Patient specific cardiac flow simulation by high performance computational fluid dynamics

SNIC 2017/1-80


SNAC Medium

Principal Investigator:

Matilda Larsson


Kungliga Tekniska högskolan

Start Date:


End Date:


Primary Classification:

20603: Medicinsk bildbehandling



In cardiovascular diagnostics, knowledge regarding general blood flow patterns and flow motions in the heart is central, since abnormalities in these are often directly related to underlying pathological changes of the myocardium or the circulatory system. Recently, anatomically accurate in-silico models in combination with complex computational fluid dynamics (CFD) solvers have enabled the study of regional blood flow behavior in-detail, and the correlation of such changes to specific patient pathologies. So far though, a vast majority of all models have been either based on generic geometries or on data originating from high-spatial resolution modalities such as MRI or CT. With such modalities unable to capture temporally transient flows, a need has been identified to create a similar flow simulation framework from ultrasound measurements [1] – a modality being available at a majority of all clinical care institutions in the world. At the Computational Technologies Laboratory (CTL) at KTH, a framework for high-performance computing (HPC) for the solution of complex systems of differential equations (such as generally present within fluid dynamical, or fluid-structure-interaction based systems) has been developed, with focus so far being put on the development of parallel adaptive algorithms within finite element modelling (FEM) [2] (the software generally used has so far been the HPC branch of the open source FEM library DOLFIN and the adaptive flow solver Unicorn. Both libraries have been parallelized using a hybrid MPI+OpenMP approach, with solvers having been successfully used to efficiently solve large scale industrial problems). Within this framework, a platform for cardiac intraventricular flow simulation has been implemented. In collaborations with the Unit for Medical Imaging at KTH, this platform has recently been extended to enable for patient-specific models of the left ventricle to be generated and evaluated with respect to blood flow behavior [3]. The work performed is though as yet at an initial stage, with the patient-specific platform being validated only in the form of a feasibility study. Therefore, a thorough robustness evaluation and validation is for 2015 planned on a larger patient-data set, to verify the usability of the model. With such performed and understanding the applicability of the model, a number of clinical as well as numerical research questions will be addressed in the following time period, starting already in 2015. For clinical use this includes relating myocardial motion disruptions to flow behavior (including flow stagnation, intraventricular thrombolic developments, atrial fibrillation, myocardial infarction etc.), whereas numerical issues may include the ability of adding more complex geometrical features (in the form of detailed valvular structures), or enabling for an extended model including fluid-structure interaction behavior (in the form of e.g. flow-myocardium communication). [1] Borazjani I. et al, Comp Math Models in Medicine, art.ID: 395081, 2013 [2] Hoffman J. et al, Computer and Fluids, 80:339-357, 2013 [3] Larsson D. et al, Abstract to be presented at IEEE International Ultrasonics Symposium in Taipei, Taiwan, 2015