AI breakthrough promises more accurate disease predictions by examining how risk factors interact
Finnish researchers develop machine learning tool that could revolutionise personalised healthcare by considering complex relationships between health factors.
Text Martti Asikainen, 1.9.2025 Photo: Martti Asikainen
A new artificial intelligence system could dramatically improve doctors’ ability to predict which patients will develop serious diseases by examining how multiple risk factors interact with each other, rather than treating them in isolation.
The breakthrough, developed by researchers at Finland’s Aalto University, represents a significant advance in personalised medicine that could help prevent conditions ranging from heart disease and diabetes to liver disorders.
Traditional risk assessment tools used by GPs and specialists typically examine individual factors such as cholesterol levels, blood pressure or smoking habits separately. But the new machine learning method, called survivalFM, recognises that human health is far more complex – and that these factors often influence each other in subtle but important ways.
“Today’s health data is incredibly complex – and so is human health,” said Dr Heli Julkunen, the study’s lead author and machine learning researcher at Aalto University. “Factors like age, lifestyle, and genetics rarely act alone; they also influence each other in subtle ways.”
The research, published in Nature Communications, could transform how the NHS and other health services identify patients at highest risk of developing serious conditions, potentially allowing for earlier intervention and more targeted prevention strategies.
Real-world testing shows promise
To test their approach, the Finnish team analysed data from the UK Biobank, which contains medical records, lab tests, lifestyle information and genetic data from around 500,000 people. The AI system was trained to predict the likelihood of developing 10 common diseases over a 10-year period.
Across most conditions tested, the new method outperformed existing prediction tools, proving especially effective at correctly identifying those who would go on to develop disease while accurately assigning lower risk scores to those who remained healthy.
The improvement comes from the system’s ability to examine combinations of risk factors simultaneously. For example, high cholesterol might pose a different level of threat depending on a patient’s age, genetic makeup or smoking habits – nuances that current tools often miss.
“Software that uses our method could help clinicians gain a better understanding of how combinations of risk factors, such as high cholesterol and smoking together, affect disease risk,” Dr Julkunen explained. “But this is only a simplified example, since the true novelty of the method lies in its ability to examine the simultaneous effects of many such risk factors.”
Transparency concerns addressed
One major advantage of the new approach is its interpretability – a crucial consideration as AI becomes more prevalent in healthcare. Unlike many “black box” AI systems, survivalFM was designed to be transparent, allowing doctors to understand exactly how it reaches its predictions.
“We see an increasing interest for interpretability in machine learning and AI applications, particularly in sensitive areas like healthcare,” said Professor Juho Rousu from Aalto University. “This method allows us to look at the model and directly see why this person was flagged as high risk.”
This transparency could be vital for gaining trust from both medical professionals and patients, addressing longstanding concerns about AI decision-making in healthcare.
Beyond healthcare applications
While the research focused on disease prediction, the method’s applications extend far beyond medicine. The team suggests it could be valuable in any field where timing matters, including engineering reliability studies and financial risk modelling.
The development comes as healthcare systems worldwide grapple with growing demands and limited resources. Tools that can more accurately identify high-risk patients could help prioritise care and prevention efforts, potentially reducing both human suffering and healthcare costs.
Current risk assessment tools used in the UK, such as QRISK3 for cardiovascular disease prediction, have proved valuable in guiding clinical decisions about prevention, screening and treatment. However, the Finnish research suggests there’s significant room for improvement through more sophisticated AI approaches.
The research was funded by the Research Council of Finland and the Technology Industries of Finland Centennial Foundation via Aalto University’s House of AI centre.
As AI continues to evolve, the challenge remains ensuring these powerful tools can be safely and effectively integrated into clinical practice – something the researchers believe their transparent approach helps address.
This article was published in collaboration with AIHealth.
Finnish AI Region
2022-2025.
Media contacts