Real-time multi-model tracking of myocardium in echocardiography using robust information fusion

Automatic myocardial wall motion tracking in ultrasound images is an important step in analysis of the heart function. Existing methods for Myocardial Wall Tracking are not robust to artifacts induced by signal dropout, significant appearance or gain control changes. We present a unified framework for tracking the myocardium wall motion in real time with uncertainty handling and robust information fusion. Our method is robust in two aspects, firstly robust information fusion is used for combining matching results from multiple appearance models and secondly fusion is performed in the shape space to combine information from measurement and prior knowledge and models. Our approach fully exploits uncertainties from the measurement, shape priors, motion dynamics, and matching process based on multiple appearance models. Experiments illustrate the advantages of our approach validating the theory and showing the potential of very accurate wall motion measurements.