Document Type

Dissertation

Date of Degree

Fall 2016

Degree Name

PhD (Doctor of Philosophy)

Degree In

Biomedical Engineering

First Advisor

Suresh M. L. Raghavan

Abstract

Cerebral aneurysm is a pathology of the circulatory system in the brain in which an arterial wall balloons into a blood-filled sac. If the aneurysm ruptures, stroke can occur and has a high probability of causing permanent disability or death. Aneurysm surgery carries a high rate of morbidity and mortality compared to the natural rate of aneurysm rupture, so physicians must take care in recommending surgery for an aneurysm patient. However, very little is known about the etiology of brain aneurysm rupture and what prognostics exist. The International Study for Intracranial Aneurysms suggested that large aneurysm size and posterior location are important factors in identifying high rupture risk. However, many small aneurysms and aneurysms in other portions of the circulation still rupture. Many studies have assessed morphological traits, identified from aneurysm appearance on diagnostic medical images, and found such traits to be different in aneurysms that ruptured and aneurysms that did not rupture. In fact, more than 50 such morphological indices have been introduced in the literature, and many of them redundantly quantify particular morphological characteristics. In order to demonstrate the prognostic ability of morphology as an indicator of rupture risk, however, a large longitudinal cohort study must be carried out. A study such as this is time-consuming and expensive, and each additional hypothesis that a particular morphological index is predictive of rupture risk would require increasing the study population size in order to fulfill the necessary statistical power requirements for a rigorous test. Thus, a minimal set of physically meaningful, independent metrics that fully describe the aneurysm morphology is needed.

In this dissertation an automated protocol was developed to process segmented medical images and extract an exhaustive set of morphological indices that quantify all relevant morphological features. Each morphological index was then analyzed for robustness to inter-user variability and for sensitivity to the particular morphological characteristic that it was designed to quantify. A factor analysis was then performed using the most robust, sensitive metrics on a population of unruptured aneurysms from five data centers and 276 patient-specific aneurysms. The results from the factor analysis were utilized to ascertain what morphological features those metrics truly described, if there were any redundancies in definition, and the variance each morphological trait described in the population. Four underlying morphological constructs were uncovered through the factor analysis. The first factor, sac size, was highly aligned with morphological indices that measured volume and one-dimensional size measurements. Sac size described 50% of the variance in the data set. The second factor, sac irregularity, was highly aligned with morphological indices that described various aspects of irregular shape. A set of variables that all were implicated in causing irregular shape, but in reality measured sac-neck size relation, also merited inclusion of a second metric to describe the variance seen in the second factor. Sac irregularity described 20% of the variance in the data set. The third factor, sac ellipticity, aligned highly with morphological indices that described the overarching ellipticity of the aneurysm sac independent of other non-spherical characteristics. Sac ellipticity described 13% of the variance in the data set. The fourth factor, sac-vessel size relation, aligned highly with morphological indices that described the size of the aneurysm sac in relation to its parent vessel. Sac-vessel size relation described 7% of the variance in the data set. All four of these factors in combination described 91% of the variance in the data set. Five morphological indices – non-planar isolation sac volume (Vnp), Voronoi diagram core evolution irregularity index (IRRvdc), tissue stretch ratio (TSR), Voronoi diagram core evolution ellipticity index (EIvdc) and size ratio (SRang) were determined to be the key indices for describing aneurysm morphology. Finally, the proposed metrics were used to test the hypothesis that aneurysms that are chosen for untreated observation are morphologically different than those that are treated – commonly referred to as selection bias. Study population was 27 patient-specific aneurysms that were placed on untreated observation (observation group) and 27 patient specific aneurysms that were size- and location-matched but were chosen for treatment (treated group). A significant difference was found in the morphological index that measured ellipticity between the two groups, indicating that physicians already commonly select highly elliptical aneurysms for treatment. This result may give insight into physicians’ choices, and merits further investigation with a larger data set for confirmation. Additionally, because the same result was replicated in both of the metrics chosen to quantify ellipticity (for both manual and automated methods), this highlighted the use of the morphological factors in determining an minimal set of independent, robust morphological indices.

Keywords

Cerebral aneurysm, Computational geometry, Morphology, Rupture

Pages

xiv, 131

Bibliography

129-131

Comments

This thesis has been optimized for improved web viewing. If you require the original version, contact the University Archives at the University of Iowa: http://www.lib.uiowa.edu/sc/contact/

Copyright

Copyright © 2016 Benjamin Micah Berkowitz

Share

COinS