A neural network only needs one lung scan to predict heart problems
Researchers have developed a deep-learning model that uses a single chest X-ray to predict the risk of death from heart disease. The findings will be presented at the Radiological Society of North America conference.A team of researchers led by the Massachusetts Cardiovascular Imaging Research Center trained a neural network to estimate the risk of death from a heart attack or stroke over the next 10 years. A single chest X-ray is sufficient for the analysis.
Atherosclerotic cardiovascular disease risk assessment is used to prescribe preventive statins. Traditionally, this uses a statistical model that accounts for many variables, including age, sex, race, systolic blood pressure, treatment for hypertension, smoking, type 2 diabetes, and blood tests.
To improve prediction efficiency and thus prescribe preventive treatment only to those who really need it, the researchers trained a neural network. They used data from 147,497 chest radiographs from 40,643 participants in a multicenter randomized trial of prostate, lung, colon and ovarian cancer screening.
The researchers tested the model using a second independent cohort of 11,430 outpatients who had routine outpatient chest radiographs and were potentially eligible for statin therapy. 1,096 of these patients had cardiovascular disease within 10 years of the radiograph; the rest had no such problems.
The results of the study showed that the risk level predicted by the neural network correlated significantly with the actual disease data: those patients who experienced problems had a significantly higher calculated risk level.