Research

Projects 2023-2025

CAIM is driving the integration of AI into healthcare. In spring 2024, we selected these five promising projects in anesthesiology, cancer care, cardiology, eye care and pathology for funding in our Fellowship program.

Anaesthesiology

Machine Learning for Causal Inference and Discovery in Anaesthesiology

Markus Huber

Department of Anaesthesiology and Pain Medicine, Inselspital, Bern University Hospital

General anesthesia is common but avoiding low blood pressure (hypotension) is crucial. Markus Huber's project aims to unravel the cause-effect relationship between hemodynamic management during anesthesia, hypotension and patient outcomes using causal inference and machine learning. This understanding can improve patient health and prevent complications like heart attacks or acute kidney injury.

Cancer Surgery

Polarimetry for use in Intraoperative Diagnostics - PolaSight™

Pablo Márquez-Neila

ARTORG Center for Biomedical Engineering Research, University of Bern

During pancreatic surgery, assessing tumor margins accurately is tough. Pablo Márquez-Neila is developing a polarimetry-based AI system that analyzes tissue in real time, helping to ensure complete tumor removal and potentially increasing surgery success rates. Inspired by space technology, this project is a collaboration between the ARTORG Center, The Institute of Tissue Medicine and Pathology and the Center for Space and Habitability at the University of Bern.

Cardiology

AI Acceleration of 3D Cardiac bSSFP MRI

Eva Sophia Peper

Department of Diagnostic, Interventional and Pediatric Radiology, Inselspital and University of Bern

Currently, cardiac imaging relies on 2D MRI. 3D MRI requires long image reconstruction times due to the need for scan acceleration methods. Eva Peper proposes using neural networks to speed up image reconstruction from hours to seconds. This AI-powered approach promises fast 3D cardiac imaging, enhancing diagnostic accuracy and patient comfort. The project collaborates with the Translational Imaging Center (TIC) at sitem-insel.

Eye Care

kerAccurate Ophthalmic Planning: Multi-fidelity DNNs, Transfer Learning & Bayesian Optimization

Miguel Angel Ariza Gracia

ARTORG Center for Biomedical Engineering Research, University of Bern

By 2050, more than 5 billion people will be affected by a refractive condition that impairs vision. Not all patients will be able to undergo laser surgery to restore their vision. Miguel Ariza is developing a platform technology for precision treatment with eye implants. It minimizes complications and enhances vision restoration outcomes. The project is being carried out in collaboration with ophthalmologists from the Inselspital Department of Ophthalmology.

Pathology

Improving MetAssist, a deep-learning model for lymph node metastasis detection, through learning from human feedback 

Amjad Khan

Institute of Tissue Medicine and Pathology (ITMP), University of Bern

Detecting lymph node metastases in colorectal cancer patients is challenging for pathologists. Amjad Khan proposes to integrate text feedback from pathologists with image predictions using vision-language models into an AI algorithm developed at the ITMP in collaboration with the Inselspital, Bern University Hospital.

Projects 2022-2024

These five projects were selected for funding by the Medical Faculty of the University of Bern through the CAIM Research Project Fund out of 20 submitted proposals.

Cardiology

Artificial Intelligence Analysis of Quantitative CMR Output Data to Better Risk Stratify Patients with Suspected Myocarditis

Prof. Dr. Dr. med. Christoph Gräni / Yasaman Safarkhanlo

Department of Cardiology, Inselspital, Bern University Hospital

Patients with suspected myocarditis (an inflammation of the myocardium) can present very heterogeneously from being asymptomatic to heart failure or arrhythmia. The 'Inflammatory Cardiomyopathy Bern Registry (FlamBeR)’ and CMRMyo registry evaluates the diagnostic and prognostic value of cardiac magnetic resonance (CMR) images in patients with suspected myocarditis with a focus on risk improving stratification. In addition to traditional CMR parameters, this project will evaluate the role of modern CMR technologies, such as T1- and T2-mapping and CMR-feature tracking (CMR-FT). CMR-FT is a finely granulated technology, which better mirrors myocardial function (compared to traditional left ventricular ejection fraction). As a post-processing image analysis with ample quantitative secondary output data, CMR-FT may include important information regarding risk stratification.

Nephrology

Machine learning models in the prediction of kidney stone recurrence

Prof. Dr. med. Daniel Fuster / PD Dr. Rémy Bruggmann

Department of Nephrology, Inselspital / Interfaculty Bioinformatics Unit, University of Bern

Kidney stones affect 10-15 % of the population worldwide and both prevalence and incidence are on the rise. The recurrence rate is high, necessitating frequent hospital visits with urological interventions that result in enormous healthcare expenditures. Risk of recurrence and stone type, which dictates prophylactic treatment, is currently poorly predictable. The aim of this project is to develop predictive machine learning (ML) tools for kidney stone type and recurrence risk using both demographic information and 24-hour urine biochemistry. 

Neuroradiology

Longitudinal follow-up of multiple sclerosis patients: effect of scanner and sequence changes using resolution adaptive neural networks

Dr. Richard McKinley / Dr. med. Piotr Radojewski

Department of Neuroradiology & Support Center for Advanced Neuroimaging (SCAN), Inselspital / Translational Imaging Center (TIC), sitem-insel

Magnetic Resonance Imaging is a vital tool in the follow-up of multiple sclerosis patients, allowing neurologists to observe disease-related changes in a patient`s brain. Of particular importance is the detection of new or enlarged lesions: this time-consuming task is currently performed by neuroradiologists. Automated tools provide a potential solution to this problem, but it is not yet known how such tools perform when a patient changes hospital or after a scanner upgrade. We will compare standard and state-of-the-art techniques in lesion detection, and analyse their robustness to these changes.

Gerontology

Development and evaluation of a digital care assistant for an old age psychiatry setting

Prof. Dr. Tobias Nef / Prof. Dr. med. Stefan Klöppel

ARTORG Center for Biomedical Engineering Research, University of Bern / Old Age Psychiatry, Bern University Psychiatry Services (UPD)

A shortage of skilled workers and pressure to reduce costs are characteristics of care for the elderly. In this project, we plan to develop and evaluate a digital care assistant to support nurses in acute care, old age, psychiatry settings. We will equip patient rooms at the UPD Waldau with sensors (e.g., pressure sensor, ambient sensors), subsequently training artificial intelligence (AI) algorithms to detect when nurses should intervene. We plan to test whether a digital nursing assistant a) is perceived as useful by nursing staff, b) reduces subjectively perceived stress, and c) reduces the distance traveled by nursing staff. 

Gynaecology

Personalized quantification of risks in menopausal women with mobile data and statistical machine learning

Prof. Dr. med. Petra Stute / Prof. Dr. David Ginsbourger

Department of Gynaecology, Inselspital / Institute of Mathematical Statistics, University of Bern

Menopause affects all women. It is a time when a woman’s body stops producing the sex hormones estrogen and progesterone which usually takes place around the age of 51. As a result, women can have noticeable symptoms like hot flashes, sleeplessness, achy joints, and weight gain. However, more critically menopause can also put women at a greater risk of chronic non-communicable diseases like cardiovascular disease, diabetes mellitus, cancer, osteoporosis, and dementia. This project will develop a digital medical device App called Navina+ for women in or after their menopause. The Navina+ App will work with data from a smart tracker, e.g. FitBit®, to drive statistical machine learning models to predict future risks of chronic disease. The woman will then be given suggestions of action, e.g. life-style changes, hormone replacement therapy etc., depending on the severity and type of risk and the informed self-management choices a woman wishes to make. 

The CAIM Research Fund Call is now closed.

Stay tuned for future calls.