M.Sc Thesis

M.Sc StudentLempel Meytal
Subject3D Speckle Tracking in Ultrasound Data
DepartmentDepartment of Biomedical Engineering
DR. Zvi Friedman
Full Thesis textFull thesis text - English Version


Assessment of left ventricular (LV) regional function is done today mainly by echocardiography, due to its widespread bedside availability, cost and non-invasive nature. Currently the analysis is generally limited to one or two dimensional (1D and 2D) scans.  However, 3D processing is more accurate and allows better use in follow-up treatment of patients than 1D or 2D processing.  In the last few years, significant improvements were made in the technology of 3D ultrasound scans, which potentially provide better information for studying and developing methods of tracking in 3D.

This study demonstrates the feasibility of assessing the myocardial LV regional motion using in-vivo dynamic 3D ultrasound acquisitions.  An algorithm was developed, for tracking motion of the LV muscle using dynamic 3D (and 2D) ultrasound acquisitions. The method was validated on a 2D phantom and 2D and 3D clinical data, using established statistical tools.

The method combines correlation-based speckle tracking with feature tracking.  The selected features are bright spots, patterns surrounding peak brightness levels in the acquisition.  The growth of the processing complexity of the algorithm was found not to be proportional to the data dimensionality, which is a great advantage over reported tracking methods.  The tracking results allow the calculation of the myocardial displacement field, from which deformation parameters, such as strain, are derived. 

The method allows measuring the persistence of bright spots.  On average, these reflectors persisted three times longer in 3D scans than in 2D scans, emphasizing the increased stability of 3D patterns over 2D patterns. 

Performances of the method on 2D data corresponded well with the processing results of a commercial product, 2D-Strain (by GE Healthcare). 

3D motions that were measured using 3D acquisitions were compared with 2D motions that were measured using 2D acquisitions of the same patient and found physiologically reasonable.  This comparison is inherently limited by the differences in characteristics of 2D and 3D scans, mainly the much lower acquisition rate of 3D scans in comparison to 2D scans (about three times lower), and of course, the extra information in 3D scans.  In the frame work of this research we could not perform a comparison to other methods in 3D.

The implementation of the 3D-Strain method in C performs fast, allowing real-time analysis, thus maintaining one of the main advantages of echocardiography over other imaging modalities.