Improvement of myocardial displacement estimation using subkernels for cross correlation between ultrasonic RF echoes

Speckle tracking is a useful diagnostic method for assessing the cardiac function from ultrasonic echoes. Some factors influence the performance of the conventional speckle tracking method. Deformation due to myocardial contraction and relaxation decreases the correlation between RF echoes. Also, if a strong echo, such as echoes from the epicardium, is included in a correlation kernel, the displacement estimated from the kernel becomes close to the displacement of the epicardium even when the myocardium in the middle of the heart wall is the target to be tracked. To overcome such a problem, in the present study, we proposed a two-step tracking method. In the first step, the displacement is coarsely estimated using a correlation kernel of size similar to that in conventional speckle tracking of (13.8°, 9.1mm) in the lateral and axial directions. The kernel is divided into small subkernels (2.8°, 1.9mm). By estimating the correlation coefficient with respect to each subkernel and averaging correlation coefficients obtained from all subkernels, the contribution of a peculiar strong echo included in one of the subkernels can be suppressed. In the second step, the residual displacement is finely estimated using a smaller kernel (13.8°, 3.6 mm). Using a smaller kernel, the influence of myocardial deformation can be suppressed, but there might be multiple regions that have echo patterns similar to that in the kernel, leading to tracking errors. Therefore, the search region in the second step is narrowed because the coarse displacement has already been estimated and compensated in the first step. In the present study, the proposed method was validated using a phantom made of urethane rubber, which was deformed by an actuator. The displacement of the phantom could be estimated with a lower error of 0.059mm by the proposed method compared to the error in the conventional method of 0.097 mm. Furthermore, the proposed method was applied to in vivo measurement of the human heart. The myocardial displacement could be estimated successfully even when the epicardium was included in a correlation kernel. These results show that the proposed method provides a more accurate and robust estimation of the myocardial displacement.