A number of genetic diseases give their petients unique facial characteristics, Down’s syndrome probably being the most famous. While some physicians are able to spot specific diseases, there hasn’t been a systematic way of using facial features to diagnose disease. Some diseases, being extremely rare, often end up going unnoticed for years while being displayed openly on the person’s face. Now researchers from Oxford University have reported in journal eLife a new computer vision software that is able to provide ‘diagnostically relevant facial gestalt information from ordinary photos.”
The algorithm behind the software was designed to learn signs of disease from a large set of photos of people with identified diagnoses. The software corrects for lighting, angle of the subject, and even the identity of the person, leaving only the relevant characteristics of a person’s face. It then makes a set of measurements between different parts of the face and compares these with what it learned from the original photo set provided. Once a match is made, the software offers its guess, which needs to be confirmed using genetic sequencing or some other method. The researchers envision this type of software automatically screening patients during regular physician checkups, providing a cheap way to identify rarely seen genetic disorders using nothing more than a smartphone.
Oxford University: Computer-aided diagnosis of rare genetic disorders from family snaps…