Title: Multi-dimensional quantification of myocardial deformation using echocardiographic imaging: development, implementation and validation
Other Titles: Multi-dimensionele quantificatie van de vervorming van de hartspier met behulp van echocardiografie: ontwikkeling, implementatie en validatie
Authors: Langeland, Stian; M0217381
Issue Date: 8-Feb-2007
Abstract: Introduction.Echocardiography is currently the most important non-invasive imaging modality in clinical cardiology. The reasons for its popularity are many.Compared to other medical imaging modalities like magnetic resonance imaging (MRI), computer tomography (CT), and nuclear imaging the ultrasound equipment is portable, easy to use, and readily available. Because of this it is currently the only bed-side imaging modality used in cardiology today.A normal clinical echo exam is performed by visually inspecting the gray-scale image and performing measurements of volumes and wall thicknesses. In addition, Doppler measurements of the blood-flow are performed in order to assess the hemodynamic performance of the heart. Most of these parameters are derived from global measurements and therefore describe the heart as a whole. However, several cardiac diseases, like ischemia orinfarction, are regional and limited to certain segments of the myocardium. Until now these lesions have been diagnosed by visual assessment ofwall motion and deformation.In recent years, strain and strain-rate imaging (SRI) have emerged as promising tools to localize and diagnose these regional myocardial diseases. Strain and strain-rate parameters describe the regional myocardial deformation as quantitative parameters. The quantitative nature of this method has the potential to reduce the inter/intra observer variability currently present in the diagnosis of regional myocardial diseaseUltrasound Based Strain Estimation.Strain is a measure of deformation and the simplest form of strain is the one-dimensional strain, which can be expressed as the change in length of a bar with respect to its initial shapeS(t) = (l(t) - l0)(l0),where S(t) is strain, l(t) is the instantaneous length at time t, and l0 is the length in the undeformed state.The current deformation imaging methodology is based on color Doppler imaging and it therefore inherits the fundamental limitations of this method. As color Doppler is only able to estimate the velocity component along the ultrasound beam, the derived deformation estimation is also limited to the component along the beam.As there are a limited number of acoustic windows through the human thorax, this imposes difficulties in assessing the different strain components. The current methodology is angle dependent, which makes the clinical interpretation more difficult and requires a high level of operator expertise. Moreover, this approach implies that only one component of the true 3D deformation of a myocardial segment is measured within one acquisition, thus limiting the information available on the myocardial deformation. Finally, to obtain reliable deformation estimates, it is necessary at present to manually track a myocardial segment throughout thecardiac cycle. This makes the technique rather time-consuming and therefore less clinically applicable.The angle dependency of the current deformation imaging methodology originates from the one dimensional nature of the velocity estimates. The solution to these problems therefore lies in finding the velocity vector rather than the velocity along the ultrasound beam. This can be obtainedby tracking patterns in the ultrasound image between consecutive frames. For each pixel in the image, angle-independent velocity estimation is performed by selecting a search pattern around that pixel in one frame and looking for a matching pattern in a search region around that pixel of the following frame . The search pattern is placed at different positions in the search region, and the similarity is measured for the overlapping area. The position where the highest similarity is found determinesthe in-plane frame-to-frame displacement relative to its initial position.Aims and Outline.The aim of this work was to develop, implement and validate algorithms that will enable the estimation of the multi-dimensional measurement of cardiac deformation. This will not only allow for the assessment of the complex deformation of the heart but will also make the technique angle independent. The former advantage will contribute to a better understanding of cardiac patho-physiology while the latter will make the techniqueless expert-dependent as the angle problem is eliminated.The work can be sub-divided into three parts:1. Development of algorithms for the estimation of multidimensional deformation of elastic objects, based on simulated data sets.2. Validation of these algorithms by measuring deformation properties of in-vitro deformable phantoms and in-vivo, both with ultrasound and an independent gold standard.3. Implementation of the optimal approach towards multi-dimensional deformation measurements in the currently used clinical research environment.Methods and results.Comparison of different similarity measurements.One of the problems to obtain robust angle independent strain estimatesis the fact that velocity estimates perpendicular to the ultrasound beam (azimuthal) are intrinsically noisier than the ones parallel to the beam (axial) . The aim of this study was to find the optimal similarity measure for the tracking of the radio frequency (RF) patterns, both in axial and azimuthal direction.Performances of the following similarity measures were investigated using simulations: cross correlation, normalized cross correlation, sum of absolute differences and sum of squared differences.Two-dimensional velocity estimation from cardiac ultrasound RF signals was not feasible using cross correlation. However, normalized cross correlation, sum of absolute differences and sum of squared differences showed accurate axial and lateral results. For smaller window lengths, sum of squared differences was found to be the preferred similarity measure for two-dimensional velocity estimation using a one-dimensional kernel.Finding the optimal processing parameters.2D velocity estimation is typically based on 2D displacement tracking of spatial features in the gray-scale (GS) or radio-frequency (RF) data using either 1D or 2D kernels. Different approaches have specific advantages and disadvantages but a trade-off between accuracy and computationalcost generally has to be made. Although some studies have been presented in the literature on the effect of these parameters, these have been limited to simulated or tissue mimicking phantom setups.The aim of this study was therefore to evaluate the influence of different estimation parameters on the strain estimate in an in-vivo setting in order to enable an optimization of computation time.B-mode RF data sets were acquired from the inferolateral wall of 4 openchest sheep. 2D displacements were calculated in each data set, based on a sum of squared differences estimator and using different algorithm parameters. Myocardial radial and longitudinal strain were simultaneouslyestimated in the inferolateral wall using this new methodology, based on 2D motion gradients. Three segment-length sonomicrometry crystals gavea continuous reference for the longitudinal and radial strain.No statistically significant differences were found for the different kernel sizes in neither axial or azimuth direction. Similarly, neither sampling frequency nor GS vs. RF tracking had a significant influence on the mean error. However, for all parameters longitudinal strain errors were significantly smaller than the radial ones (p < 0.01).It has been shown in the literature that, under ideal conditions, the variance in the strain estimator is highly affected by parameters as window size and sampling-frequency. These results did not reproduce in the in-vivo setting indicting that other sources of noise are more dominant. Short, 1D, kernels might thus be advantageous in 2D strain estimation since they give a similar accuracy at a reduced computational cost.Validation in tissue mimicking phantoms.The aim of this study was to validate this methodology in a phantom setup. A tubular thick-walled tissue mimicking phantom was fixed in a watertank. Varying the intraluminal pressure resulted in cyclic radialdeformation. The 2D strain was calculated from the 2D velocity estimates, obtained from 2D radio-frequency (RF) tracking, using a 1D kernel. Additionally, ultrasonic micro crystals were implanted on the outer and inner wall of the tube in order to give an independent measurement of the instantaneous wall thickness. The two methods were compared by means of linear regression, the correlation coefficient and Bland-Altman statistics.As expected, the strain estimates dominated by the azimuth velocity component were less accurate than the ones dominated by the axial velocity component. Correlation coefficients were found to be r=0.78 for the former estimates and r = 0.83 was found for the latter. Given that the overall shape and timing of the 2D deformation were very accurate (r=0.95 andr=0.84), these results were within acceptable limits for clinical applications.Validation in-vivo.The aim of this study was to validate this new methodology in an in-vivo setting using sonomicrometry for comparison. In 5 open chest sheep, ultrasound data were acquired in a parasternal long axis view. Using the new methodology, simultaneous measurements of radial and longitudinal strain were performed in the inferolateral wall from single data sets. Segment-length sonomicrometry crystals were used as the reference. After baseline acquisitions, deformation was modulated by pharmacologically changing the inotropic state of the myocardium and by inducing ischemia. The ultrasonically estimated peak systolic radial and longitudinal strain components were validated against sonomicrometry by means of the intra class correlation coefficient and Bland-Altman analysis.For both strain components, good agreements were found between the ultrasound and the sonomicrometry measurements as shown by Bland-Altman statistics. For the radial strain estimates, the intra class correlation coefficient was found to be ICCC = 0.72, while for the longitudinal component, it was ICCC = 0.80.Application to TEE imaging.The aim of this study was to investigate the possibility of measuring angle independent two-dimensional strain during TEE imaging.Nine consecutive patients, referred for a clinical TEE examination, were included in the study. Transgastric short and long axis TEE images were acquired followed by transthoracic parasternal long and short axis images.From the Bland-Altman analysis, the limits of agreement were found to be 3.3+-9.3% and 0.3+-12.1% for the radial and longitudinal strain respectively. Linear regression gave a correlation coefficient of r=0.74 for the radial data while no significant correlation was found for the longitudinal data.Analysis of 2D strain was found to be feasible for TEE imaging. This method enables automatic tracking of the region of interest. Unfortunately, the assessment of longitudinal deformation was found to be not accurate enough for current clinical use.
Publication status: published
KU Leuven publication type: TH
Appears in Collections:Cardiovascular Imaging and Dynamics

Files in This Item:

There are no files associated with this item.

Request a copy


All items in Lirias are protected by copyright, with all rights reserved.