Non-Rigid Image Registration Based Strain Estimator for Intravascular Ultrasound Elastography

Abstract

Intravascular ultrasound elastography (IVUSe) could improve the diagnosis of cardiovascular disease by revealing vulnerable plaques through their mechanical tissue properties. To improve the performance of IVUSe, we developed and implemented a non-rigid image-registration method to visualize the radial and circumferential component of strain within vascular tissues. We evaluated the algorithm's performance with four initialization schemes using simulated and experimentally acquired ultrasound images. Applying the registration method to radio-frequency (RF) echo frames improved the accuracy of displacements compared to when B-mode images were employed. However, strain elastograms measured from RF echo frames produce erroneous results when both the zero-initialization method and the mesh-refinement scheme were employed. For most strain levels, the cross-correlation-initialization method produced the best performance. The simulation study predicted that elastograms obtained from vessels with average strains in the range of 3%–5% should have high elastographic signal-to-noise ratio (SNRe)–on the order of 4.5 and 7.5 for the radial and circumferential components of strain, respectively. The preliminary in vivo validation study (phantom and an atherosclerotic rabbit) demonstrated that the non-rigid registration method could produce useful radial and circumferential strain elastograms under realistic physiologic conditions. The results of this investigation were sufficiently encouraging to warrant a more comprehensive in vivo validation.