Microbubble Tracking with a Nonlinear Motion Model

The reconstruction of the microvasculature and the estimation of its blood flow from contrast-enhanced ultrasound images is very promising for tumor characterization and therapy control. The application to the clinic, e.g. to ultrasound sequences of breast tumors, implies the challenge of low frame rates that are typical for clinical ultrasound devices. It has been shown that the vessel reconstruction can be performed by tracking microbubbles in consecutive frames. In this work, the Markov chain Monte Carlo data association algorithm is used which has proven to be a reliable tracking algorithm, also for those low frame rates. The algorithm is currently based on a linear motion model. However, depending on the tortuosity, vessels consist of curved courses with varying curvature radii. To better match these vessels, a nonlinear motion model is introduced. To handle the model's nonlinearity, an unscented Kalman filter is implemented. Simulations of different vessel scenarios show that the nonlinear motion model leads to a more robust tracking of MBs flowing along tortuous vessels that are acquired with low frame rates. For a lower degree of tortuosity, the linear and the nonlinear motion model perform similar.