Algorithm Developed to Detect Possible Heart Disease by Selfies
The ethical concerns over utilizing easily obtainable data has become controversial.
A research conducted in China saw the development of an algorithm that could detect how likely someone is to experience heart disease just by looking at their face through their selfies. It is quite scary to just let artificial intelligence undertake these kinds of tasks for humanity. The diagnosis of coronary artery disease is about to be one of them.
The study was published in the European Heart Journal.
%80 correct detection
As the study suggests, there are several possible symptoms which are linked to heart disease and visible to the naked eye. These symptoms include alopecia, the absence of hair on certain parts of the body, such as baldness; xanthelasmata, sort of yellow moles around eyelids; or arcus corneae, which mostly manifests white opaque rings around the cornea. These are the facial features the developed algorithm looks for to calculate the risk of heart disease through photos.
The possibility of heart disease was detected correctly in 80% of the cases. And 61% of those who did not have a high likelihood of developing a heart condition were also diagnosed correctly as well.
Let’s not forget the ethical concerns the study brought about. Taking pictures of someone or getting them on social platforms is easy. The data obtained by the algorithm will be distinctively individual, which makes the researchers consider the possible collection by unrelated institutions. However, they agree that privacy is the key.
Apparently, further trials and developments are on the way, as the researchers do not want the users to panic and create long lines before the clinics for further testing.
“Our ultimate goal is to develop a self-reported application for high-risk communities to assess heart disease risk in advance of visiting a clinic. This could be a cheap, simple, and effective of identifying patients who need further investigation. However, the algorithm requires further refinement and external validation in other populations and ethnicities.” explained Professor Zhe Zheng, lead
For the algorithm to analyze heart disease patterns, 5,796 patients from eight hospitals in China were enrolled in the study between July 2017 and March 2019. Each patient’s photos were shot at four different angles- one frontal, two side profiles, and one looking down at the top of the head. By taking them under different imaging procedures such as coronary angiography or coronary computed tomography angiography, conditions of the patients’ blood vessels were examined as well.
Radiologists reviewed the participants’ images to estimate the likelihood of heart disease based on how many of their blood vessels were narrowed by 50% and the location of the vessels in the body. That’s how they created the base of the algorithm.
Reportedly, the algorithm had “a moderate performance.” Additional clinical data had no impact on the development of the algorithm’s performance. Hence the algorithm appeared successful in detecting potential heart disease only by the photos.