AI learns to predict 10-year risk of heart disease from a single chest X-ray

Scientists developed a deep-learning model that, from a single chest X-ray, forecasts a 10-year risk of death from heart disease. The study results presented at the annual meeting of the Radiological Society of North America.

Standard cardiovascular risk assessment is used to guide prophylactic statin therapy. The classic analysis relies on statistical methods that account for demographics, biomarkers, pathologies and lifestyle factors. The researchers say these data are not always available, reducing prediction accuracy.

To improve risk prediction for heart disease, the researchers developed a deep-learning model. They trained it on 147,497 chest X-rays from 40,643 participants in a multi-centre randomized screening trial for cancer of the prostate, lungs, colon, and ovaries.

They then validated the algorithm on data from 11,430 other patients who previously might have been candidates for statin therapy. Of these, 1,096 developed cardiovascular disease in the 10 years after the X-ray. The remaining patients did not have such problems.

As a result, the AI model predicts risk with high accuracy. Among patients who went on to develop cardiovascular disease the predicted risk level was significantly higher.

According to the study’s lead author, Jacob Weiss, the algorithm could help identify people who would benefit from starting statin therapy.

“Based on a single existing chest X-ray, our deep-learning model predicts future serious adverse cardiovascular events with comparable performance and added value to the established clinical standard,” he said.

Weiss added that further studies are needed to validate the model, which in the future could serve as a decision-support tool for clinicians.

Earlier in September, researchers created an AI, that identifies diseases from X-ray images.

Subscribe to ForkLog news on Telegram: ForkLog AI — all news from the AI world!