Flow Based Interpretation of Dynamic Contrast Imaging Data

The measurement of perfusion and filtration are important clinical parameters used in diagnosis, follow-up, and therapy. The aim of this project is to investigate a novel approach for the interpretation of dynamic medical imaging with emphasis on blood distribution and flow. Typical applications include characterisation of strokes and planning and evaluation of cancer treatment. Medical image acquisition techniques like computerised tomography (CT), magnetic resonance imaging (MRI), or positron emission tomography (PET), can all be applied in a dynamic setting where the evolving distribution of an injected contrast agent is gathered together as a temporal sequence of images. Quantitative tissue characterisation (e.g. blood perfusion) from such data is currently performed locally by applying tracer-kinetic methodology to a single region of interest (ROI) or a voxel at a time. The tracer concentration for an individual voxel will be considered as a flow problem. By modelling the flow from first principles and calibrate the models to observations via systematic assimilation techniques, our goal is to advance understanding and clinical utility of dynamic imaging interpretation. In addition, this research aims to produce knowledge and technology that contributes to ICT solutions for enhancing productivity and efficiency within the health sector. To constuct the geometry of the model, we have been working with techiques to detect blood vessels from MR images. To this goal, we have been in contract with a research group in Jena that has developed a MR-sequence that is better suited to segmentation of blood vessels than those from our project partners. We have solved the problem of joining disconnected segmented parts of blood vessels. On the perfusion side, we have been working with several different models. One of the key issues is how to treat the transport of fluid from the vessel terminals to the neighbouring voxels in a way that fluid mass is conserved and numerical simulation is stable. We have at the moment at least two models, a simplified one (which is less accurate but computationally less demanding) and several more complex models (which are more accurate but significantly more expensive from the computational point of view). We have set up a methodology for parameter estimation and performed preliminary tests on the frog tongue (2D) and section of the brain (3D). Testing and validating is ongoing. The work with validation of the models shows realistic results. For the simplified model in 2D, we are able to produce blood pressure data that correspond to experimental results in literature (within measurement uncertainty) for the frog tongue, for a variety of radii. Our simulations are less accurate when the pixel size is in the same order of magnitude as the diameters of the blood vessels, at the border transition to the capillary tissue. This indicates that our model needs to be made even more accurate in this region. We look at this in the two models (both the simplified one and the one that is more accurate), where one "favours" flow in a similar was as a shortcut between arteries and veins. Preliminary results provide better agreement with experimental results from MRI with contrast, and these new models are now being tested as a reference to compare and validate methods of perfusion (such as "maximum slope") used clinically by the partner in Sheffield (UK), where the PhD student has been in a stay abroad for 3 months. Since March 2020, the work has been delayed, among others due to the closure of society in Norway, Europe and the world as a result of the covid-19 pandemic. As a result, conferences have been canceled, including conferences where contributions were accepted. Other conferences have been moved forward in time, while some have been virtual. This has led to a delay in our research and publishing activity, as well as a delay in the doctoral program for the PhD student. In the last project period, we have worked to further refine the computational perfusion models, in particular how the blood is transferred to the continuous tissue from the endpoints of the representable blood vessels. We have also worked with simulation of blood occlusions in vessels (infarct simulations). These simulations are very promising in as they show tissues/organs that might be damaged and how the system autoregulates, if so. The simulations might also become an important planning tool for clinical interventions, like how better plan a vessel by-pass in the occurrence of a stroke or infarction.

- Økt tverrfaglig og internasjonalt samarbeid (med miljø i Jena og Sheffield) - Økt samarbeid med bedrifter (NordicNeuroLab) - Utviklet metoder som har klinisk potensial i behandling av infarkt. Vi er i kontakt med medisinske miljø ved Haukeland Universitetssykehus for å se på muligheter for et fremtidig fellesprosjekt.

The measurement of perfusion and filtration are important clinical parameters used in diagnosis, follow-up, and therapy. By utilising complementary and well documented research skills, the aim is to investigate a novel approach towards interpretation of dynamic medical imaging with emphasis on blood distribution and flow. Typical applications include characterisation of strokes and planning and evaluation of cancer treatment. Medical image acquisition techniques like computerised tomography (CT), magnetic resonance imaging (MRI), or positron emission tomography (PET), can all be applied in a dynamic setting where the evolving distribution of an injected contrast agent is gathered together as a temporal sequence of images. Quantitative tissue characterisation (e.g. blood perfusion) from such data is currently performed locally by applying tracer-kinetic methodology to a single region of interest (ROI) or a voxel at a time. This project will formulate and investigate an alternative interpretation strategy where the tracer concentration for an individual voxel will be considered in the context of a global flow problem that connects all voxels in the image domain. By modelling the flow between voxels from first principles and calibrate the models to observations via systematic assimilation techniques that include rigorous error estimates, our goal is to advance understanding and clinical utility of dynamic imaging interpretation. Towards this end, we have assembled a team that includes established expertise in flow modelling, medical imaging, and data assimilation. In addition, this research aims to produce knowledge and technology that contributes to ICT solutions for enhancing productivity and efficiency within the health sector.