Date on Master's Thesis/Doctoral Dissertation


Document Type

Doctoral Dissertation

Degree Name

Ph. D.


Electrical and Computer Engineering

Committee Chair

Amini, Amir A.

Author's Keywords

3D SinMed; 3D CSPAMM; MRI; image processing; cardiac deformations


Heart--Diseases--Diagnosis; Diagnostic imaging; Magnetic resonance imaging; Imaging systems in medicine


Many cardiac diseases, such as myocardial ischemia, secondary to coronary artery disease, may be identified and localized through the analysis of cardiac deformations. Early efforts for quantifying ventricular wall motion used surgical implantation and tracking of radiopaque markers with X-ray imaging in canine hearts [1]. Such techniques are invasive and affect the regional motion pattern of the ventricular wall during the marker tracking process and, clearly are not feasible clinically. Noninvasive imaging techniques are vital and have been widely applied to the clinic. MRI is a noninvasive imaging technique with the capability to monitor and assess the progression of cardiovascular diseases (CVD) so that effective procedures for the care and treatment of patients can be developed by physicians and researchers. It is capable of providing 3D analysis of global and regional cardiac function with great accuracy and reproducibility. In the past few years, numerous efforts have been devoted to cardiac motion recovery and deformation analysis from MR imaging sequences. In order to assess cardiac function, there are two categories of indices that are used: global and regional indices. Global indices include ejection fraction, cavity volume, and myocardial mass [2]. They are important indices for cardiac disease diagnosis. However, these global indices are not specific for regional analysis. A quantitative assessment of regional parameters may prove beneficial for the diagnosis of disease and evaluation of severity and the quantification of treatment [3]. Local measures, such as wall deformation and strain in all regions of the heart, can provide objective regional quantification of ventricular wall function and relate to the location and extent of ischemic injury. This dissertation is concerned with the development of novel MR imaging techniques and image postprocessing algorithms to analyze left ventricular deformations. A novel pulse sequence, termed Orthogonal CSPAMM (OCSPAMM), has been proposed which results in the same acquisition time as SPAMM for 2D deformation estimation while keeping the main advantages of CSPAMM [4,5]: i.e., maintaining tag contrast through-out the ECG cycle. Different from CSPAMM, in OCSPAMM the second tagging pulse orientation is rotated 90 degrees relative to the first one so that motion information can be obtained simultaneously in two directions. This reduces the acquisition time by a factor of two as compared to the traditional CSPAMM, in which two separate imaging sequences are applied per acquisition. With the application of OCSPAMM, the effect of tag fading encountered in SPAMM tagging due to Tl relaxation is mitigated and tag deformations can be visualized for the entire cardiac cycle, including diastolic phases. A multilevel B-spline fitting method (MBS) has been proposed which incorporates phase-based displacement information for accurate calculation of 2D motion and strain from tagged MRI [6, 7]. The proposed method combines the advantages of continuity and smoothness of MBS, and makes use of phase information derived from tagged MR images. Compared to previous 2D B-spline-based deformation analysis methods, MBS has the following advantages: 1) It can simultaneously achieve a smooth deformation while accurately approximating the given data set; 2) Computationally, it is very fast; and 3) It can produce more accurate deformation results. Since the tag intersections (intersections between two tag lines) can be extracted accurately and are more or less distributed evenly over the myocardium, MBS has proven effective for 2D cardiac motion tracking. To derive phase-based displacements, 2D HARP and SinMod analysis techniques [8,9] were employed. By producing virtual tags from HARP /SinMod and calculating intersections of virtual tag lines, more data points are obtained. In the reference frame, virtual tag lines are the isoparametric curves of an undeformed 2D B-spline model. In subsequent frames, the locations of intersections of virtual tag lines over the myocardium are updated with phase-based displacement. The advantage of the technique is that in acquiring denser myocardial displacements, it uses both real and virtual tag line intersections. It is fast and more accurate than 2D HARP and SinMod tracking. A novel 3D sine wave modeling (3D SinMod) approach for automatic analysis of 3D cardiac deformations has been proposed [10]. An accelerated 3D complementary spatial modulation of magnetization (CSPAMM) tagging technique [11] was used to acquire complete 3D+t tagged MR data sets of the whole heart (3 dynamic CSPAMM tagged MRI volume with tags in different orientations), in-vivo, in 54 heart beats and within 3 breath-holds. In 3D SinMod, the intensity distribution around each pixel is modeled as a cosine wave front. The principle behind 3D SinMod tracking is that both phase and frequency for each voxel are determined directly from the frequency analysis and the displacement is calculated from the quotient of phase difference and local frequency. The deformation fields clearly demonstrate longitudinal shortening during systole. The contraction of the LV base towards the apex as well as the torsional motion between basal and apical slices is clearly observable from the displacements. 3D SinMod can automatically process the image data to derive measures of motion, deformations, and strains between consecutive pair of tagged volumes in 17 seconds. Therefore, comprehensive 4D imaging and postprocessing for determination of ventricular function is now possible in under 10 minutes. For validation of 3D SinMod, 7 3D+t CSPAMM data sets of healthy subjects have been processed. Comparison of mid-wall contour deformations and circumferential shortening results by 3D SinMod showed good agreement with those by 3D HARP. Tag lines tracked by the proposed technique were also compared with manually delineated ones. The average errors calculated for the systolic phase of the cardiac cycles were in the sub-pixel range.