The first trustworthy artificial intelligence (AI) model for tailoring management of heart failure patients is being developed by the AI4HF project, launched this month in Utrecht, the Netherlands. The ground-breaking project is being conducted by a consortium of international partners including the European Society of Cardiology (ESC).
“An innovative personalised risk calculator will be created to pinpoint the most beneficial treatments for each heart failure patient,” said AI4HF coordinator Professor Folkert Asselbergs of Amsterdam Heart Centre. “The model will incorporate data on symptoms and lifestyle behaviours, blood tests, electrocardiograms and cardiac imaging, and will be co-created with patients and clinicians.”
Heart failure is the leading cause of hospitalisation in people over the age of 65.1 About half of hospital readmissions are related to co-existing conditions, multiple medications, and disabilities related to heart failure.2 The prognosis of heart failure is worse than many forms of cancer.3 Management is challenging as the condition has many causes and manifestations, from decreased quality of life to regular hospitalisations, heart attack, and premature death.
Professor Asselbergs said: “A personalised medicine approach is needed where we tailor the advice and treatment we give to individual patients including medication, diet, exercise, pacemakers and cardiac resynchronisation therapy based on early prediction of their risk of poor outcomes. Projections indicate that the number of patients living with heart failure will be 46% higher in 2030 due to an ageing population and unhealthy lifestyles, so it’s important that we act now.”4
The largest-ever dataset of heart failure patients will be harnessed to develop the AI model during this pioneering four-year project, which is funded by the European Health and Digital Executive Agency (HaDEA) and involves 16 organisations around the world.5 The inclusion of hundreds of thousands of patients with heart failure in Europe, South America and Africa will result in novel analyses across populations, clinical settings and ethnic groups.
To achieve its ambitious goal, the AI4HF consortium will leverage a unique blend of resources and tools. Real-world health data will be obtained from BigData@Heart (www.bigdata-heart.eu) and integrated using the FAIR4Health platform (www.fair4health.eu) following best practice recommendations for building trustworthy AI tools established by FUTURE-AI (www.future-ai.eu).
Patient privacy will be preserved by using a federated learning approach to train the AI model. This means that the model will be sent to clinical centres in Europe, Africa and South America for onsite training using local data, and the resulting individual models will be combined at a central location. Patient data will always remain at the local centre and will not be shared. A cutting-edge AI-patient interface will be created to provide clear and accessible information on personal risk and ways to lower that risk including lifestyle changes.
Professor Asselbergs said: “AI4HF promises to benefit patients with heart failure by adapting management to individual needs. In addition, a state-of-the-art AI passport will be introduced which uses new methods to continuously update the model following its deployment in real-world practice and will act as a framework for developing trustworthy AI solutions across all areas of health.”