Model-Based Estimation of Intracardiac Blood Flow Velocity Patterns Based on Ultrasound Imaging
MetadataShow full item record
There is a general concession within the field of cardiology that deeper knowledge of the complex intracardiac blood flow patterns would likely lead to improved methods for diagnosing cardiovascular diseases and deficiencies. It has already been shown that two-dimensional blood velocity estimates is of great clinical use for the diagnosis of congenital heart defects. However, methods currently in development suffer from substantial measurement noise and intrinsic artefacts, requiring both spatiotemporal filtering and manual intervention to yield useful velocity estimates. This thesis presents and evaluates possible benefits of using a model-based Unscented Kalman filter approach based on the Navier-Stokes equations for incompressible fluid flow and a more accurate measurement model to offer adaptive filtering and fusion of Blood Speckle Tracking (BST) and Colour Flow Imaging (CFI) velocity estimates with automatic Doppler aliasing correction. The fundamental model prediction and sensor fusion capabilities of the method was first proved using a Computational Fluid Dynamics (CFD) simulation of a simplified model of a neonatal Left Ventricle (LV) as ground truth, and a simple tube flow simulation with simplified additive measurement noise. Further, the filter performance was evaluated by filtering BST and CFI estimates from simulated ultrasound (US) images of the simulated LV, and comparing the resulting velocity estimates with the CFD simulation velocities. Lastly, as a proof-of-concept, velocity estimates from an actual US recording of a neonate was filtered using the presented method, and compared to the results of a conventional spatiotemporal filtering approach. The results show that the more accurate nonlinear measurement model with a modified estimator variance measure significantly improves the fusion of BST and CFI velocity estimates, by handling and correcting for known artefacts; but the nonlinear blood flow model is not robust enough to provide the overall filtering method with improved velocity estimates. It is therefore concluded, that further development of the blood flow model is necessary before the modified Unscented Kalman filter could be of use for multidimensional intracardiac blood velocity estimation, to aid in more accurate diagnosis of cardiovascular diseases.