Periodic Reporting for period 1 - MultiPRESS (Multiscale Imaging of Cardiovascular Pressure Gradients – a Paradigm Shift in Hemodynamic Risk Prediction)
Période du rapport: 2023-05-01 au 2025-10-31
Regional quantification of cardiovascular pressure gradients is critical for diagnosis, treatment planning, and risk prediction of many cardiovascular diseases. Still, for a large number of conditions, non-invasive assessment is obstructed by inherent method limitations, and a wide range of cardiovascular instances exist where regional pressure behaviour remains unexplored. The MultiPRESS project main objective is to develop a novel imaging paradigm for non-invasive assessment of cardiovascular pressure gradients, overcoming critical limitations of existing techniques through a unique multiscale approach. Doing so, the MultiPRESS project seeks to – for the first time – extend non-invasive hemodynamic risk prediction into previously inaccessible cardiovascular domains, advance our knowledge of complex hemodynamic behaviour, and tackle remaining urgent clinical challenges across the heart, aorta, and brain.
Using deep integration of advanced full-field magnetic resonance imaging (4D Flow MRI), super-resolution networks, and physics-informed image processing, a set of core developments will allow for unique, comprehensive image-based pressure gradient assessment across (1) spatial (big/small vessels), (2) temporal (fast/slow flows), and (3) flow (laminar/turbulent) scales, with developments consistently validated in dedicated in-silico, in-vitro, and in-vivo cohorts. These developments will then be utilized on a set of core applications across (4) cardiovascular scales (heart/aorta/brain), addressing urgent clinical challenges and extending image-based pressure gradient quantification through previously inaccessible domains. Based in a unique multidisciplinary setting at Scandinavia’s largest university hospital, successful delivery of MultiPRESS will represent a paradigm shift in clinical hemodynamic risk prediction, and pave way for new scientific knowledge revitalizing risk stratification of complex cardiovascular disease across the heart, aorta, and brain.
Using deep integration of advanced full-field magnetic resonance imaging (4D Flow MRI), super-resolution networks, and physics-informed image processing, a set of core developments will allow for unique, comprehensive image-based pressure gradient assessment across (1) spatial (big/small vessels), (2) temporal (fast/slow flows), and (3) flow (laminar/turbulent) scales, with developments consistently validated in dedicated in-silico, in-vitro, and in-vivo cohorts. These developments will then be utilized on a set of core applications across (4) cardiovascular scales (heart/aorta/brain), addressing urgent clinical challenges and extending image-based pressure gradient quantification through previously inaccessible domains. Based in a unique multidisciplinary setting at Scandinavia’s largest university hospital, successful delivery of MultiPRESS will represent a paradigm shift in clinical hemodynamic risk prediction, and pave way for new scientific knowledge revitalizing risk stratification of complex cardiovascular disease across the heart, aorta, and brain.
The MultiPRESS objectives were divided into two core sections: one devoted to core developments (CDs), and one devoted to subsequent core applications (CAs). During the first two years of the grant, focus has been on the CDs, developing novel utilities for rapid image acquisitions, efficient super-resolution networks, and coupled physics-informed image analysis. More specifically:
• Establishing the use of residual ensemble learning to extend super-resolution 4D Flow MRI across arbitrary vascular sections, with a focus on spatial super-resolution (CD.1)
• Extending the idea of residual super-resolution into the temporal domain (CD.2) establishing a first approach in the temporal super-resolution 4D Flow MRI domain
• Mapping out both the potentials and challenges of contemporary generative networks for super-resolution imaging, clarifying where these benefit clinical use (in resolving boundary flows) as well as when challenges overtake benefit (e.g. non-convergence; hallucinations, etc.)
• Implementing the use of sparse convolutions to enable simultaneous spatiotemporal super-resolution; a concept previously limited by the computational curse of dimensionality when moving to higher-order dimensions
• Developing physics-driven imaging algorithms for estimating relative pressure through turbulent flow fields (CD.3) validating their performance in-silico
While CA will commence primarily in the second half of the project, some achievements can be listed as:
• Approved ethical applications for all clinical acquisitions (heart, aorta, and brain; CA1.-3.)
• Validation of imaging approach for super-resolution relative pressures in narrow intracranial vessels (CA.1)
• Prospective imaging trial for aortic dissection established with planned study start in October 2025 (CA.3)
• Establishing the use of residual ensemble learning to extend super-resolution 4D Flow MRI across arbitrary vascular sections, with a focus on spatial super-resolution (CD.1)
• Extending the idea of residual super-resolution into the temporal domain (CD.2) establishing a first approach in the temporal super-resolution 4D Flow MRI domain
• Mapping out both the potentials and challenges of contemporary generative networks for super-resolution imaging, clarifying where these benefit clinical use (in resolving boundary flows) as well as when challenges overtake benefit (e.g. non-convergence; hallucinations, etc.)
• Implementing the use of sparse convolutions to enable simultaneous spatiotemporal super-resolution; a concept previously limited by the computational curse of dimensionality when moving to higher-order dimensions
• Developing physics-driven imaging algorithms for estimating relative pressure through turbulent flow fields (CD.3) validating their performance in-silico
While CA will commence primarily in the second half of the project, some achievements can be listed as:
• Approved ethical applications for all clinical acquisitions (heart, aorta, and brain; CA1.-3.)
• Validation of imaging approach for super-resolution relative pressures in narrow intracranial vessels (CA.1)
• Prospective imaging trial for aortic dissection established with planned study start in October 2025 (CA.3)
The technical impact of the project lies in pushing the use of 4D Flow MRI and non-invasive hemodynamic risk mapping into previously inaccessible domains. Here, the development and expansion of super-resolution 4D Flow MRI-specific networks achieved this, together with physics-driven approaches quantifying output. Further uptake and success will be contingent on clinical usability, to be attempted through in-line implementations on clinical systems.