High blood pressure has long been regarded as one of the most significant risk factors for cardiovascular disease. However, the impact of hypertension appears to vary greatly from person to person. Whilst one patient may develop damage to the heart, brain or kidneys even with relatively mild increases in blood pressure, another may remain largely symptom-free for years. Researchers at the University of Oxford have now developed an AI model that reveals these differences.
The new model, called HyperScore, can assess the extent of organ damage caused by high blood pressure even before serious cardiovascular complications arise. According to the researchers, this could be an important step towards more personalised prevention and treatment.
Beyond the blood pressure monitor
In everyday practice, hypertension is primarily assessed on the basis of blood pressure readings. However, according to the researchers, this provides only a limited picture of the actual damage to health caused by the condition. To develop HyperScore, they combined hundreds of clinical and imaging parameters. They looked at changes in, amongst other things, the heart, brain, kidneys, blood vessels, lungs, liver and metabolism.
Using machine learning, the researchers analysed data from over 27,000 participants in the UK Biobank. The results were then validated using data from a further 5,500 participants from the US Atherosclerosis Risk in Communities study. The analysis showed that hypertension does not develop in a single, uniform way. Instead, the AI model identified six distinct disease patterns, referred to by the researchers as ‘HyperTrajectories’. Some patients showed abnormalities primarily affecting the heart, whilst in others it was the brain, blood vessels, kidneys or metabolic processes that were most severely affected.
Six patterns of organ damage
A striking finding, published in Circulation, is that the severity of organ damage is not always directly linked to the level of blood pressure. People with similar blood pressure readings could differ significantly in their risk of future health problems. The researchers found that individuals with a high HyperScore had a significantly greater risk of cardiovascular conditions such as strokes, heart failure and kidney disease. This risk remained evident even when traditional blood pressure measurements failed to adequately distinguish between high- and low-risk patients.
According to lead author Mohanad Alkhodari, this underlines the need not to view hypertension solely as a number on a blood pressure monitor. AI can help to better understand how the condition affects individual patients’ bodies, thereby enabling a more personalised approach.
The brain: a key indicator
One of the most striking findings of the study concerns the brain. Changes visible on MRI scans proved to be among the strongest indicators of hypertension-related damage. According to co-researcher Winok Lapidaire, this is consistent with growing evidence that high blood pressure can affect the brain long before patients develop symptoms. This makes early detection particularly important.
The researchers emphasise that HyperScore is currently still a research tool and is not suitable for routine clinical use. Further validation is required before the method can be deployed in everyday clinical practice. Nevertheless, they see great potential. In the future, similar AI models could help to identify patients at risk of serious complications at an earlier stage. Furthermore, the technology could contribute to the development of new, more personalised treatment strategies for hypertension.
According to Professor of Cardiovascular Medicine Paul Leeson, the research demonstrates how the combination of AI and advanced imaging can reveal hidden patterns of organ damage that remain invisible with traditional blood pressure measurements. This could open up a new way of better understanding the consequences of hypertension and treating them in a more targeted manner.
Wearable hypertension alerts
Earlier this year we reported on a study that suggested that the Apple Watch's new hypertension notification feature could help identify previously undiagnosed high blood pressure, but its effectiveness varies considerably across different population groups. Approved by the U.S. FDA in 2025, the feature uses optical sensors to analyze blood flow patterns and alerts users when their data indicate a possible risk of hypertension.
Researchers from the University of Utah and the University of Pennsylvania found that receiving an alert substantially increases the likelihood that a user has hypertension, particularly among older adults. For people aged 60 and over, an alert raises the probability from 45% to 81%, while the absence of an alert still leaves a 34% chance of having hypertension. Similar differences were observed across racial and ethnic groups. The researchers conclude that the Apple Watch could become a valuable population screening tool, but stress that it should be viewed as a prompt for medical evaluation rather than a replacement for clinical diagnosis.
References
Circulation (research)
