Research
Our group focuses on advancing digital biomarkers—objective, quantifiable physiological and behavioral data collected through digital devices. These biomarkers enable continuous, non-invasive disease monitoring at a fraction of the cost of traditional episodic follow-ups and clinic visits. By frequently tracking disease progression over extended periods, digital biomarkers support proactive care, enabling timely adjustments to treatments and medications while reducing the need for in-person appointments. Additionally, these technologies are integral to hybrid and fully decentralized clinical trials, offering the ability to capture surrogate endpoints and provide critical insights into the efficacy and effectiveness of investigational therapies.
Projects
This project aims to develop and evaluate ALEX, an open-source digital health assistant that supports remote asthma monitoring and personalized self-management in adolescents. The platform integrates multimodal data from connected respiratory devices and wearables to generate continuous insights into asthma control and guide user-friendly coaching. The study assesses the feasibility and scalability of this approach for pediatric asthma care, with a focus on low- and middle-income country settings.
This project, in collaboration with our industry partner, CARE, investigates how biological age can be accurately estimated using blood biomarkers and advanced predictive models, and how personalized digital health interventions can influence the aging process. By combining longitudinal real-world data from the CARE platform with state-of-the-art linear and machine learning approaches, the project aims to support more precise, data-driven strategies for healthy aging beyond chronological age.
EvoMorph is a framework for generating counterfactual explanations for time series prediction models using multi-objective evolutionary optimization. Instead of arbitrary perturbations, EvoMorph modifies time series while preserving morphological characteristics and physiological plausibility, enabling interpretable explanations of how time-series predictions could change. The resulting counterfactuals can also be used to probe model uncertainty by exploring realistic signal variations, revealing low data-density regions where predictions become unstable.
In collaboration with the University Hospital Zurich, this project investigates machine learning methods for improving diagnostic utility in central nervous system disorders using routinely collected cerebrospinal fluid (CSF) biomarkers. The research focuses on developing models that support classification across major CNS disease categories while accounting for uncertainty in model predictions. These methods are translated into a clinician-facing inference tool designed to support reliable and interpretable diagnostic decision-making.
This project aims to develop a scalable and accessible method for estimating biological age using lifestyle behaviors captured by digital devices such as smartwatches and wearables. Lifestyle behaviors are closely linked to our health and aging, and by analyzing continuously collected accelerometer data to quantify physical activity, circadian rhythms, sleep patterns, and movement behaviors, we can gain insights into how these factors influence biological aging.
The sensitivity of voice to fluid accumulation inspires the development of a novel monitoring tool for heart failure self-management. This longitudinal study aims to predict the health status of stable heart failure patients using audio data collected through mobile devices.
CLAID is an open-source initiative to develop, validate and share AI models, digital biomarkers and healthcare applications. Our goal is to transfer research findings from the lab into the real-world. We invite researchers, clinicians and developers to use and contribute packages and to participate in our community.
While many commercial wearables are available for circadian rhythm monitoring, their efficacy remains underexplored. By conducting a two-week longitudinal observational study, we validated two commercial devices, finding they outperformed research-grade tools for circadian rhythm monitoring.
This project explores the digital potential of assessing inflammatory biomarkers, such as C-reactive protein (CRP), in systemic inflammation. Currently, CRP measurement relies on invasive blood tests, and no noninvasive, digital methods exist for continuous remote monitoring. Developing a noninvasive, digital CRP biomarker offers unprecedented potential for remote monitoring, early detection of exacerbations, and timely intervention.
This project aims to develop a cutting-edge smartphone-based cough monitoring system designed to detect, record, and analyze cough patterns in real-time. By leveraging advanced algorithms and the smartphone’s built-in sensors, the system will provide an accessible, non-invasive tool for continuous respiratory health monitoring.