Researchers at the University of Oxford have developed an artificial intelligence (AI) tool that may help doctors better understand how high blood pressure damages different organs in different people - potentially paving the way for more personalised treatment in future.

The study, published in Circulation, used machine learning to analyse imaging and clinical data from more than 27,000 people in the UK Biobank, alongside external validation in a further 5,500 participants from the US-based Atherosclerosis Risk in Communities (ARIC) study.

This research was supported by the UKRI Medical Research Council, National Institute for Health and Care Research Biomedical Research Centre: Oxford, and the British Heart Foundation.

While clinically the focus remains only on blood pressure readings, the researchers combined hundreds of measures from heart scans, brain MRI, blood vessels, kidneys, liver, body composition and blood tests to build a broader picture of hypertension-related organ damage.

The team developed an AI-derived score, called "HyperScore", designed to estimate the extent of damage caused by high blood pressure across multiple organs before any major cardiovascular event. They also discovered that there appear to be six distinct patterns, or "HyperTrajectories", of hypertension-related disease that people develop, with some groups showing predominantly heart, brain, vascular, kidney or metabolic changes.

The researchers found that people with higher HyperScores were more likely to experience future cardiovascular problems, even when blood pressure readings alone did not fully distinguish risk.

Dr Mohanad Alkhodari, first author of the study and a current visiting researcher at the Radcliffe Department of Medicine's Clinical Cardiovascular Research Facility (CCRF), carried out this work during his DPhil in the department. He said: 'High blood pressure affects people very differently. Some individuals develop significant damage to the heart, brain or kidneys even when blood pressure is only mildly elevated, while others appear relatively protected despite longstanding hypertension.

'Our findings suggest that AI methods may help us move beyond treating hypertension based purely on blood pressure numbers, towards a more personalised understanding of how the disease affects the body.'

Potential to identify those at higher risk of major complications

The researchers say the work could eventually support earlier identification of people who are starting to develop problems that could lead to stroke, heart failure or kidney disease. The approach could also be used to help develop new and more personalised treatments. However, they caution that the approach is still at an early stage and is not yet ready for routine clinical use.

Professor Paul Leeson, Professor of Cardiovascular Medicine and senior author from the University of Oxford's Radcliffe Department of Medicine said: 'This study shows the potential of combining AI with multi-organ imaging to better understand the hidden effects of hypertension and how they vary between different people.

'What we have shown is that these computational approaches can uncover patterns of organ damage that are difficult to detect using blood pressure measurements alone.'

Brain changes strongest indicators of hypertensive-related damage

The study also found that brain changes detected on MRI scans were among the strongest indicators associated with hypertension-related damage. 'This reinforces growing evidence that high blood pressure can affect the brain long before symptoms appear,' said Dr Winok Lapidaire, co-first author of the study from the University of Oxford's Radcliffe Department of Medicine.

'Our preliminary follow-on research suggests that simpler clinical tests, such as ECGs or routine health measurements, could eventually provide similar insights without the need for extensive imaging.'

Jill Jones, Head of Global Health Strategy at UKRI Medical Research Council, added: 'This important study demonstrates the value of integrating data across multiple organ systems using advanced machine learning approaches.

'While further research is needed to bring these techniques to the clinic, this work represents a meaningful step towards one of MRC's key ambitions, enabling earlier identification of disease to support more timely and personalised intervention.'

ENDS

Notes to editors

The paper is titled: Contrastive machine learning to quantify hypertensive multi-organ damage and identify new disease phenotypes: a multi-national multi-modal study.

The study used data from:

UK Biobank Imaging Enhancement Study

Atherosclerosis Risk in Communities (ARIC) study

Please use the contact details below to obtain a copy of the paper.

Media contacts

Natalie Gidley, Communications Officer, Radcliffe Department of Medicine, University of Oxford: natalie.gidley@rdm.ox.ac.uk / 07789864645

Natalie Gidley Dip CIPR. MCIPR

Radcliffe Department of Medicine, University of Oxford

MRC Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford OX3 9DU

Email: natalie.gidley@rdm.ox.ac.uk