Motion Estimation in Static Magnetic Resonance Elastography

Abstract

Elastography is the imaging of the biomechanical properties of a tissue to detect and diagnose abnormal pathologies in a variety of disease conditions. Static Magnetic Resonance Elastography (MRE) is a modality of elastography that uses Magnetic Resonance Imaging (MRI) principles for data acquisition from a biological sample under external loading. An estimation of the mechanical deformation of the loaded sample from its Magnetic Resonance (MR) images constitutes a key component of the static MRE. Efforts in this area of research have mainly been focused on developing data acquisition protocols and motion estimation algorithms for producing high quality elastography images. So far, however, progress made in static MRE remains limited in both clinical and experimental fields. This dissertation work performed a comprehensive investigation of the data acquisition, pre-processing, and motion analysis stages of the static MRE modality. First, a mechanical device was introduced to reliably apply repetitive external compression to the sample. The design of this device and how it was interfaced with the scanner for gated data acquisition are described in detail. Next, MRI basics are summarized, and the use of tagged MRI sequence as the data acquisition protocol is justified. Optimal parameters that led to the best quality tagged MRI data were determined by taking the repetitiveness of the compression and the use of tag lines into consideration. Lastly, two reliable motion estimation algorithms were implemented and successfully tested on a variety of synthetic and real MRE data. After adjusting the parameters of the techniques using the prior knowledge of the features of the tagged MR images, both Iterative and One-step Optical Flow (OF) algorithms consistently produced acceptable results. It was found, that while applied to the real data, the Iterative OF algorithm slightly outperforms the One-step OF algorithm. The results of the testing are provided and discussed. This research is interdisciplinary and embraces concepts from the fields of Physics, Image Processing, Computer Vision, Algorithmics, Electrical Engineering, and Biomedical sciences. Future extensions of the research include a variety of studies on phantoms with an inclusion, small oncology animal models, and possibly followed by clinical human research that would contribute to improving the reliability, accuracy, and speed of tumor detection. Other possible applications may involve processing of different types of MRI data, such as cardiac tagged gated MRI.