CIBM DS CHUV-HUG SECTION

Imaging for Precision Medicine

Section Head: Dr. Jonas Richiardi, MER (CHUV-UNIL)

The CIBM Data Science CHUV HUG Imaging for Precision medicine Section focuses on developing machine learning methods to integrate imaging with other types of data, in particular -omics data. The ambition is to leverage this combination to improve diagnostic, prognostic, subtyping, and treatment planning for individual patients. To support this aim, the section helps develop data science tools and infrastructure to interface with hospital IT before fundamental research, during fundamental research, and after fundamental research, including methods reusable across other areas of biomedical imaging.

RESEARCH TOPICS

Advanced Stroke Analytics

Description: Stroke is a leading cause of death and disability worldwide. New treatments drastically improve patient outcomes, but increasingly need personalised views of brain perfusion and clinical factors. Current software to process stroke images is not robust, leading to incorrect estimates of brain tissue at risk, and gives imprecise patient recovery prognosis, leading to unnecessary surgery and high societal and economic costs for rehabilitation. This project aims at developing multimodal machine learning algorithms robust to heterogeneous imaging protocols, in order to improve ischemic tissue volumetry and localization and improve treatment planning for individual patients. We also aim to improve outcome prediction. This research is supported by Innosuisse (#43087.1 IP-LS).

Investigator: Jonathan Rafael Patino Lopez (CHUV), Richard Mospan (CHUV), Jonas Richiardi (CHUV/UNIL)

Collaborators: Richard McKinley (UNIBE, Inselspital), Roland Wiest (UNIBE, Inselspital), Bénédicte Maréchal (Siemens), Tobias Kober (Siemens), Diego Zeiter (Inselspital/Siemens), Elda Fischi (CHUV), Hannes Gubler (EPFL), Adrien Jammot (CHUV/INSA), Patric Michel (CHUV), Guillaume Saliou (CHUV), Alexander Salerno (CHUV), Silvia Pistocchi (CHUV)

Integrative subtyping of heart failure

Description: Heart failure with preserved ejection fraction (HFpEF) represents around 50% of heart failure cases, and has poor prognosis. One of the reasons for this failure may relate to the existence of distinct HFpEF subtypes that all require customized treatment, although these subtypes have not been conclusively identified. The goal of this project is to develop develop machine learning methods using a combination of cardiac MRI sequences (qMRI, cine…), genomics, transcriptomics, and metabolomics data to derive new biomarkers and patient subtypes with differentiated survival. This is part of a CHUV-HUG project funded by an SNSF Sinergia grant (#202276) which also comprises image acquisition developments, patient recruitment, and wet lab analyses.

Investigator: Costa Georgantas (CHUV), Jaume Banus Cobo (CHUV/UNIL), Jonas Richiardi (CHUV)

Collaborator: Ruud van Heeswijk (CHUV), Roger Hullin (CHUV), Philippe Meyer (HUG), Augustin Ogier (CHUV), Pauline Calarnou (CHUV), Angela Rocca (CHUV), Tamila Abdurashidova (CHUV), Jean-Paul Vallée (HUG), Jean-François Deux (HUG), Anam Fatima (UNIGE), Lindsey Crowe (HUG), Léo Gribinski (CHUV), Zoltan Kutalik (UNIL)

Multi-institution computable biomedical imaging platform

Description: The biomedical imaging research community needs collaborative and secure access to shared medical imaging data, a goal that CIBM, led by its founding institutions—CHUV, UNIL, EPFL, UNIGE, and HUG— is aiming to achieve. By consolidating efforts across these institutions, we aim to securely connect imaging data sources across multiple sites to a common inter-institutional platform, by developing efficient open-source data pipelines that transform clinical data into research-ready, depersonalised formats with consistent coding across imaging and clinical data. Together with consistent inter-institution guidelines for data governance, this will enable secure and easy sharing of imaging data. The platform will be connected to large-scale storage and downstream compute capabilities, enabling scalable image processing and machine learning workloads.

Investigator: Jonas Richiardi (CHUV), Sébastien Courvoisier (UNIGE), Solange Zoergiebel (CHUV), Marc Daverat (HUG), Dimitri Van De Ville (EPFL/UNIGE), Axel Couturier (EPFL)

Collaborators: Jean-Paul Vallée (HUG), Romain Breches (HUG), Nicolas Roduit (HUG), Shanoir Consortium (INRIA)