According to the authors of a new study published today in the European Heart Journal, sending a “selfie” to a doctor can be an inexpensive and simple way to detect heart disease.
The study is the first to show that it is possible to use a deep learning computer algorithm to detect coronary artery disease (CAD) by analyzing four photographs of a person’s face.
Although the algorithm needs to be developed and tested in large groups of people from different ethnic groups, the researchers say it has the potential to be used as a screening tool to detect possible cardiovascular disease in people in the general population or higher May identify risk groups that may be referred for further clinical investigation.
“To the best of our knowledge, this is the first work to show that artificial intelligence can be used to analyze faces to detect heart disease. This is a step towards the development of an intensive learning-based tool that can be used to assess the risk of heart disease, either in outpatient clinics or through patients to perform their screenings ies selfie ‘ for taking.
It can guide further clinical trials or clinical journeys, ”said Professor Zheng Zheng, who led the research and vice president of the National Center for Cardiovascular Diseases and Fuwai Hospital, Chinese Academy of Medical Sciences and Peking Union Is the vice president of the medical college. ” Beijing, People’s Republic of China.
He said: “Our ultimate goal is to develop a self-reported application for high-risk communities to assess the risk of heart disease before coming to a clinic. This can be an inexpensive, simple and effective way to identify patients who need further investigation. However, the algorithm requires further refinement and external validation across other populations and ethnicities. ”
It is already known that some facial features are associated with an increased risk of heart disease. These include thin or brown hair, wrinkles, ear lobe creases, Xanthelsmetta (small, yellow deposition of cholesterol under the skin, usually around the eyelids) and arcus cornea (fat and cholesterol deposits that form a bluish white, gray or blue opaque.
Appears as a ring) in the outer edges of the cornea). However, they are difficult for humans to predict and quantify the risk of heart disease successfully.
Prof. Zheng, Professor Jiang-Yang Jie, director of the Institute of Brain and Cognition at the Department of Automation at Tsinghua University, Beijing, and other colleagues enrolled 5,796 patients from eight hospitals in China for the study between July 2017 and March 2019 is.
The patients underwent imaging procedures to examine their blood vessels, such as coronary angiography or coronary computed tomography angiography (CCTA). They were randomly divided into training (5,216 patients, 90%) or validation (580, 10%) groups.
Trained research nurses took four facial photographs with digital cameras: a frontal, two profiles and a view of the top of the head. They also interviewed patients to collect data on socioeconomic status, lifestyle and medical history.
Radiologists reviewed patients’ angiograms and assessed the degree of heart disease, depending on how many blood vessels were narrowed by 50% or more (% 50% stenosis), and their location. This information was used to create, train, and validate intensive learning algorithms.
The researchers then tested the algorithm on 1,013 patients from nine hospitals in China, who were enrolled between April 2019 and July 2019. Most of the patients in all groups were of Han Chinese ethnicity.
They found that the algorithm demonstrated existing methods of predicting heart disease risk (Diamond-Forester model and CAD consortium clinical score). In the validation group of patients, the algorithm detected heart disease in 80% of cases (true positive rate or ‘sensitivity’) and correctly detected heart disease was not present in 61% of cases (true negative rate or ‘specificity’ ) Belongs to. The sensitivity was 80% and specificity 54% in the test group.
Pro. Gee said: “The algorithm had moderate performance, and additional diagnostic information did not improve its performance, meaning it could easily be used to predict potential heart disease based on facial photographs. The cheeks, forehead, and nose contributed more information to the algorithm than other facial areas.
However, we need to improve specificity to a false positive rate because 46% of patients may have anxiety and discomfort. , As well as potentially overloaded clinics with patients requiring unnecessary tests. ”
The limitations of the study, along with the need for testing in other ethnic groups, include the fact that only one center in the test group was different to the centers that provided patients with the algorithm to develop, given its generality May limit to other populations.