Myocarditis (inflammation of the heart muscle) is usually caused by viruses, e.g. in Covid-19 disease. However, it can also be induced by medication, toxic substances or in the context of a rheumatological disease. Clinical assessment is difficult due to widely varying symptoms, from fatigue to chest pain, palpitations, shortness of breath, and, rarely, sudden cardiac death, the latter associated with sports activity. Nowadays, cardiac magnetic resonance imaging (CMR) is usually performed when myocarditis is suspected. In some cases, the data obtained from these scans do not provide a sufficiently personalized risk assessment and respective optimal treatment options.
Review study points to new approaches
In a scientific review in collaboration with the University of Tübingen, the Bristol Heart Institute and Harvard Medical School, a research group led by Prof. Christoph Gräni, MD, PhD, from Inselspital and the University of Bern assessed various CMR parameters in terms of their importance for diagnosis, prognosis and monitoring of myocarditis.
“From the comparison of the different used diagnostic tools, we can derive novel approaches for future research and development. Next, we will determine how artificial intelligence (AI) can assist us in a rapid and comprehensive evaluation of the many different clinical parameters and image data,” according to Christoph Gräni. “To this end, I am pleased that we have received funding from the Bern Center for Artificial Intelligence in Medicine (CAIM) to pursue this promising research direction.”
Make complex cardiac function data readable with AI
Using CMR, more than 1000 measurements can be collected per patient, including parameters on anatomy, tissue characterization of the heart muscle and pericardium (e.g., inflammation or scars), and heart muscle function data. Physicist Yasaman Safarkhanlo, who is a PhD student under the supervision of Prof. Gräni at the Department of Cardiology at the Inselspital, explains: “Only AI can evaluate these many variables quickly in their entirety. We want to let the data speak to better understand what exactly happens during myocarditis. This is a novel approach that does not start from our previous understanding of physiology and looks for a known feature in the images. With our project, we're rather starting from the data aspect to see what new correlations we discover on the images – so, ultimately, we can allow better treatment in the future."