Abstract

Magnetic Resonance Imaging (MRI) is an imaging modality that acquires an image with little to no damage to the tissue. MRI does not introduce foreign particles or high energy radiation into the body, making it one of the least invasive medical imaging modalities. MRI can achieve excellent soft tissue contrast and is therefore useful for diagnosis of a wide variety of diseases. While there are a wide variety of available techniques for generating contrast in MRI, there are still many open areas for research. For example, many tissues in the human body exhibit such rapid signal decay that they are difficult to image with MRI: they are "MRI invisible". Furthermore, some of the newer MRI imaging techniques have not been fully validated to ensure that they are truly revealing accurate information about the underlying anatomical microstructure that they purport to image. This dissertation focuses on the development of new techniques in two distinct areas. First, a novel method for accurately assessing the MRI signal decay properties of tissues that are normally MRI invisible, such as tendons, ligaments, and certain pathological chemical deposits in the brain, is presented. This is termed "ultrashort MRI relaxometry". Second, two new image processing algorithms that operate on high resolution images of stained histological slices of the ex vivo brain are presented. The first of these image processing algorithms allows the semi-automated extraction of nerve fiber directionality from the histological slice images, a process that is normally done manually, is incredibly time consuming, and is prone to human error. This new technique represents one significant step in the complicated problem of attempting to validate a popular MRI technique, Diffusion Tensor Imaging (DTI), by ensuring that DTI results correlate with the true underlying physiology revealed by histological slicing and staining. The second of these image processing algorithms attempts to extract and segment regions of different "cytoarchitectonic characteristics" from stained histological slices of ex vivo brain. Again, traditional cytoarchitectonic segmentation relies on manual segmentation by an expert neuroanatomist, which is slow and sometimes inconsistent. The new technique is a first step towards automated this process, potentially providing greater accuracy and repeatability of the segmentations in a much shorter time. Together, these contributions represent a significant contribution to the body of MR imaging techniques, and associated image processing techniques for validation of newer MR neuroimaging techniques against the gold standard of stained histological slices of ex vivo brain.

Degree

PhD

College and Department

Ira A. Fulton College of Engineering and Technology; Electrical and Computer Engineering

Rights

http://lib.byu.edu/about/copyright/