Comparison of geometric regularization methods for 2D myocardial strain estimation in the mouse

Myocardial strain quantification in the mouse based on 2D speckle tracking using real time ultrasound data sets is feasible but still challenging. This is due to the low frame rate (about 300 Hz) in comparison with the high heart rate (500-700 bpm) which causes significant de-correlation between successive frames. The aim of this work was twofold: first, to analyze which motion estimator performs best in the context of real time cardiac murine data sets and second, to study the effect of different geometric regularizers on the initial motion assessment using a simulated mouse heart. More stable strain estimates were obtained making use of normalized cross-correlation followed by spatially limiting the curvature of the motion field.