Hyper-modal and Unobtrusive Detection of Atrial Fibrillation in Geriatric Patients (HypAFib)

• Applicants: Professor Dr. Leo Cornelius Bollheimer; Professor Dr.-Ing. Steffen Leonhardt
• Subject Area: Biomedical Systems Technology ; Biogerontology and Geriatric Medicine
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 527346822

Cardiovascular diseases are the most common cause of death for people aged 65 years and above. One of these cardiovascular diseases is atrial fibrillation (AFib), which is associated with an increased risk for stroke, heart failure, dementia, and an overall increased mortality. The prevalence of AFib increases with age and is 17.8% in people aged 85 years and above. Of particular interest is asymptomatic AFib: screening studies have shown that depending on the selected study population (regarding pre-conditions and risk factors), 10.5% up to 30% of the participants had previously undiagnosed AFib. The objective of the HypAFib project is the research on data fusion methods for early AFib detection based on unobtrusive and "hyper-modal" measurements of physiological signals in bed. The proposed hyper-modal measurement system comprises both multi-modal sensors integrated into the bed and a ceiling-mounted multi-spectral camera system. The bed-integrated sensors employ capacitive electrocardiogram, reflection photoplethysmography, and magnetic impedance monitoring. The multi-spectral camera system includes both photoplethysmography imaging (PPGI) as well as infrared thermography (IRT). Due to its unobtrusiveness, the proposed hyper-modal measurement system is especially well suited for patients who do not tolerate conventional contact-based monitoring using ECG (e.g. those in delirium or with dementia). Our previous work has demonstrated the suitability of the different modalities to measure physiological signals. Therefore, the focus of the HypAFib project is on research of fusion algorithms of several weak and potentially corrupted data streams. On the one hand, multi-spectral image fusion and multi-spectral 3D-tracking should improve robustness of the camera system against motion artefacts. On the other hand, robustness should also be achieved through an intelligent fusion algorithm which recognizes links among the multiple modalities, rejects invalid signals, and fuses valid signals to robustly compute physiological parameters and signal features. AFib detection based on these parameters and features will be performed by a classifier, for which machine learning and especially deep learning approaches will be investigated. The HypAFib project plans for one lab study with healthy subjects and one clinical trial with AFib patients. Data acquired in these two studies will be used for design and evaluation of the algorithms. Additionally, data augmentation techniques will be investigated to obtain the necessary amount of training data for machine learning and deep learning. The HypAFib project is intended to eventually offer patients an easy and casual screening for AFib during a stay in hospital or in a nursing home. Such early AFib detection aims at avoiding AFib-related secondary diseases and thus to increase life expectancy and life quality.

• DFG Programme: Research Grants