Tissue Doppler imaging optical flow (TDIOF): a combined B-mode and tissue Doppler approach for cardiac motion estimation in echocardiographic images.

Abstract

The quantitative analysis of cardiac motion from echocardiographic images helps clinicians in the diagnosis and therapy of patients suffering from heart disease. Quantitative analysis is usually based on tissue Doppler imaging (TDI) or speckle tracking. These methods are based on two techniques which to a large degree are independent: the Doppler phenomenon and image sequence processing. Herein, to increase the accuracy of the speckle tracking technique and to cope with the angle dependence of TDI, a combined approach dubbed tissue Doppler imaging optical flow (TDIOF) is proposed. TDIOF is formulated based on the combination of B-mode and Doppler energy terms minimized using algebraic equations and is validated on simulated images, and in vivo data. It was observed that the additional Doppler term is able to increase the accuracy of speckle tracking, compared to two popular motion estimation and speckle tracking techniques (Horn-Schunck and block matching methods). This observation was more pronounced when noise was present. The magnitude and angular error for TDIOF applied to simulated images, when comparing estimated motion with ground-truth motion, were 15% and 9.2°/frame, respectively. As an additional validation, echocardiography-derived strains were compared to tagged MRI-derived myocardial strains in the same subjects. The correlation coefficient (r) between the TDIOF-derived radial strains and tagged MRI-derived radial strains value was 0.83 (P < 0.001). The correlation coefficient ( r) for the TDIOF-derived circumferential strains compared to the tagged MRI-derived circumferential strains was 0.86 (P < 0.001). The comparison of TDIOF-derived and block matching speckle tracking and Horn-Schunck optical flow strain values using student t-test demonstrated superiority of TDIOF (95% confidence interval, P < 0.001).