Cardiovascular diseases are the leading cause of death and disability in Europe. There are now many AI applications that can assist doctors with diagnosis and treatment, but their use is fragmented, and there is no common European approach to evaluating them. As a result, hospitals often do not know which applications are reliable, for which patients they are effective, and how they fit into the care process. This large-scale European study aims to change that.
Interpreting an ECG requires specialized knowledge that is not available everywhere or at all times. As a result, conditions such as asymptomatic atrial fibrillation or early-stage heart muscle disease sometimes go undetected. At the same time, many patients undergo a cardiac ultrasound as a precaution, even though the results later show nothing abnormal.
In the first clinical study led by the UMCG, a specific AI application analyzes the standard ECG recorded in outpatient clinics and emergency departments. It checks for recognizable patterns associated with arrhythmias, cardiomyopathies, and structural heart defects. In all cases, the physician continues to make the medical decisions.
The researchers are investigating whether this application makes it possible to determine earlier which patients need a cardiac ultrasound. They are also examining whether asymptomatic atrial fibrillation can be detected earlier, so that targeted cardiac rhythm monitoring can follow. Furthermore, they are investigating whether this allows for faster decisions regarding next steps for patients with chest pain in the emergency department. Finally, the researchers are determining when an ECG check is advisable before starting, for example, oncology medication, so that treatment can begin sooner if necessary.
The study is being conducted at five hospitals in the Netherlands, Spain, Greece, Estonia, and the Czech Republic, as well as at a cardiology clinic in Slovakia, and involves several thousand ECGs.
Echocardiography and MRI measurements of the heart are indispensable for diagnosis and monitoring. These measurements require a great deal of manual work, and the results can vary between evaluators and between hospitals. This makes it difficult to compare studies and leads to delays in receiving results.
In the second clinical study, two AI applications automatically perform echocardiogram and MRI measurements of the heart. A lab technician or physician reviews and approves each measurement before it is included in the report. This involves 2,000 to 3,000 echocardiograms and 500 to 1,000 MRI scans in hospitals across five European countries, including the Netherlands. The goal is to determine whether this approach reduces the time required for measurements and report generation and whether it leads to less variation in the assessments.
Pim van der Harst, professor and head of the Department of Cardiology at the UMCG, is leading both studies. He emphasizes that the application will be introduced step by step: “Only once it becomes clear that the software’s results are reliable and do not influence patient care will healthcare providers begin using the results in practice. And then there will be clear protocols regarding how to respond to a specific finding.”
According to Van der Harst, it can only be considered a success if it offers benefits in practice. “Just because an algorithm performs well in a study doesn’t mean it benefits the patient. We’re examining what changes in the doctor’s office and in the emergency room: do we make the diagnosis sooner, do we spare patients unnecessary tests, and do healthcare providers have more time? And does the same hold true in other European hospitals? Only then can we say that AI works for cardiac care.”
Van der Harst is optimistic. “If expectations are met, heart diseases will be detected earlier, patients will undergo fewer tests that yield no results, the results of an echocardiogram or MRI will be available faster, and healthcare providers will spend less time on manual measurements. The results will be entered into a European database, so that other hospitals and governments can also see which applications work, for whom, and in what setting.”
CHAIN (European Cardiovascular Health Data and AI Network) is a consortium of 53 partners from 20 European countries: universities, hospitals, technology companies, ministries of health, and patient organizations. Six clinical studies in 13 countries demonstrate the benefits of AI in everyday healthcare. The UMCG is leading two of these studies. Within CHAIN, Pim van der Harst also serves as chair of the Pilot Adoption Board. The European Union’s EU4Health program is contributing 20 million euros to CHAIN.
