Oxford AI Detects Early Heart Failure Risk From Routine CT Scans With 86% Accuracy Across 72,000 Patients
In Brief
Researchers at the University of Oxford have developed an AI system that detects subtle, invisible changes in heart fat from routine CT scans, predicting heart failure risk up to five years ahead with 86% accuracy across 72,000 patients.
Researchers at the University of Oxford have developed an artificial intelligence system that can estimate a patient’s risk of developing heart failure up to five years in advance, achieving 86% accuracy in validation across more than 72,000 patients. The approach does not require additional testing, specialist intervention, or new medical equipment, as it relies on cardiac CT scans that are already routinely performed in clinical practice.
The work, led by Professor Charalambos Antoniades and published in the Journal of the American College of Cardiology, addresses a long-standing limitation in cardiology: heart failure is typically diagnosed only after significant structural damage has already occurred, at which point preventive options are often limited. The proposed system shifts attention to early biological changes that precede visible symptoms by several years.
At the centre of the model is an unconventional data source: the fat surrounding the heart, known as pericardial adipose tissue. While traditionally overlooked in routine scan analysis, this tissue appears to reflect underlying inflammatory and metabolic changes occurring in the heart muscle itself.
According to the researchers, these fat deposits gradually alter their texture in response to stress in the cardiovascular system, creating patterns that are not detectable through standard human interpretation of imaging results. The AI system is designed to identify these subtle variations and translate them into a quantified risk estimate for future heart failure.
Reading Signals The Human Eye Cannot See
Cardiac CT imaging is widely used across the UK’s National Health Service to investigate chest pain and assess coronary artery disease, with hundreds of thousands of scans performed annually. In typical clinical workflows, radiologists focus primarily on arterial blockages and visible abnormalities, while surrounding fat tissue receives limited analytical attention.
The Oxford model repurposes this overlooked data layer by analysing textural features within pericardial fat. Using machine learning techniques trained on anonymised CT data from more than 59,000 NHS patients, the system learned to associate specific imaging patterns with later development of heart failure over long-term follow-up periods.
In validation testing involving 13,424 additional patients, the model produced an 86% accuracy rate in predicting five-year heart failure risk. Individuals classified in the highest-risk group were found to be approximately 20 times more likely to develop the condition than those in the lowest category, with an estimated one-in-four probability of onset within five years.
Importantly, the system generates risk scores automatically, without requiring manual input from clinicians. This positions it as a potential decision-support tool rather than a replacement for existing diagnostic processes.
From Cardiac Scans To Any Chest CT — And A Path To The NHS
The broader ambition of the research is to extend the technology beyond cardiac-specific imaging. The team is currently working on adapting the model to analyse standard chest CT scans, including those used in lung cancer screening and respiratory diagnostics. Given the significantly higher volume of chest CT imaging compared with cardiac-specific scans, such an adaptation could substantially increase the reach of the system.
Clinically, the implications are tied to earlier intervention. By identifying high-risk patients years before symptoms appear, healthcare providers could adjust monitoring strategies, initiate preventative treatments earlier, and prioritise resources more effectively. With heart failure already affecting more than one million people in the UK, the potential impact on long-term healthcare demand is considerable.
Plans are now underway to seek regulatory approval for integration into routine radiology workflows within the NHS. If adopted, the system would operate in the background of standard imaging procedures, producing automated risk assessments at no additional cost or change in scanning protocols.
The research was supported by the British Heart Foundation and the National Institute for Health and Care Research Biomedical Research Centre in Oxford. It reflects a broader shift in medical imaging, where artificial intelligence is increasingly used not only to detect existing disease but also to infer future risk from subtle, previously underutilised biological signals embedded in routine scans.
Disclaimer
In line with the Trust Project guidelines, please note that the information provided on this page is not intended to be and should not be interpreted as legal, tax, investment, financial, or any other form of advice. It is important to only invest what you can afford to lose and to seek independent financial advice if you have any doubts. For further information, we suggest referring to the terms and conditions as well as the help and support pages provided by the issuer or advertiser. MetaversePost is committed to accurate, unbiased reporting, but market conditions are subject to change without notice.
About The Author
Alisa, a dedicated journalist at the MPost, specializes in crypto, AI, investments, and the expansive realm of Web3. With a keen eye for emerging trends and technologies, she delivers comprehensive coverage to inform and engage readers in the ever-evolving landscape of digital finance.
More articles
Alisa, a dedicated journalist at the MPost, specializes in crypto, AI, investments, and the expansive realm of Web3. With a keen eye for emerging trends and technologies, she delivers comprehensive coverage to inform and engage readers in the ever-evolving landscape of digital finance.
