issues that, unlike the genetic disorder, can be treated.Face recognition, either in still or moving images, is anything but new in our world of ever smarter computers leaving no more room for anonymity, but this type of technology has definitely found a perfect match in this field, and has a high chance of becoming a breakthrough in the medical world.

               However, Nellåker warns that his team’s new technology will never aim to replace traditional diagnosis, but simply to assist it, or sometimes improve, for example it in parts of the world where local professionals might lack the required expertise: “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”.

 

   

Sponsored links

       

 

Sources:
http://www.sciencealert.com.au/news/20142706-25763.html
http://www.newscientist.com/article/dn25776-computer-spots-rare-diseases-in-family-photos.html#.U7biMdJdVIG
http://www.independent.co.uk/news/science/facial-recognitiontechnology-used-tospot-genetic-disorders-9558032.html
http://www.smithsonianmag.com/smartnews-science/software-could-spot-rare-genetic-disorders-family-photos-selfies-180951840/
http://www.sciencedaily.com/releases/2014/06/140624093322.htm
http://au.ibtimes.com/articles/557498/20140630/rare-diseases-family-photos-identified-computer-program.htm#.U7bi_9JdVIE
http://www.dpag.ox.ac.uk/news/computer-aided-diagnosis-of-rare-genetic-disorders-from-family-snaps
http://ghr.nlm.nih.gov/condition/apert-syndrome
http://www.independent.co.uk/news/science/facial-recognitiontechnology-used-tospot-genetic-disorders-9558032.html
http://www.institut-biotherapies.fr/en/biotherapies/genetic-diseases-and-rare-diseases/genetic-diseases/

Published by Andreea Dobre

Share

Facebook comments (0)

u2know comments (0)

Add a comment

I have account on u2know
E-mail
Password
Remember me Forgot password?
Comment
Notify me of follow up comments via e-mail