The CAIM projects from five different medical disciplines combine the opportunities of data-driven medicine and the latest data science and AI technologies to solve some of today's greatest challenges in acute medical care, the prevention of high prevalence health risks, and the provision of tailored just-in-time care for chronic diseases. They envision tangible patient benefits that can realistically be achieved within the next decade through the joint expertise of interdisciplinary project teams. Projects will start on 1 March 2022.
Myocarditis is an inflammation of the myocardium and caused by different underlying pathogens, including viruses. It is believed to be responsible for 20 percent of all sudden cardiac arrests worldwide and is the third leading cause of death in young adults. Further, myocarditis has made headlines in the last two years as a consequence of Covid 19 infection (or its vaccination). Due to its heterogeneous presentation, myocarditis is a challenge to diagnose based on solely clinical information. Since myocarditis can cause heart failure or dangerous arrhythmia, it is important to identify affected patients quickly and reliably to initiate the appropriate treatment. Cardiac Magnetic Resonance (MR) has become the primary diagnostic tool in this clinical setting and inherits the potential to improve risk stratification.
Prof. Dr. med. Christoph Gräni, PhD, Director of Cardiac Imaging at the Department of Cardiology at Inselspital, Bern University Hospital, together with his PhD student Yasaman Safarkhanlo, a physicist in Cardiac Imaging, now aims to develop a machine learning algorithm that allows automatic cardiac MRI analysis to classify patients according to their risk of a major adverse cardiovascular event. The project team is also collaborating with Harvard University in Boston, USA.
Up to 15 percent of the population develops kidney stones - extremely painful crystallizations of salts in the kidney - which often require surgical removal. Since kidney stones frequently recur, they cause high annual treatment costs as well as a severe impairment of life quality for those affected. Depending on the type of kidney stone, recurrence can be largely averted with targeted prevention. But so far, stone types cannot be distinguished sufficiently well to design prevention programs tailored to each patient. In addition, individual prediction of recurrence risk is currently not possible.
Prof. Dr. med. Daniel Fuster, Department of Nephrology and Hypertension at Inselspital, in collaboration with PD Dr. Rémy Bruggmann, Interfaculty Bioinformatics Unit, University of Bern, therefore, wants to develop a machine learning tool that can determine kidney stone type and recurrence risk based on demographic information and urinary parameters which are commonly available in clinical routine.
Multiple sclerosis (MS) typically affects younger people in the prime of their lives (onset often between 20 and 40 years of age). The chronically progressive neurological condition necessitates lifelong care in a medical center. To ensure that patients maintain good quality of life, drug therapy must be adapted to the individual disease progression; this requires regular MRI examinations to detect change in lesion load. However, if a patient changes care centers (e.g., due to a move) or if a hospital upgrades its imaging technology, the different imaging parameters can impede this seamless monitoring because the image data are not comparable.
Dr. Richard McKinley, University Institute of Diagnostic and Interventional Neuroradiology (DIN) and Support Center for Advanced Neuroimaging (SCAN) Inselspital, and Dr. med. Piotr Radojewski, DIN, SCAN and Translational Imaging Center (TIC) sitem-insel, intend to change that. In collaboration with the Department of Neurology, Inselspital, they are working on a 3D segmentation of various MRI modalities that can seamlessly compare different MRI imaging standards using a form of geometric deep learning. The goal is for every MS patient to receive individually adapted treatment, independent of location and parameters used.
A single nurse is in sole charge of up to 30 inpatients during a night shift. Especially in psychiatric hospitals, patients often require nursing support and intervention overnight. However, if a patient raises a nursing alert, the nurse has no way of knowing how urgently he or she needs help. This can be very stressful if, for example, the nurse is helping at one end of the ward and receives two calls from the other end.
Prof. Tobias Nef, Head of Gerontology at the ARTORG Center for Biomedical Engineering Research, University of Bern, and Prof. Dr. med. Stefan Klöppel, Head of Geriatric Psychiatry, University Psychiatric Services Bern (UPD), therefore propose to develop a digital care assistant that supports caregivers in assessing the urgency of an intervention for a patient. The system is to analyze multimodal sensor data from the rooms and identify situations that require urgent nursing intervention (for example, if a person has suffered a fall or is experiencing anxiety) so that caregivers are not exposed to undue stress. The team works closely with the NeuroTec research platform at sitem-insel.
Every woman experiences menopause, with varying degrees of symptoms. What they all share is an increased risk of obesity, diabetes, osteoporosis, cardiovascular disease as well as cancer. Today, women are no longer ready to simply accept these risks, but seek individualized care during menopause, allowing them the greatest possible autonomy over their own health and enabling them to self-mitigate menopausal risks.
Prof. Dr. med. Petra Stute, Head of the Menopause Center, Department of Gynecology and Obstetrics, Inselspital, wants to empower women in midlife to determine their personal risk portfolio for acute or chronic diseases via a digital health app. Using a common smartwatch, the app collects various data types from the woman and is designed to, for instance, indicate whether she is currently susceptible to an infection or whether her risk of diabetes is increasing. Prof. David Ginsbourger from the Institute of Mathematical Statistics and Actuarial Sciences at the University of Bern works on probabilistic prediction of risk parameters using data from different sources including wearables and questionnaires, with a focus on quantifying and reducing associated uncertainties.