Contenu et objectifs du travail de recherche
Basé sur des observations cliniques, ce projet part du postulat qu’il existe plusieurs sous-types d’HFpEF distincts, chacun nécessitant une approche thérapeutique différente.
Nous enrichirons considérablement les données disponibles pour caractériser et grouper les patients HFpEF. A cette fin, nous développerons des nouvelles techniques d’imagerie par résonance magnétique (IRM) permettant d’étudier la structure et la fonction du cœur de manière extrêmement précise. Nous recueillerons également des données génétiques et métaboliques détaillées. Nous développerons finalement de nouvelles techniques de modélisation par apprentissage automatique, qui permettront de combiner toutes ces informations, d’identifier des sous-groupes de patients, et d’étudier leur évolution clinique différenciée dans le temps.
Contexte scientifique et social du projet de recherche
Ce projet multidisciplinaire va mettre à disposition des techniques d’IRM plus précises permettant leur application chez le patient cardiologique. Crucialement, la distinction des sous-groupes différents parmi des patients HFpEF fournira la base pour des approches thérapeutiques au bénéfice des patients.
Background - Heart failure with preserved ejection fraction (HFpEF) is an increasingly prevalent and fatal condition that is estimated to affect 2% of the general population. Risk factors and aspects of the pathophysiology associated with HFpEF have been identified in recent years, but no pharmacological strategy that interacts with either HFpEF-related risk factors or detected pathological mechanisms has improved cardiovascular outcome in these patients. One of the reasons for this failure may well relate to the existence of distinct HFpEF subtypes that all require customized treatment, although these subtypes have not been conclusively identified. Heterogeneity of the study population may therefore explain the lackluster results in many clinical trials. Even worse, pharmacological studies on the basis of currently identified subtypes also resulted in neutral outcomes, implying that our present mostly clinics-based comprehension of HFpEF does not provide the basis necessary for successful therapeutic strategies. Therefore, new strategies that go beyond clinical information to identify HFpEF subtypes need to be studied, and we here propose an integrative approach in order to assemble the most complete information on HFpEF to date. Ideally, this approach should integrate results from modalities that dive deep into the anatomy, genetics, and metabolism and that are based on cutting-edge technology not yet available for clinical routine. Quantitative magnetic resonance imaging (qMRI) is particularly suited for the anatomy part, since qMRI can be used to map cardiac structure, function, hemodynamics, and tissue composition with high resolution and reproducibility. Together with data from genomic and metabolomic analyses, such rich imaging data are the ideal substrate for applied machine learning (ML). ML has recently been used to revolutionize various domains of health care based on its unbiased approach to data. The application of ML-based strategies to this unique wealth of multi-modal data from HFpEF patients is therefore very timely, may lead to a breakthrough in the understanding of the HFpEF pathophysiology, and will in parallel advance the cutting-edge technologies of qMRI and ML by themselves. Goal - We aim to perform the first unbiased and in-depth subtyping of HFpEF patients and the underlying pathophysiology with respect to specific clinical outcomes, which has the groundbreaking potential to pinpoint targets of therapeutic intervention. To this end, we will create the largest HFpEF databank containing the richest collection of clinical, genomic, metabolomic, and quantitative MRI data to date, and develop novel qMRI and machine learning techniques tailored to HFpEF. Methods - Our deliberate research plan brings together a multidisciplinary team with documented expertise in heart failure, MRI, machine learning, genomics, and metabolomics, and employs state-of-the-art equipment. In a first Work Package, high-resolution 3D quantitative MRI for the precise characterization of the heart will be designed, simulated, and optimized in vitro, in healthy volunteers, and in HFpEF patients. Simultaneously, in the second Work Package we will develop new ML algorithms that will learn what the subtypes of HFpEF are from multi-modal data, first with the large existing UK Biobank, then with our own more expansive data as it is acquired. Regular feedback between research groups and specific, targeted clinical outcomes will ensure the relevance and applicability of our analyses. The combination of these new MRI and machine learning techniques will be applied in the third Work Package, where we will perform an observational clinical study that includes clinical workup, quantitative MRI, genomics, and metabolomics in 550 patients and 50 healthy volunteers. The ML-based multi-modal clustering of this data will then identify the subtypes and their significantly different characteristics, which will be linked to subtype-specific clinical outcomes, pathophysiological mechanisms, and possible therapies.Anticipated impact - The unprecedented resolution and precision of the new cardiac qMRI acquisitions coupled with a uniquely rich dataset with genetic and metabolic dimensions will thoroughly profile the HFpEF subtypes as well as their specific characteristics and outcomes. This project will thus pave the way for a paradigm shift in the comprehension and specification of HFpEF and will furthermore identify potential targets of therapeutic intervention. Simultaneously, this project opens up fundamentally new application of qMRI and ML for research and discovery in clinical medicine.
Last updated:01.03.2022
