M.Sc Thesis

M.Sc StudentZagury Nir
SubjectQuantification of Regional Radial Myocardial Strain by
Wavelet Denoising
DepartmentDepartment of Biomedical Engineering
Full Thesis textFull thesis text - English Version


The ability to quantify regional cardiac deformation automatically and across the different segments of the left ventricle (LV) plays a crucial role in diagnosis of myocardial dys-function and therapeutic planning. Automatic measurement of the radial strain, in particular, has the potential to define the regional myocardial function, where any visual interpretation is subjective, experience dependent, and causes high inter-observer and intra-observer variability. 

The proposed approach for radial strain measurement can be divided into three main stages: Imaging, Tissue tracking and a Post-Processing stage. Imaging of the LV was done by an ultrasound imaging system in order to acquire a sequence of B-mode images during the cardiac cycle, at three short axis planes. Tissue tracking was performed by following the displacement of each "tracking point" within the image. Local displacements and velocities of the myocardial tissue were measured by tracking the displacement of speckles during time. The last stage, the Post-Processing stage, included calculation of the radial strain, after implementation of several signal processing techniques, including wavelet de-noising.

Validation of the sub-stages was done by using Matlab simulations, Field II simulations and Hydro-gel phantoms. The proposed technique provided accurate measurement of the radial strain, in the different simulations and phantoms. Specific components of the wavelet coefficients, that encode simulated area with small stretching, were founded at the different resolution levels. The hydro-gel phantom that included a stiff area, which mimics small ischemic area, demonstrated that it was possible to measure more than half of the stiff area, in each frame. 

Clinical data acquired from a group of healthy human subjects, was analyzed. The results were compared to the results obtained by employing a commercial system (Echo-PAC, GE Healthcare Inc.). Radial strain estimates showed a good agreement with the results obtained by the more established commercial system. The measured radial strain at the different segments, which was measured by the commercial system, showed high similarity between the different segments, while the proposed method revealed the physiological differences among the various segments. The similarity obtained by the commercial system indicates that a robust smoothing procedure has been used. Additional comparison was performed between the proposed automatic technique and manual measurements, which showed high correlations, and thus validated the results of the proposed method.