The ClinicalDigitalExperiences project aims to systematically evaluate clinician experiences of digital tools in hospitals and clinical settings across the east of Switzerland. We are conducting a qualitative interview study to evaluate experiences, strengths, weaknesses, and derive recommendations for the future digitalisation trajectory of clinical systems. We are particularly interested in experiences with diverse implementations of electronic health records, electronic patient records, knowledge management, decision support and artificial intelligence.
We have collaborated closely with the Human Behaviour-Change Project in developing a comprehensive knowledge system for behavioural science to advance the science of health-related behaviours. We have developed an explainable and semantically constrained system for predicting the outcomes of behaviour change interventions for hypothetical behaviour change scenarios based on machine learning over data and semantics from the published literature. The prediction system uses data extracted from the published literature together with the logical structure of the domain as represented in the Behaviour Change Intervention Ontology which we helped to develop.
Together with collaborators in Magdeburg, we are working on developing human-centered approaches to semantic (meaningful) Artificial Intelligence, driving a paradigm shift in the development of artificial intelligence technologies which embed meaningful semantics at the heart of predictive models through neuro-symbolic architectures. Our approaches allow for meaningful constraints on system performance to be expressed in terms that humans understand, enable the learning to benefit from prior human knowledge reducing the volumes of data needed for training, and give explanations for predictions based on meaningful categories at different hierarchical levels.
We are collaborators providing expertise on semantics and data harmonisation in the Sinergia project Connecting properties of the micro- and macrovasculature from multimodal imaging through genetics and deep learning to better understand vascular pathomechanisms and predict disease risk led by the Computational Biology group at UNIL. This project aims to use genome-wide association together with deep learning in processing images of vasculatures across different imaging modalities in order to derive novel predictive features that have biological meaning. We support the project’s complex data semantics and harmonisation efforts, and explore the association of learned features to clinical knowledge.
Are you interested in doing your master’s thesis project with us? Get in touch!