Your doctor might be able to use your selfies to spot potential problems
By Caitlin Finlay
We take selfies to document special occasions, locations, and future cherished memories, but did you know that your doctor might be able to use your selfies to detect health risks? According to a study recently published in the European Heart Journal, photos of a person’s face could be used to detect heart disease.
Researchers worked with eight hospitals in China, where data was collected from 5,796 patients undergoing an angiogram—an imaging procedure to assess blood vessels. Four photos were taken of each patient—a frontal photo, two profile photos, and a photo of the top of the head. The data from the patients were split into two groups: a training group and a validation group. The training group consisted of 5,216 patients whose data would train a computer algorithm—the sort of artificial intelligence (AI) that predicts your preferences in Google, for example—and the remaining data from the other 580 patients would be used to validate what the algorithm had learned.
The photos and information from the training group were used to teach an algorithm to look for specific facial features that have been associated with an increased risk for heart disease and assess the risk based on the patients’ health information. The associated facial features include wrinkles, an ear-lobe crease, thinning or grey hair, and two types of cholesterol deposits: zanthelasmata (found around the eyelids) and arcus corneae (which affects the outer edges of the cornea). While these facial features are challenging for humans to use as reliable predictors or to quantify someone’s risk of heart disease, AI and algorithms might give us the tools to make the job easier.
The researchers validated the algorithm by comparing its results with angiograms evaluated by a radiologist. They also used photos and health data from 1,013 additional patients from nine hospitals in China and compared the algorithm’s performance with the results of existing methods of predicting risk of heart disease—the algorithm was found to out-perform existing methods.
Notably, the limitations to the study were a high rate of false positives (up to 46%) and a limited distribution of ethnic backgrounds, which would need to be addressed in future research.
“To our knowledge, this is the first work demonstrating that artificial intelligence can be used to analyze faces to detect heart disease,” said Professor Zhe Zheng, who led the research and is the vice director of the National Centre for Cardiovascular Diseases and the vice president of Fuwai Hospital in Beijing, China. “It is a step towards the development of a deep learning-based tool that could be used to assess the risk of heart disease, either in outpatient clinics or by means of patients taking ‘selfies’ to perform their own screening.”
Another study, recently published in Nature Medicine, demonstrated that smartphone cameras could be used to detect diabetes. The smartphone camera and flash were used to measure photoplethysmography (PPG) signals, which detect vascular damage due to diabetes. The study was able to train an algorithm using PPG recordings from 53,870 patients who had been diagnosed with diabetes. The algorithm was then validated in two groups comprising 7,806 and 181 subjects. When the algorithm was tested, it was able to identify patients who did not have diabetes with an accuracy of 92% to 97%, and able to identify patients with diabetes with an accuracy of 82%.