Artificial Intelligence

New AI Tool Predicts Heart Failure 5 Years Early

Is AI the future of early heart disease detection or a dangerous tool that risks deepening healthcare inequality?

Published APR 8

New AI Tool Predicts Heart Failure 5 Years Early

Story

Image credit: Unsplash

The Facts

In their research published in the Journal of the American College of Cardiology on Wednesday, researchers at the University of Oxford said they have developed an AI tool that can predict a person's risk of heart failure up to five years before the condition develops, using routine cardiac CT scans.
Daily Mail
Guardian
The tool detects textural changes in fat surrounding the heart that signal inflammation in the heart muscle — changes invisible to the human eye and undetectable through standard medical imaging tests.
Radcliffe Department of Medicine, University of Oxford
Independent
The AI was trained and validated using data from 72,751 adults across nine NHS centers in England between 2007 and 2022, predicting heart failure risk within five years with 86% accuracy.
JACC Journals
Daily Mirror
Patients in the highest risk group were around 20 times more likely to develop heart failure than those in the lowest risk group and had approximately a one-in-four chance of developing the condition within five years.
Daily Mail
Independent
Around 350,000 patients are referred for cardiac CT scans in the U.K. each year. Researchers are seeking regulatory approval to add the tool to routine scan analysis in NHS radiology departments.
Radcliffe Department of Medicine, University of Oxford
Daily Mirror
Researchers also plan to adapt the tool for use with any chest CT scan, not only cardiac scans. The British Heart Foundation funded the study.
JACC Journals
Guardian

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Techno-optimist narrative

This tool gives doctors a powerful weapon against one of medicine's most devastating conditions. Trained on over 70,000 patients across nine NHS Trusts, the tool identifies subtle changes in fat texture around the heart with 86% accuracy. Rolling this out nationwide could save countless lives by catching heart failure before irreversible damage sets in.

British Heart Foundation

Beginners in AI on X

Techno-skeptic narrative

Tools like this carry serious risks if built on biased data, and deploying them without an equity framework could deepen healthcare disparities. Algorithmic bias stems from flawed training datasets, poor variable selection and a lack of diverse research teams. Responsible AI development demands continuous bias monitoring, external validation across demographic groups and meaningful patient input.

Nature