Xunchang Chen

Strain rate imaging using two-dimensional speckle tracking

Strain rate images (SRI) of the beating heart have been proposed to identify non-contracting regions of myocardium. Initial attempts used spatial derivatives of tissue velocity (Doppler) signals. Here, an alternate method is proposed based on two-dimensional phase-sensitive speckle tracking applied to very high frame rate, real-time images. This processing can produce high resolution maps of the time derivative of the strain magnitude (i.e., square root of the strain intensity). Such images complement traditional tissue velocity images (TVI), providing a more complete description of cardiac mechanics. To test the proposed approach, SRI were both simulated and measured on a thick-walled, cylindrical, tissue-equivalent phantom modeling cardiac deformations. Real-time ultrasound images were captured during periodic phantom deformation, where the period was matched to the data capture rate of a commercial scanner mimicking high frame rate imaging of the heart. Simulation results show that SRI with spatial resolution between 1 and 2 mm are possible with an array system operating at 5 MHz. Moreover, these images are virtually free of angle-dependent artifacts present in TVI and simple strain rate maps derived from these images. Measured results clearly show that phantom regions of low deformation, which are difficult to identify on tissue velocity-derived SRI, are readily apparent with SRI generated from two-dimensional phase-sensitive speckle tracking.

3-D Correlation-Based Speckle Tracking

Widely-used 1-D/2-D speckle tracking techniques in elasticity imaging often experience significant speckle decorrelation in applications involving large elevational motion (i.e., out of plane motion). The problem is more pronounced for cardiac strain rate imaging (SRI) since it is very difficult to confine cardiac motion to a single image plane. Here, we present a 3-D correlation-based speckle tracking algorithm. Conceptually, 3-D speckle tracking is just an extension of 2-D phase-sensitive correlation-based speckle tracking. However, due to its high computational cost, optimization schemes, such as dynamic programming, decimation and two-path processing, are introduced to reduce the computational burden. To evaluate the proposed approach, a 3-D bar phantom under uniaxial compression was simulated for benchmark tests. A more sophisticated 3-D simulation of the left ventricle of the heart was also made to test the applicability of 3-D speckle tracking in cardiac SRI. Results from both simulations clearly demonstrated the feasibility of 3-D correlation-based speckle tracking. With the ability to follow 3-D speckle in 3-D space, 3-D speckle tracking outperforms lower-dimensional speckle tracking by minimizing decorrelation caused by pure elevational translation. In other words, 3-D tracking can push toward solely deformation-limited, decorrelation-optimized speckle tracking. Hardware implementation of the proposed 3-D speckle tracking algorithm using field programmable gate arrays (FPGA) is also discussed.

Lateral speckle tracking using synthetic lateral phase

In traditional speckle tracking, lateral displacement (perpendicular to the beam direction) estimates are much less accurate than axial ones (along the beam direction). The accuracy of lateral tracking is very important whenever spatial derivatives of both axial and lateral displacements are required to give a full description of a two-dimensional (2-D) strain field. A number of methods have been proposed to improve lateral tracking by increasing the sampling rate in the lateral direction. We propose an alternate method using synthetic lateral phase (SLP). The algorithm, a direct analog of the phase zero-crossing approach used in axial displacement estimation, synthesizes the lateral phase first, then performs a zero-crossing detection on this synthetic phase to obtain lateral displacement estimates. The SLP is available by simply eliminating either the positive or negative half of the lateral spectrum of the original analytic signal. No new data need to be acquired for this procedure. This new algorithm was tested on both simulations and measurements from a cardiac phantom model. Results show that the method greatly improves the accuracy of lateral tracking, especially for low strain cases (/spl les/1%). The standard deviation of the estimation error of the lateral normal strain obtained with this approach has an approximate factor of 2-3 improvement for low strain cases. The conceptual and computational simplicity of this new method makes it a practical approach to improve lateral tracking for elasticity imaging.

Exploiting strain-hardening of tissue to increase contrast in elasticity imaging

Most biological tissues show strain hardening. For example, direct mechanical measurements on human prostate show up to a threefold increase in Youngs modulus over a 10% deformation. In addition, strain-hardening behavior differs significantly between tissue types. In conventional elasticity imaging these effects produce strain dependent elastic contrast. Furthermore, image quality is suboptimal because softer tissues are imaged at higher strains than stiffer tissues. By applying linear processing on many small subsets over a large total deformation, strain-hardening behavior was measured for an agar-gel based phantom. Such processing can address strain-hardening problems in conventional elasticity imaging, and produce images of the elastic nonlinearity itself. Finally preliminary results for human prostate are presented.

Microwave-induced thermal imaging of tissue dielectric properties

A new imaging method, microwave-induced thermal imaging (MITI), was developed to differentiate tissue based on thermal and dielectric properties. Image contrast depends on temporal strain in tissue, which was determined by one-dimensional speckle tracking using a phase-sensitive, correlation-based technique. The underlying mechanisms were analyzed and experimental results on biologic tissue agreed well with theoretical predictions. Because of its strong contrast between water-bearing and lipid-bearing tissue, the technique may enhance existing intravascular ultrasound (IVUS) imaging systems to identify vulnerable arterial plaque.

Evaluation of 2-D speckle tracking based strain rate imaging (SRI) using a 3-D heart simulation model

Preliminary 2D SRI results from a porcine heart model showed significant average residual accumulation error (RAE) in the total displacement, similar to widely reported baseline drift artifacts in SRI. Therefore, to understand the performance of 2D SRI methods in a controlled 3D environment, a left-ventricle simulation model with physiological deformation was developed. Torsion was added to move tissue in and out of the image plane. A phased-array imaging system was simulated to generate 2D image sequences for typical cardiac scans. Results show the presence of RAE, which reaches a minimum at a frame rate (FR) of 75 frame/cycle corresponding to 2-3% peak frame-to-frame strain. RAE increases as torsion increases. In addition. the RAE map also spatially correlates well with both out-of-plane motion and cross-correlation coefficient maps. These results suggest that out-of-plane motion plays a significant role in RAE. To optimize SNR in displacement estimates a retrospective processing strategy employing different frame intervals for different periods in a cardiac cycle is presented.

Clinical application and technical challenges for intracardiac ultrasound imaging catheter based ICE imaging with EP mapping

A 9F combination intracardiac imaging and electrophysiology mapping catheter has been developed and tested to help guide diagnostic and therapeutic intracardiac electrophysiology procedures. A 7.5 MHz, 64 element, side looking phased array was used for sector scanning from the tip of the catheter. Multiple electrophysiology (EP) mapping sensors were mounted as ring electrodes just proximal to the array for electrocardiographic synchronization of ultrasound images. The catheter has been used in vivo in a porcine animal model and has demonstrated useful intracardiac echocardiographic (ICE) visualization of both cardiac tissue and electrophysiology catheters in the right atrium. The catheter performed well in high frame rate imaging, color flow imaging, and strain rate imaging of atrial and ventricular structures.

Temporal and spatial registration for cardiac strain rate imaging

Current strain rate imaging (SRI) suffers from Doppler principle limitations. Here we present an alternative 2D correlation-based speckle tracking method for SRI. To fully characterize local deformation over a cardiac cycle with Lagrangian strain estimates, in addition to spatial registration by referencing displacement measurements back to its original geometry, we perform temporal registration to compensate for different sampling times between beams. This algorithm was tested on both simulations and measurements from a porcine cardiac model. Simulations show strain rate errors of 30% of maximum without time registration. Strain rate with temporal and spatial registration show good agreement with theoretical predictions. In addition, the tracking algorithm is robust, producing almost no residual error in displacements accumulated over one complete cardiac cycle. Preliminary results from a porcine interventricular septum also show high repeatability in both longitudinal displacement and strain rate estimates between cardiac cycles.

Application of ultrasonic thermal imaging in IVUS systems

Sudden disruption of vulnerable coronary plaque is considered by pathologists as the most frequent cause of acute coronary syndromes. These plaques are characterized by a large lipid-rich core with abundant inflammatory cells and a thin fibrous cap, which can be potentially detected with thermal strain imaging (TSI) using IVUS. A phantom experiment using an IVUS array demonstrates the concept and results agree reasonably well with predictions. The in vivo application of this technique faces the major challenge of tissue motion. We propose a practical imaging scheme to minimize mechanical strains caused by tissue motion based on a linear least squares fitting strategy. To test this scheme, thermal strains were artificially superimposed on computed mechanical strain images from clinical data, and results suggest a 1-2/spl deg/C temperature rise is possible for in vivo plaque detection.