An investigation of microstructural white matter changes in Alzheimer’s disease and healthy aging using diffusion tensor imaging

(1)

An Investigation of Microstructural White Matter Changes in Alzheimer’s Disease and Healthy Aging Using Diffusion Tensor Imaging

by

Chantel Dana Mayo

B.Sc. Honours, University of Winnipeg, 2013 A Thesis Submitted in Partial Fulfillment

of the Requirements for the Degree of MASTER OF SCIENCE in the Department of Psychology

(2)

Supervisory Committee

An Investigation of Microstructural White Matter Changes in Alzheimer’s Disease and Healthy Aging Using Diffusion Tensor Imaging

by

Chantel Dana Mayo

B.Sc. Honours, University of Winnipeg, 2013

Supervisory Committee

Jodie R. Gawryluk, Department of Psychology Supervisor

Mauricio A. Garcia-Barrera, Department of Psychology Departmental Member

(3)

Abstract

Supervisory Committee

Jodie R. Gawryluk, Department of Psychology Supervisor

Mauricio A. Garcia-Barrera, Department of Psychology Departmental Member

Background: Given that brain pathology precedes clinical symptoms in Alzheimer’s

disease (AD), identifying pre-symptomatic biomarkers is critical in order to implement symptom-delaying strategies as early as possible. Magnetic resonance imaging (MRI) is an ideal method for detecting early brain changes in Alzheimer's disease, as it is non-invasive, easily repeatable, and widely available. To date, MRI biomarker research has largely focused on neuronal loss in grey matter, but there is a lack of research on white matter and its relationship with cognitive performance. Diffusion tensor imaging (DTI) is a MRI-based technique that is particularly sensitive to microstructural white matter characteristics, making it an ideal method to study white matter changes. Methods: Longitudinal DTI and clinical data from the Alzheimer’s Disease Neuroimaging Initiative 2 database were used to examine the 1) within-group microstructural white matter

changes in individuals with AD and healthy aging controls at baseline and year one; 2) the between-group microstructural differences in individuals with AD and controls at both time points; and 3) the relationship between white matter and cognitive performance at both time points. Results: 1) Within-group: Tract-based Spatial Statistics reveal that individuals with AD have reduced fractional anisotropy (FA) and increased mean diffusivity (MD) in the corpus callosum; internal and external capsule; corona radiata; posterior thalamic radiations; superior and inferior longitudinal fasciculus;

(4)

fronto-occipital fasciculus; cingulate gyri; fornix; uncinate fasciculus; tapetum; medial lemniscus; cerebellar and cerebral peduncle; and hippocampal cingulum at year one compared to baseline. Controls also had reduced FA and increased MD at year one compared to baseline, but such changes were less extensive and did not include the hippocampal cingulum. 2) Between-group: Relative to controls, individuals with AD had lower FA and higher MD in the corpus callosum, internal and external capsule; corona radiata; posterior thalamic radiation; superior and inferior longitudinal fasciculus and fronto-occipital fasciculus; cingulate gyri; fornix; uncinate fasciculus; tapetum and hippocampal cingulum. 3) There was a positive relationship between FA and an ADNI-derived memory composite score in individuals with AD. Conclusion: The results revealed that DTI holds potential as an AD biomarker given its sensitivity to detect microstructural white matter characteristics. Longitudinal tracking of brain imaging and AD clinical signs in large cohorts are necessary to further evaluate potential clinical utility.

(5)

Within-group Microstructural White Matter Changes in AD ... 49 Between-group Differences in AD and Healthy Controls ... 51 The Relationship Between Microstructural White Matter and Cognitive Performance ... 53 Study Limitations ... 56 Conclusion ... 57 Chapter 3: Future Directions for Diffusion Tensor Imaging in Alzheimer’s Disease Using the ADNI Database ... 59 References ... 64 Appendix ... 74

(7)

List of Tables

Table 1. Participant Demographics and Clinical Scores at Baseline and Year One ... 28

Table 2. Regions Showing Reduced Fractional Anisotropy at Year One Compared to Baseline in i) Individuals with Alzheimer’s Disease and in ii) Healthy Controls (p <0.05, Corrected for Multiple Comparisons). ... 31

Table 3. Regions Showing Increased Mean Diffusivity at Year One Compared to Baseline in i) Individuals with Alzheimer’s Disease and in ii) Healthy Controls (p <0.05,

Corrected for Multiple Comparisons). ... 33

Table 4. Regions Showing Low Fractional Anisotropy in Individuals with Alzheimer’s Disease Compared to Healthy Controls at i) Baseline and ii) Year One (p <0.05,

Table 5. Regions Showing High Mean Diffusivity in Individuals with Alzheimer’s Disease Compared to Healthy Controls at i) Baseline and ii) Year One (p <0.05,

Table 6. Regions Where ADNI-MEM Composite Scores are Positively Associated with Fractional Anisotropy at i) Baseline and ii) Year One in Individuals with Alzheimer’s Disease (p <0.05, Corrected for Multiple Comparisons). ... 39

Table 7. Regions Where ADNI-MEM Composite Scores are Negatively Associated with Mean Diffusivity at i) Baseline and ii) Year One in Healthy Controls (p <0.05, Corrected for Multiple Comparisons). ... 41

Table 8. Regions Where ADNI-EF Composite Scores are Positively Associated with Fractional Anisotropy at i) Baseline and ii) Year One in Individuals with Alzheimer’s Disease (p <0.07, Corrected for Multiple Comparisons). ... 43

Table 9. Regions Where ADNI-EF Composite Scores are Positively Associated with Fractional Anisotropy at i) Baseline and ii) Year One in Healthy Controls (p <0.05, Corrected for Multiple Comparisons). ... 45

Table 10. Regions Where ADNI-EF Composite Scores are Negatively Associated with Mean Diffusivity at i) Baseline and ii) Year One in Healthy Controls (p <0.05, Corrected for Multiple Comparisons). ... 47

(8)

List of Figures

Figure 1. Alzheimer’s disease defined by the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association criteria (panel A); and AD as a neurocognitive disorder as defined by the Diagnostic and Statistical Manual of Mental Disorders 5th edition (panel B). ... 4 Figure 2. Eigenvalues (λ1, λ2, λ3) are equal (isotropic diffusion; panel A) and eigenvalues

are unequal (anisotropic diffusion; panel B). ... 12 Figure 3. Flow diagram of participant selection for individuals with Alzheimer’s disease and healthy controls. Final selection criteria included individuals with Alzheimer’s disease (AD) or healthy controls (CN) from the ADNI2 database with axial diffusion-weighted magnetic resonance imaging scans at baseline and year one. ... 27 Figure 4. Results of within-group Tract-Based Spatial Statistics white matter analysis showing pattern of reduced fractional anisotropy (red) at year one compared to baseline in individuals with Alzheimer’s disease (panel A) and in controls (panel B; p <0.05, corrected for multiple comparisons) ... 32 Figure 5. Results of within-group Tract-Based Spatial Statistics white matter analysis showing pattern of increased mean diffusivity (blue) at year one compared to baseline in individuals with Alzheimer’s disease (panel A) and in controls (panel B; p <0.05,

corrected for multiple comparisons) ... 34 Figure 6. Results of between-group baseline (panel A) and year one (panel B) Tract-Based Spatial Statistics white matter analysis showing pattern of lower fractional anisotropy (red) overlaid on the mean fractional anisotropy skeleton (green) in individuals with Alzheimer’s disease compared to controls (p < 0.05, corrected for

multiple comparisons) ... 36 Figure 7. Results of between-group baseline (panel A) and year one (panel B) Tract-Based Spatial Statistics white matter analysis showing pattern of mean diffusivity (blue) overlaid on the mean fractional anisotropy skeleton (green) in individuals with

Alzheimer’s disease compared to controls (p < 0.05, corrected for multiple comparisons) ... 38 Figure 8. Images showing regions where ADNI-MEM composite scores are positively associated with fractional anisotropy (red) at baseline (panel A) and year one (panel B) in individuals with Alzheimer’s disease (p <0.05, corrected for multiple comparisons) ... 40

(9)

Figure 9. Images showing regions where ADNI-MEM composite scores are negatively associated with mean diffusivity (blue) at baseline (panel A) and year one (panel B) in healthy controls (p <0.05, corrected for multiple comparisons) ... 42 Figure 10. Images showing regions where ADNI-EF composite scores are positively associated with fractional anisotropy (red) at baseline (panel A) and year one (panel B) in individuals with Alzheimer’s disease (p <0.07, corrected for multiple comparisons. ... 44 Figure 11. Images showing regions where ADNI-EF composite scores are positively associated with fractional anisotropy (red) at baseline (panel A) and year one (panel B) in healthy controls (p <0.05, corrected for multiple comparisons. ... 46 Figure 12. Images showing regions where ADNI-EF composite scores are negatively associated with mean diffusivity (blue) at baseline (panel A) and year one (panel B) in healthy controls (p <0.05, corrected for multiple comparisons) ... 48

(10)

Acknowledgments

I would like to express my appreciation to all who contributed to this work. In

particular, I would like to thank my supervisor, Dr. Jodie Gawryluk, for endless support and guidance over the past two years. In addition, I would like to thank my committee member, Dr. Mauricio Garcia-Barrera, and co-investigators, Drs. Erin Mazerolle, John Fisk, and Lesley Ritchie for thoughtful input and contributions. I am truly grateful to have such amazing mentors. I would also like to thank Mom, Dad, Carly, Brent, and my cohort members (Abbi, Drew, Kristen, Raquel, and Vivien) for continued encouragement and love along the way. Finally, I would like to express my gratitude to the Canadian Institutes of Health Research for funding this thesis.

Data collection and sharing for this project was funded by the Alzheimer's Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies;

(11)

Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for

Research and Education, and the study is coordinated by the Alzheimer's Disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.

(12)

Dedication

(13)

Chapter 1: Using Diffusion Tensor Imaging to Identify White

Matter Biomarkers in Alzheimer’s Disease

Alzheimer’s Disease

Alzheimer’s disease (AD) is the most common form of dementia. Approximately 747 000 Canadians are affected by AD or a related dementia and this number is expected to climb to 1.4 million Canadians by 2031 (Alzheimer’s Society of Canada, 2012). In the United States, 5.2 million Americans are affected by AD. The Alzheimer’s Association (2014) estimates that the incidence of AD will soon reach 1 million; this means one new case will develop every 33 seconds.

AD is a neurocognitive disorder that is characterized by progressive

neurodegeneration and cognitive decline (Alzheimer’s Association, 2014). Memory loss is often the initial and primary concern, but many other cognitive domains can be

affected (Alzheimer’s Association, 2014).

The traditional criteria for diagnosing AD was proposed in 1984 by the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA; McKhann et al., 1984). Using these criteria, an individual could be diagnosed with 1) Probable AD, 2) Possible AD, or 3) Definite AD (Figure 1A).

According to NINCDS-ADRDA criteria, probable AD is characterized by deficits in two or more areas of cognition; progressive worsening of memory and other cognitive functions; no disturbance of consciousness; onset between ages 40 to 90; and dementia

(14)

confirmed by neuropsychological tests. The diagnosis is often supported by evidence of a progressive deterioration of cognitive functioning; impaired activities of daily living; a positive family history of AD; as well as laboratory tests (normal lumbar puncture,

normal or non-specific electroencephalography changes, and evidence of cerebral atrophy using computerized tomography). Possible AD is characterized by the presence of a single, progressive cognitive deficit; the absence of other neurological or psychiatric disorders sufficient to cause dementia; and variations in the onset, presentation or clinical course. Finally, Definite AD is characterized by fulfilling the criteria for Probable AD, along with histopathological evidence obtained via biopsy or through post-mortem autopsy (McKhann et al., 1984).

Alternate, but related AD criteria are available in the Diagnostic and Statistical Manual of Mental Disorders 5th edition (DSM 5; American Psychiatric Association, 2013), often considered the “gold standard” for mental health diagnoses. In the DSM 5, AD is a neurocognitive disorder and is defined by the following criteria:

Evidence of significant cognitive decline from a previous level of performance in one or more cognitive domains … based on 1) concern of the individual, a

knowledgeable informant, or the clinician that there has been a significant decline in cognitive function; and 2) a substantial impairment in cognitive performance, preferably documented by neuropsychological testing or, in its absence, another quantified clinical assessment (American Psychiatric Association 2013, p. 602). Neurocognitive disorders may fall into a major or mild category depending on the

severity of the presenting signs symptoms (Figure 1B). In major neurocognitive disorder, cognitive impairments interfere with activities of daily living, but in minor

(15)

neurocognitive disorder, activities of daily livings are not disrupted. In both cases, the cognitive impairment is not better explained by delirium or by another mental disorder.

Neurocognitive disorders may be caused by a number of disorders or diseases (e.g., dementia, traumatic brain injury, substance use, Human Immunodeficiency Virus,

Parkinson’s disease, Huntington’s disease). Neurocognitive Disorder Due to AD is characterized by an insidious onset and gradual progression of cognitive impairment.

In Major Neurocognitive Disorder Due to Probable AD there is 1) evidence of a positive family history of AD through genetic testing; or 2a) evidence of memory and learning decline along with declines in a second cognitive domain (e.g., complex attention, executive function, language, perceptual motor, social cognition); 2b)

progressive decline in cognition; and 2c) and no evidence of other conditions that may be contributing to cognitive decline. In the absence of this evidence, a diagnosis of Major Neurocognitive Disorder Due to Possible AD is provided.

In Mild Neurocognitive Disorder Due to Probable AD, there is evidence of a positive family history of AD through genetic testing. In the absence of this evidence, a diagnosis of Mild Neurocognitive Disorder Due to Possible AD is given when there is also clear evidence of memory and learning, progressive decline in cognition, and no evidence of other conditions that may be contributing to cognitive decline (DSM 5; American Psychiatric Association, 2013).

(16)

Figure 1. Alzheimer’s disease defined by the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association criteria (panel A); and AD as a neurocognitive disorder as defined by the Diagnostic and Statistical Manual of Mental Disorders 5th edition (panel B).

In terms of neuropathology, neurofibrillary tangles and beta amyloid plaques are hallmark features of AD (Hampel et al., 2014). Abnormalities in tau protein within the cells lead to microtubule separation and abnormal tau aggregation, which disrupts the transport of nutrients and other essential molecules (Hardy & Higgins, 1992; Perl, 2010). Abnormal splicing of amyloid pre-cursor protein by an enzyme called beta secretase leads to the formation of beta amyloid plaques, which are toxic, and interfere with neuron-to-neuron communication at the synapse (Hardy & Higgins, 1992; Perl, 2010). Such changes in the brain lead to damage to the neurons and, eventually, cell death (Alzheimer’s Association, 2014; Hampel et al., 2014; Hardy & Higgins, 1992; Perl,

(17)

2010). The typical progression of this neurodegeneration was first characterized by Braak and Braak (1991) in a post-mortem study of 83 brains. The entorhinal cortex was among the earliest affect regions, followed by increasing limbic involvement (e.g., hippocampal and parahippocampal regions) increasing association cortex involvement, and finally, sensory-motor area involvement. Because individuals with AD often survive to the late stages of the disease, post-mortem brains usually have widespread pathology, which makes identification of clinical progression difficult (Perl, 2010).

Although there are neuropathological features associated with AD that can be detected in vivo, AD diagnosis is only confirmed following post-mortem observation of neurofibrillary tangles and beta amyloid plaques in the brain (Hampel et al., 2011; Sperling et al., 2011; Wurtman 2015). There are ongoing efforts to establish criteria to identify AD progression through clinical and laboratory assessment (e.g., Albert et al., 2011; Jack et al., 2011; McKhann et al., 2011; Sperling et al., 2011), but late clinical diagnosis is common among those living with AD. In practice, an AD diagnosis is typically provided by a clinical neuropsychologist after cognitive decline has already occurred. By the time of clinical diagnosis, an individual with AD may have widespread damage to neurons, cortical thinning, ventricular enlargement and demyelination (Gold, 2012).

Once a diagnosis is given, there is no cure for AD, and available treatment options are limited. Approved options for treatment of AD include cholinesterase inhibitors (e.g. Donepezil) and NMDA antagonists (i.e. Memantine), which are pharmacological

treatments aimed at reducing the rate of cognitive decline (Howard et al., 2012; Mayo Clinic, 2015; Robinson & Keating, 2006;). Other non-pharmacological therapies may be

(18)

used to maintain cognitive function through compensation strategies, but disease-modifying therapies are not yet available.

At present, there is emerging research focused on developing interventions to further delay disease progression. The development of an intervention that could delay the onset of AD by 5 years would reduce the number of individuals living with AD by approximately 57% (Sperling et al., 2011), saving $283 billion in American health care costs (for a detailed breakdown, see Alzheimer’s Association, 2015). Furthermore, there is evidence that early identification of an AD-trajectory and subsequent intervention would improve quality adjusted life years (e.g., Barnett, Lewis, Blackwell, & Taylor, 2014). Together, these data highlight the need to identify those individuals at risk of first developing AD, as well as indices of their disease progression. Ideally, effective

symptomatic treatments and preventative strategies would be introduced prior to the development of significant neurodegeneration (e.g., Jack et al., 2012; Vos et al., 2013; Petersen, 2013).

Biomarkers for Early Identification

There is increasing support that brain pathology precedes clinical symptoms, which has contributed to a surge of researchers seeking to identify pre-symptomatic change through the use of biomarkers (Sperling et al., 2011; Wurtman 2015). A biomarker is any measurable biological index of the presence of a disease or the risk of developing a disease in the future (Alzheimer’s Association, 2014). Ideally, a successful biomarker would 1) facilitate early identification and aid in diagnoses; 2) allow clinicians to monitor disease progression; 3) give clinicians a tool to monitor treatment response (Hampel et al., 2011; Hampel et al., 2014; Wurtman 2015). To date there is no consensus on which

(19)

biomarkers hold the greatest diagnostic utility (Dubois et al., 2014), but there are a number of biomarkers that are currently being investigated for AD, as outlined below (see Hampel et al., 2014 for a thorough review).

Cerebrospinal Fluid

Molecular changes in the brain are reflected in cerebrospinal fluid (CSF; Blennow & Zetterberg, 2009). Promising candidate CSF biomarkers include the 42 amino acid form of amyloid beta (Aβ42), which is thought to reflect the total deposition of amyloid (plaques) in the brain; total tau (t-tau), which is thought to reflect neurodegeneration; and phosphorylated tau (p-tau), which correlates with pathological neurofibrillary changes (i.e., tangle formation; Blennow, Hampel, Weiner & Zetterberg, 2010; Hampel et al., 2014; Sheltens et al., 2016). Unfortunately CSF is obtained through a lumbar puncture, which is often avoided due to its perceived invasiveness and risk of post-procedure complications (Blennow & Zetterberg, 2009).

Positron Emission Tomography

Positron Emission Tomography (PET) is also a widely used biomarker tool. In PET imaging, a molecular isotope (tracer) is injected; this tracer accumulates in brain regions that have an affinity for that molecule (Budinger & Van Brocklin, 2013). Pittsburgh compound B (PiB) is a recently developed tracer that is able to cross the blood brain barrier and bind to beta amyloid in the brain (Chetelat et al., 2010; Ikonomovic et al., 2008). Previous research has shown that individuals with AD have a greater retention of PiB on PET scans (Ikonomovic et al., 2008).

(20)

Structural Imaging

Traditionally, structural brain imaging (e.g., computed tomography (CT), magnetic resonance imaging (MRI)) has only been used in clinical settings to rule out other

medical conditions (e.g., hemorrhage, tumour, etc.; de Gois Vasconcelos et al., 2009). Although less precise than post-mortem histochemical analysis, neuroimaging allows for the observation of the brain in vivo (Amlien & Fjell, 2014). MRI, specifically, is an ideal biomarker tool. It is non-invasive, easily repeatable, and widely available in many hospital and research settings (de Gois Vasconcelos et al., 2009).

Magnetic Resonance Imaging

MRI is based on the principles of nuclear magnetic resonance (NMR; Brown, Cheng, Haacke, Thompson & Venkatesan, 2014). According to NMR theory, atomic nuclei possess spin. When exposed to an external magnetic field (B0), the hydrogen nuclei spin on their axis (or precess) and align parallel or anti-parallel to the direction of the magnetic field. Typically, more nuclei align parallel (low energy state) rather than anti-parallel (high energy state), to the direction of the magnetic field, which results in a longitudinal bulk magnetization. The rate of precession for each nuclei is determined by Larmor frequency (ω), which suggests that the rate of the precession is related to the strength of the magnetic field:

where γ is the gyromagnetic ratio, a constant value for each type of nuclei

(21)

protons in the low energy state will flip to the high energy state, reducing the longitudinal magnetization, and the protons begin to precess together, creating transverse

magnetization. Together, this causes the bulk magnetization to rotate 900 from the longitudinal plane to the transverse plane When the RF pulse is removed, some of the high energy protons will return to the low energy state (T1 or “Spin-Lattice” relaxation) and the protons that were precessing together begin to repel each other due to their positive charge (T2 or “Spin-Spin” relaxation), releasing energy that can be detected by the MRI’s receiving coil.

Different types of tissue have distinct T1 and T2 relaxation rates, allowing for the construction of contrasted grey-scale images. A T1 weighted image maximizes the differences in the longitudinal plane; CSF appears dark grey (long T1), grey matter appears light grey (medium T1), and white matter appears white (short T1).

In contrast, T2 weighted image maximize the differences in the transverse plane; CSF appears white (long T2), grey matter appears grey (short T2) and white matter appears dark grey (short T2; Brown et al., 2014).

Consistent with post-mortem findings, anatomical MRI findings have indicated widespread whole brain atrophy in individuals with AD, along with enlarged ventricles and decreased hippocampal volume (Perl, 2010; Teipel et al., 2010). In vivo estimations of pathological progression using brain imaging have found that the neurodegenerative process begins years before diagnosis. For example, in a longitudinal study of preclinical AD, Bernard and colleagues (2014) found that individuals who were later diagnosed with AD (5 years after baseline) had smaller left amygdalohippocampal volumes at baseline than individuals who remained healthy. Furthermore, these individuals with AD had a

(22)

higher rate of atrophy in the temporal and parietal cortices at each of the 4 time points examined. There is hope individuals on an AD trajectory can be identified much earlier in the disease progression (i.e., before symptoms are apparent). To date, the best established MRI biomarker for AD is decreased hippocampal volume (Gold, 2012; Hampel et al., 2014). Although this biomarker has been widely used in research settings for more than 20 years, only recently (2011) have attempts to standardize begun (Hampel et al., 2014). Unfortunately, such large-scale (macroscopic) measures are only indicative of the later stages of AD pathology. Therefore, these are likely insufficient measures for early detection of AD (de Gois Vasconcelos et al., 2009). The Amyloid Cascade Hypothesis (Hardy & & Higgins, 1992) predicts that axonal degeneration (white matter) precedes neuronal loss (grey matter), suggesting that white matter biomarkers may lead to earlier identification of AD. Despite this, there remains a paucity of research on white matter in AD relative to grey matter studies.

Using DTI to Assess Microstructural White Matter Diffusion in Biological Tissue

One promising tool to detect microstructural white matter in the brain is an MRI technique known as diffusion weighted MRI (Alexander et al., 2007; Mori & Zhang, 2009). Diffusion weighted MRI is based on water diffusion within the brain (Soares, Marques, Alves & Sousa, 2013). Diffusion of water occurs both within and between brain cells; these patterns of water diffusion vary in different types of tissue and are influenced by the presence of biological barriers (Gold, 2012; Mori & Zhang 2009; Soares et al., 2013). Water unconstrained by biological barriers diffuse equally, in all directions; this is known as random or isotropic diffusion. In contrast, biological barriers restrict water

(23)

movement in a perpendicular direction; this is known as anisotropic diffusion (Gold, 2012).

Diffusion Tensors

A diffusion tensor is a mathematic model of this water diffusion in 3-dimensional space derived from diffusion measurements obtained through 6 or more diffusion gradient directions (Jones & Leemans, 2011; Jellison et al., 2014). It can be represented numerically in a diagonally symmetric 3 by 3 covariance matrix (Alexander et al., 2007; Jellison et al., 2014). Diagonalization of this matrix generates three eigenvectors (ε1, ε2,

ε3,) and three corresponding eigenvalues (λ1, λ2, λ3). Eigenvectors represent the direction

of the maximum water diffusion, and eigenvalues represent the magnitude of the water diffusion for each vector (Jellison et al., 2014; Soares et al., 2013; Stebbins & Murphy, 2009).

It is typical to visualize the diffusion tensor as an ellipsoid shape (Jellison et al., 2014). In this case, the eigenvectors define the direction of the principle axes, while the eigenvalues define the radius of the ellipsoid (Alexander et al., 2007). With isotropic diffusion, the eigenvalues are approximately equal and the tensor approaches a spherical shape (Figure 2). With anisotropic diffusion, however, the eigenvalues are unequal, and the tensor becomes elliptical (i.e., deviates from the spherical shape; Alexander et al., 2007, Jellison et al., 2014).

Eigenvalues are influenced by changes in tissue microstructure (e.g., due to aging, brain trauma or disease). Thus, in modeling the diffusion weighted MRI data through this diffusion tensor imaging (DTI), researchers can detect in vivo microstructural changes in the brain that are not detected by conventional MRI (Alexander et al., 2007; de Gois

(24)

Vasconcelos et al., 2009; Gold, 2012; Jones & Leemans, 2011; Soares et al., 2013).

Figure 2. Eigenvalues (λ1, λ2, λ3) are equal (isotropic diffusion; panel A) and eigenvalues

are unequal (anisotropic diffusion; panel B). DTI Indices

There are a number of common DTI measures used to assess microstructural characteristics in the brain. Two of the most commonly reported DTI indices include fractional anisotropy (FA) and 2) mean diffusivity (MD):

FA. Measured on a scale from 0 (isotropic diffusion) to 1 (anisotropic diffusion),

FA is a measure of the degree of directionality of the water diffusion (Amlien & Fjell, 2014; de Gois Vasconcelos et al., 2009; Mori & Zhang, 2009; Soares et al., 2013; Stebbins & Murphy, 2009). FA is calculated using the following formula:

where λ1, λ2, λ3 represent the eigenvalues of the diffusion tensor, and λbar represent the

(25)

MD. Unlike FA, MD does not provide any information regarding the direction

associated with of the diffusion (Stebbins & Murphy, 2009). Instead, MD is a measure of the water diffusion rate (Soares et al., 2013). MD is calculated using the following

formula:

where λ1, λ2, λ3 represent the eigenvalues of the diffusion tensor

Water Diffusion in Healthy Brain Tissue

Water diffusion is generally isotropic in CSF and nearly so in grey matter, but is anisotropic in white matter (Alexander et al., 2007; Jones & Leemans, 2011). Because water diffusion is completely isotropic in CSF, it has very low FA. Although biological barriers such as cell membranes impede some of the water diffusion in grey matter, water diffusion is still largely isotropic, and thus, grey matter also has lower FA (e.g. FA = 0.3; Keller et al., 2013). In contrast to CSF and grey matter, white matter is highly organized into parallel fibre bundles. While water is free to diffuse along the direction of the axonal fibres, water diffusion perpendicular to the fibres is greatly restricted by tightly packed axons and the myelin surrounding the axons. Therefore, white matter water diffusion is highly anisotropic and has high FA, (de Gois Vasconcelos et al., 2009; Stebbins & Murphy, 2009; Zhang, Xu, & Kantarci, 2013), especially in regions with a large number of parallel fibres (e.g. corpus callosum, FA =0.8; Keller et al., 2013).

(26)

these changes occur during the healthy aging process. For example, Burzynska and colleagues (2010) examined DTI indices from 80 young adults (aged 20 to 32) and 63 older adults (aged 60 to 71). Tract-based spatial statistics (TBSS) showed an age-related reduction in FA within a number of white matter structures, including the corona radiata; the white matter of the superior, inferior, middle and frontal gyri; the white matter of the precuneus; the white matter of the superior parietal lobule; the dorsal cingulum; the fornix; the forceps minor and major; the internal and external capsule; the sagittal stratum; and the parahippocampal white matter.

Water Diffusion in Unhealthy White Matter

When pathological processes affect white matter in the brain, as is the case in AD, alterations in the water diffusion properties beyond normal aging are observed. Typically, decreases in FA occur along with increases in MD in white matter (e.g., Douaud et al., 2011; Shu, Wang, Qi, Li & He, 2011). High MD is indicative of high water diffusion in the brain; this may occur as a result of tissue breakdown as is the case in brain trauma or disease (Gold, 2012). Similarly, decreased FA is also indicative of progressive loss of barriers (e.g., myelin) that restrict water diffusion in healthy white matter. Several factors may influence FA value including myelination level and axon density (Bosch et al., 2012). As such, measures of FA are thought to provide an in vivo index of

microstructural integrity in white matter tissue. (Stebbins & Murphy, 2009). Although microstructural changes occur over the course of healthy aging, the regions affected in healthy and unhealthy aging are not the same. Damoiseaux and

colleagues (2009) used TBSS to compare healthy aging to AD. Results suggested that the white matter regions affected differ in healthy aging and AD. Specifically, lower FA was

(27)

primarily found in the frontal, parietal, and subcortical areas when healthy elderly were compared to healthy young (i.e., indicative of “healthy aging”). In contrast, lower FA was also observed in the left anterior temporal lobe when individuals with AD were compared to the healthy elderly. Therefore, to understand the white matter involvement unique to AD, comparisons to healthy age-matched controls is essential when determining regional differences in DTI indices.

Regional Differences in DTI Indices

There are a number of brain regions that have consistently shown altered FA and MD patterns within the growing body of literature on AD (see Appendix). In comparing a large number of DTI studies, Stebbins and Murphy (2009) found that low FA and high MD were commonly found in all lobes of the brain, including medial temporal lobe structures: the entorhinal cortex, parahippocampal white matter, and posterior cingulum – regions known to be involved in memory function (Stoub, deToledo, & Dickerson, 2014). Additionally, these patterns of white matter alterations were predominantly posterior (Stebbins & Murphy, 2009).

One of the first DTI meta-analyses found widespread alterations in both FA and MD in individuals with AD across 41 region-of-interest (ROI) studies (Sexton, Kalu, Filippini, MacKay, & Ebmeier, 2011). Specifically, low FA was reported in the white matter of frontal and temporal lobes; the genu and splenium of the corpus callosum; the anterior, middle, and posterior cingulum; the parahippocampal cingulum, the uncinate fasciculus; and the superior longitudinal fasciculus. Similarly, high MD was observed in the white matter of all lobes of the brain; the genu and splenium the corpus callosum; the posterior cingulum; the uncinate fasciculus; and the hippocampus. A more recent

(28)

meta-analysis of 13 whole-brain TBSS studies also reported low FA and high MD in AD (Acosta-Cabronero & Nestor 2014). When AD was compared to healthy aging controls, widespread white mater alterations were usually observed in the parietal, temporal and prefrontal white matter, including association fibres and inter-hemispheric connections via the corpus callosum.

The majority of DTI studies to date have employed cross-sectional designs. As such, there is a great need for repeated within and between-group longitudinal follow-up to identify meaningful observations for predicting conversion rates from healthy aging to AD (Hampel et al., 2011).

Longitudinal Change in DTI Indices

Few longitudinal studies have been published to date, and of the published studies, many have small sample sizes (Amlien & Fjell, 2014). In one longitudinal study,

Kitamura and colleagues (2013) scanned individuals with AD once at baseline and again approximately 1.5 year later. FA decreases were observed within the bilateral uncinate fasciculi. Genc and colleagues (2016) examined white matter changes in a small cohort of individuals with AD (N = 18) at 6-month follow-up compared to a baseline scan. In addition to significant FA reductions (9.6%) and MD increases (32.2%) in the whole white matter skeleton, there were also FA decreases and MD increases in the body and splenium of the corpus callosum and right superior longitudinal fasciculus.

Unfortunately, these longitudinal changes were not compared to changes within a healthy control group so the extent to which such changes were influenced by normal aging is unknown.

(29)

participants every 3 months for 1 year. Within- and between-group (AD and controls) comparisons were made. Within-groups, AD had MD increases in the inferior and anterior fornix. No FA changes were observed over time in AD. Healthy controls had increased MD in the inferior cingulum and decreased MD in the posterior cingulum. Low FA and high MD was observed in the fornix and splenium of individuals with AD

relative to healthy controls across all time points. Individuals with AD also had larger MD increases in the inferior cingulum relative to healthy controls at 6 and 12 months. Overall, preliminary studies suggest that there are microstructural white matter changes across time, including decreased FA and increased MD in a number of regions; however, given the small sample sizes and limited control groups for comparison, further investigation is needed in large cohorts of individuals with AD and healthy controls.

Relationship Between DTI Indices and Cognitive Performance

A number of studies have also examined whether there is a relationship between the microstructural white matter characteristics detected by DTI and cognitive

functioning. Some investigators have observed a correlation between DTI indices and general cognitive status as measured by the Mini Mental Status Exam (MMSE). For example, Bozzali and colleagues (2002) found a relationship between MMSE score and overall white matter FA (positive relationship) and MD (negative relationship) in

individuals with AD. Scrascia and colleagues (2014) also observed a positive relationship between MMSE scores and FA, but this correlation was confined to the frontal white matter. In contrast, other studies have not observed a relationship between MMSE scores and DTI indices (e.g., Ibrahim et al., 2009; Stahl et al., 2007).

(30)

Other studies have looked beyond the MMSE and employed more extensive neuropsychological testing. Bosch and colleagues (2012) examined the relationship between whole-brain DTI indices and four neuropsychological composite scores: 1) memory composite (Consortium to Establish a Registry for AD (CERAD)’s Recall of Constructional Praxis; Grober and Buschke Free and Cued Selective Reminding Test); 2) executive composite (Digit Span Backward, Symbol Search, and Similarities from the Wechsler Adult Intelligence Scale III; Controlled Oral Word Association Test); 3) language composite (Boston Naming Test; Auditory Comprehension of the Boston Diagnostic Aphasia Examination); 4) visuoperceptive-visuospatial (Incomplete Letters, Number Location of the Visual Object and Space Perception). Here, high FA and low MD were related to better performance on the memory composite when assessing AD, amnestic mild cognitive impairment (MCI) and healthy controls simultaneously. When the correlation analysis was confined to the amnestic MCI and AD groups, only FA was correlated with scores on memory tests. These results suggest that FA may be most strongly related to memory performance in AD. Serra and colleagues (2010) also

observed a relationship between FA and performance on immediate and delayed memory tests. Specifically, there was a positive relationship between anterior thalamic radiation and fornix FA and immediate recall (as assessed by the Rey Complex Figure Test) and corpus callosum FA and delayed recall (as assessed by the 15 Rey’s Word List).

Finally, there is evidence to suggest that the region of the brain considered influences the relation between DTI indices and cognitive performance. Huang and Auchus (2007) found DTI indices (e.g., FA) in the temporal lobe were correlated with performance on CERAD episodic memory tests (Word List Memory Total; Word List

(31)

Delayed Recall), indices in the frontal lobe were correlated with performance on CERAD episodic memory and executive function tests (Trails A), diffusion indices in the parietal lobe were correlated with general cognition (MMSE).

Rationale for the Investigation of DTI as a Biomarker for AD In summary, DTI holds potential as a biomarker for early identification of AD given its sensitivity to detect microstructural characteristics in the brain that are related to cognitive performance. However, multi-year tracking of brain imaging and AD clinical signs are necessary to evaluate the utility of white matter measures as an AD biomarker. The cross-sectional studies to date are methodologically insufficient to detect meaningful change across time, and to predict progression from healthy aging to AD.

Continued work is needed to investigate longitudinal DTI in conjunction with clinical data in a large sample. Ultimately, better characterizing the longitudinal

microstructural white matter changes in AD, as compared to healthy aging, may lead to earlier pre-symptomatic biomarkers for AD, using non-invasive neuroimaging.

(32)

Chapter 2: An Investigation of Microstructural White Matter

Characteristics in Alzheimer’s Disease and Healthy Aging Using

Diffusion Tensor Imaging

Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline (Alzheimer’s Association, 2014). Although a number of cognitive domains can be affected (e.g., orientation, attention, problem solving, constructional praxis), memory loss is often the initial and primary concern (Alzheimer’s Association, 2014). Approximately 747 000 Canadians are affected by AD or a related dementia; this number is expected to climb to 1.4 million Canadians by 2031 (Alzheimer’s Society of Canada, 2012). In the United States, 5.2 million Americans are affected by AD and the Alzheimer’s Association (2014) estimates that one new case develops every 33 seconds.

Currently, there is no cure for AD. Emerging research has focused on both the identification of biomarkers, and possible treatment options that could be administered at the earliest time point to delay disease progression. Indeed, there is increasing support that brain pathology precedes clinical symptoms, up to 10 to 20 years before the dementia stage (Sperling et al., 2011; Wurtman 2015). The majority of MRI-based biomarker research has focused on loss in grey matter structures. Consistent with post-mortem findings, anatomical MRI findings have indicated widespread whole brain grey matter atrophy in individuals with AD, including enlarged ventricles and decreased hippocampal volume (Perl, 2010; Teipel et al., 2010). To date, the best-established MRI biomarker for

(33)

AD is decreased hippocampal volume (Gold, 2012; Hampel et al., 2014). Unfortunately, these larger-scale (macroscopic) measures are only indicative of the later stages of AD pathology. Therefore, these are unlikely to be sufficient measures for the earliest detection of AD (de Gois Vasconcelos et al., 2009). The Amyloid Cascade Hypothesis (Hardy & Higgins, 1992) predicts that axonal degeneration (white matter) precedes neuronal loss (grey matter), suggesting that white matter biomarkers may lead to earlier identification of AD. Despite this, there remains a paucity of research on early white matter in AD.

One promising tool to detect early white matter alterations in the brain is diffusion tensor imaging (DTI; Alexander, Lee, Lazar, & Field, 2007; Soares, Marques, Alves & Soursa, 2013; Mori & Zhang, 2009). There are a number of common DTI measures used to assess microstructural alterations in the brain including: 1) fractional anisotropy (FA) and 2) mean diffusivity (MD). FA is a measure of the degree of directionality of water diffusion (Alexander et al., 2007; Amlien & Fjell, 2014; Soares et al., 2013; Stebbins & Murphy, 2009), while MD is a measure of the mean water diffusion rate (Soares et al., 2013). When pathological processes affect white matter in the brain, decreases in FA are often observed along with increases in MD in white matter (e.g., Douaud et al., 2011; Shu, et al., 2011).

To date there are a number of brain regions that have consistently shown altered FA and MD patterns within the growing body of literature on AD. For example, in

comparing a large number of DTI studies, Stebbins and Murphy (2009) found that low FA and high MD were commonly observed in the white matter of all of the brain lobes, including the medial temporal lobe structures– regions known to be involved in memory

(34)

function (Stoub, deToledo, & Dickerson, 2014). One of the first DTI meta-analyses found widespread differences in both FA and MD in individuals with AD across 41 ROI studies (Sexton, Kalu, Filippini, MacKay, & Ebmeier, 2011). Specifically, low FA was reported in the frontal and temporal lobes; the genu and splenium of the corpus callosum; the anterior, middle, and posterior cingulum; parahippocampal cingulum, uncinate

fasciculus; and superior longitudinal fasciculus. Similarly, high MD was observed in all of the lobes of the brain; the genu and splenium the corpus callosum; posterior cingulum; uncinate fasciculus; and the hippocampus. A more recent meta-analysis of 13 whole-brain TBSS studies also reported low FA and high MD in individuals with AD (Acosta-Cabronero & Nestor 2014). When AD was compared to healthy aging controls, these white matter alterations were usually observed in the parietal, temporal and prefrontal white matter, including the association fibres and inter-hemispheric connections via the corpus callosum.

Despite the number of cross sectional studies using DTI in AD, there have been few longitudinal studies have been published to date, and of the published studies, most have had small sample sizes (Amlien & Fjell, 2014). In one study, Kitamura and colleagues (2013) scanned individuals with AD at baseline and approximately 1.5 year later and observed FA decreases within the uncinate fasciculus. However, as no healthy control group comparisons were made at the second time point, it cannot be determined if such longitudinal changes reflect normal aging. In another study, Norwrangi and colleagues (2013) assessed individuals with AD and control participants every 3 months for 1 year and made within- and between-group comparisons of DTI indices in 8 ROIs. Individuals with AD had MD increases in the anterior and inferior fornix over one year while lower

(35)

FA and higher MD was also observed in the fornix and splenium of individuals with AD relative to healthy controls across all time points. Individuals with AD also had larger MD increases in the inferior cingulum relative to healthy controls at 6 and 12 months.

A number of studies have also examined whether microstructural white matter alterations detected by DTI are related to cognitive functioning in those with AD. Some investigators have observed a relationship between DTI indices and general cognitive status as measured by the Mini Mental Status Exam (MMSE; e.g., Bozzali et al., 2002; Scrascia et al., 2014), while others have not (e.g., Ibrahim et al., 2009; Stahl et al., 2007). In a study that employed more extensive neuropsychological testing, Bosch and

colleagues (2012) examined the relationship between whole-brain DTI indices and four neuropsychological composite scores: 1) memory composite (Consortium to Establish a Registry for AD’s Recall of Constructional Praxis; Grober and Buschke Free and Cued Selective Reminding Test); 2) executive composite (Digit Span Backward, Symbol Search, and Similarities from the Wechsler Adult Intelligence Scale III; Controlled Oral Word Association Test); 3) language composite (Boston Naming Test; Auditory

Comprehension of the Boston Diagnostic Aphasia Examination); 4) visuoperceptive-visuospatial ( Incomplete Letters, Number Location of the Visual Object and Space Perception). High FA and low MD were related to better performance on the memory composite when individuals with AD, amnestic mild cognitive impairment (MCI) and healthy controls were considered simultaneously. However, when the analysis was confined to the AD and the MCI group, only FA was correlated with scores on memory tests, suggesting that FA may be most strongly related to memory performance. Serra and colleagues (2010) also observed a relationship between FA values in the right anterior

(36)

thalamic radiation and fornix and performance on immediate recall (Rey Complex Figure Test) as well as between corpus callosum FA and delayed (15 Rey’s Word List) recall tests. Huang and Auchus (2007) examined relations of cognitive and DTI indices within specific brain regions and found that temporal lobe FA was correlated with performance on episodic memory tests (Word List Memory Total; Word List Delayed Recall); frontal lobe FA was correlated with performance on CERAD episodic memory and executive function tests (Trails A); and parietal lobe FA was correlated with general cognition (MMSE). These limited studies demonstrate the need for longitudinal studies of the potential DTI measures to predict conversion from healthy aging to AD (Hampel et al., 2011).

The current study used diffusion tensor imaging (DTI) and clinical data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database to examine 1) within-group microstructural white matter changes in individuals with AD and healthy controls at baseline and year one; 2) between-group differences in individuals AD and healthy controls at both time points; and 3) the relationship between white matter and cognitive performance at both time points.

Considering the data available from limited longitudinal studies of AD and cross-sectional studies of AD compared to healthy aging, it is hypothesized that 1) individuals with AD will have decreased FA and increased MD across time; 2) individuals with AD will have lower FA and higher MD as compared to controls at both time points. In particular, given the AD neurodegeneration observed in grey matter (Braak & Braak, 1991), we predict that white matter in the medial temporal lobe will be more greatly affected in AD, compared to controls. Finally, it is hypothesized that 3) there will be a

(37)

relationship between DTI indices and cognitive performance in AD and controls at both time points.

Method and Materials

All data were obtained from the Alzheimer’s Disease Neuroimaging Initiative 2 (ADNI2) database (http://adni.loni.usc.edu), the third of three data collection phases (ADNI1, ADNIGO, ADNI2). The ADNI, led by Principal Investigator Dr. Michael W. Weiner, was launched in 2003 with the goal of testing whether longitudinal brain

imaging, biological markers, and neuropsychological assessment can be used together to measure the progression of AD. Data was collected from approximately 200 individuals with AD, 600 individuals with mild cognitive impairment (MCI), and 200 healthy controls across the first two collection phases (ADNI1 and ADNIGO) from nearly 55 sites in North America. The third phase of the ADNI (ADNI2) is now underway. ADNI2 is currently collecting data from 200 individuals with AD, 300 individuals with MCI, and 150 healthy controls. For more information, please see www.adni-info.org.

Participants

Full eligibility criteria for the ADNI are described in the ADNI2 procedures manual (Alzheimer’s Disease Neuroimaging Initiative, 2008). Individuals with AD met

NINCDS/ADRDA criteria for probable AD (McKhann et al., 1984), demonstrated abnormal memory function on the Wechsler Memory Scale (WMS) II (< 8 for 16 years education and above), had a MMSE score between 20 and 26, and had a Clinical Dementia Rating of 0.5 (very mild) or 1.0 (mild).

(38)

score within the normal range on the WMS Logical Memory II (> 9 for 16 years of education and above), have a MMSE score between 24 and 30, and a Clinical Dementia Rating of 0 (none).

Data were collected from 34 individuals with AD (mean age = 75.8 ± 7.6; 10 females; MMSE = 23.59 ± 1.74; Logical Memory II = 1.65 ± 1.94,) and 33 healthy age-matched controls (mean age = 73.0 ± 6.6; 16 females; MMSE = 29.03 ± 1.26, Logical Memory II = 11.70 ± 2.84) at two time points: baseline and year one (see Figure 3 for flow diagram of participant selection). The mean number of days from baseline to year one was 394 ± 25 for AD, and 403 ± 54 for healthy controls. Demographic information is shown in Table 1.

(39)

Figure 3. Flow diagram of participant selection for individuals with Alzheimer’s disease and healthy controls. Final selection criteria included individuals with Alzheimer’s disease (AD) or healthy controls (CN) from the ADNI2 database with axial diffusion-weighted magnetic resonance imaging scans at baseline and year one.

ADNI1, ADNIGO, ADNI2

MRI data

N=1832

ADNI2

MRI data

N=1183

ADNI2

Axial DTI

N=269

ADNI2

Axial DTI + AD + CN

N=113

ADNI2

Axial DTI + AD + Baseline + Year 1

N= 34

ADNI2

Axial DTI + CN + Baseline + Year 1

N= 35*

*2 removed due to insufficient

# of slices

(40)

Table 1.

Participant Demographics and Clinical Scores at Baseline and Year One

Baseline Year One Within-group

AD CN AD vs. CN AD CN AD vs. CN AD CN Age 75.8 ± 7.6 73.0 ± 6.6 p = 0.104 76.9 ± 7.7 74.1 ± 6.5 p = 0.114 - - # Males 24 17 p = 0.113 - - - - - # Females 10 16 - - - - - Education 15.7 ± 2.9 16.4 ± 2.8 p = 0.347 - - - - - ADNI-MEM -0.58 ± 0.34 0.86 ± 0.55 p < 2.20 e-16 _{-0.88 ± 0.60} _{0.95 ± 0.60} _{p < 2.20 e}-16 _{p < 0.015} _{p = 0.540} ADNI-EF a _{-0.64 ± 0.79} _{0.86 ± 0.75} _{p < 2.94 e}-11 _{-0.89 ± 0.81} _{0.84 ± 0.66} _{p < 6.67 e}-14 _{p = 0.168} _{p = 0.916}

a_{ADNI-EF composite scores were unavailable for 1 CN at year one .: N = 32}

All ADNI participants provided informed written consent approved by each sites’ Institutional Review Board. Secondary data use for the current study was approved by the Human Research Ethics Board at the University of Victoria, in British Columbia,

Canada.

Image Acquisition

MRI data were downloaded from the ADNI2 database. All participants underwent whole-brain MRI scans according to the ADNI protocol. Images were acquired from 3T MRI scanners (GE Medical Systems) from 7 North American acquisition sites. Axial diffusion weighted image data were acquired with a spin echo echo planar imaging sequence. Scan parameters are as follows: acquisition matrix = 256 x 256, voxel size = 1.4 x 1.4 x 2.7mm3, flip angle = 90 degree, number of slices = 59. There were 46 images acquired for each scan: 41 diffusion-weighted images (b = 1000 s/mm2_{) and 5} non-diffusion-weighted images (b = 0 s/mm2). Repetition time (TR) varied across scanning sites, but was approximately 12500 to 13000 ms.

(41)

Neuropsychological Data

All neuropsychological data were downloaded from the ADNI2 database. All participants completed a battery of neuropsychological tests, as outlined in the ADNI2 procedures manual (Alzheimer’s Disease Neuroimaging Initiative, 2008). Two composite scores derived from the ADNI neuropsychological battery were used in this study: 1) ADNI-MEM, a memory composite score; and 2) ADNI-EF, an executive functioning composite score. Composite scores were chosen to minimize the effects of potential outlying performance on a single test item. ADNI-MEM was derived by confirmatory factor analysis using data from the Rey Auditory Verbal Learning Test (RAVLT), ADAS-Cog, WMS Logical Memory, and MMSE (See Crane et al., 2012 for full details on psychometric development). ADNI-EF was derived by item response theory using data from the Wechsler Adult Intelligence Scale – Revised (WAIS-R) Digit Symbol Substitution, Digit Span Backwards, Trails A and B, Category Fluency, and Clock Drawing (See Gibbons et al., 2012 for full details on psychometric development).

Data Analysis

Image preprocessing. All data analyses were performed in Functional MRI of

the Brain Software Library (FSL) version 5.0 (Analysis Group, FMRIB, Oxford, UK, http://fsl.fmrib.ox.ac.uk; Smith et al., 2004). Diffusion weighted images were corrected for eddy current distortions and head movement using Eddy Current Correction, and non-brain tissue was removed using Brain Extraction Tool (Smith, 2002). Brain-extracted images were then visually inspected to ensure brain tissue was not removed.

Image analysis. FA maps were created using DTIfit and input into TBSS to

(42)

2006). First, all participants’ FA data were non-linearly aligned to common space. Second, the mean FA image was created and thinned (threshold FA = 0.2) to create the mean FA skeleton. Third, each participant’s FA data was projected onto the thresholded mean FA skeleton. Voxelwise statistical analysis of the white matter skeleton was

performed using Randomise, FSL’s nonparametric permutation inference tool. Threshold free cluster enhancement was used to correct for multiple comparisons. TBSS was also performed for MD; non-linear registration was applied to MD data and all participants’ MD data was merged into a 4D file. Each participant’s MD data was projected onto the mean FA skeleton before applying voxelwise statistics.

Statistical comparisons. Within-group contrast comparisons were made for individuals with AD from baseline to year one, and for healthy controls from baseline to year one. Additionally, between-group contrast comparisons were made between

individuals with AD and healthy controls at both baseline and at year one. Correlations were examined between participants’ diffusion metrics (FA; MD) and ADNI composite scores (ADNI-MEM; ADNI-EF) at baseline and at year one. White matter regions were identified with Johns Hopkins University’s white matter atlas included in FSL (Mori et al., 2008; Wakana et al., 2007).

Results

Within-group Microstructural White Matter Changes in Individuals with AD and Healthy Controls at Baseline and Year One

The within-group FA analysis showed that individuals with AD had reductions in FA in multiple regions including the hippocampal cingulum at year one compared to baseline (Table 2; Figure 4A). Controls also had reduced FA in similar regions, but these

(43)

alterations were less extensive, and did not include the hippocampal cingulum (Figure 4B).

Table 2.

Regions Showing Low Fractional Anisotropy at Year One Compared to Baseline in i) Individuals with Alzheimer’s Disease and in ii) Healthy Controls (p <0.05, Corrected for Multiple Comparisons).

Alzheimer’s Disease Healthy Controls

Genu, body and splenium of the corpus callosum

Genu, body, and splenium of the corpus callosum

Internal capsule Internal capsule

External capsule External capsule

Anterior, superior and posterior corona

radiata Anterior, superior and posterior corona radiata Posterior thalamic radiation Posterior thalamic radiation

Superior and inferior longitudinal

fasciculus Superior and inferior longitudinal fasciculus Superior and inferior fronto-occipital

fasciculus

Superior and inferior fronto-occipital fasciculus

Cingulate gyri Cingulate gyri

Fornix Fornix

-- Corticospinal tract

Uncinate fasciculus Uncinate fasciculus

Tapetum Tapetum

Medial lemniscus Medial lemniscus

Superior, inferior and middle cerebellar

peduncle Superior, inferior and middle cerebellar peduncle

Cerebral peduncle Cerebral peduncle

(44)

Figure 4. Results of within-group Tract-Based Spatial Statistics white matter analysis showing pattern of reduced fractional anisotropy (red) at year one compared to baseline in individuals with Alzheimer’s disease (panel A) and in controls (panel B; p <0.05, corrected for multiple comparisons). Images on overlaid on mean fractional anisotropy skeleton (green) and T1-weighted MNI152_T1_1mm standard template provided by Functional MRI of the Brain’s Software Library.

(45)

The within-group MD analysis showed that individuals with AD also had increased MD in multiple regions including the hippocampal cingulate at year one compared to baseline (Table 3; Figure 5A). Controls also had increased MD at year one compared to baseline, but once again, these alterations were less extensive than in AD, and did not include the hippocampal cingulum (Figure 5B).

Table 3.

Regions Showing Increased Mean Diffusivity at Year One Compared to Baseline in i) Individuals with Alzheimer’s Disease and in ii) Healthy Controls (p <0.05, Corrected for Multiple Comparisons).

Alzheimer’s Disease Healthy Controls

Anterior, superior and posterior corona radiata

Posterior thalamic radiation Posterior thalamic radiation Superior and inferior longitudinal

fasciculus

Superior and inferior longitudinal fasciculus

Fornix Fornix

Uncinate fasciculus --

Tapetum Tapetum

Medial lemniscus Medial lemniscus

Superior, inferior and middle cerebellar peduncle

Cerebral peduncles Cerebral peduncles

(46)

Figure 5. Results of within-group Tract-Based Spatial Statistics white matter analysis showing pattern of increased mean diffusivity (blue) at year one compared to baseline in individuals with Alzheimer’s disease (panel A) and in controls (panel B; p <0.05,

corrected for multiple comparisons). Images on overlaid on mean fractional anisotropy skeleton (green) and T1-weighted MNI152_T1_1mm standard template provided by Functional MRI of the Brain’s Software Library.

(47)

Between-group Differences in Individuals with AD and Healthy Controls at Baseline and Year One

At baseline, between-group TBSS revealed that individuals with AD had lower FA relative to controls (Table 4; Figure 6A). At year one, individuals with AD also had lower FA relative to controls, and these alterations appeared more widespread than at baseline (Figure 6B).

Table 4.

Regions Showing Low Fractional Anisotropy in Individuals with Alzheimer’s Disease Compared to Healthy Controls at i) Baseline and ii) Year One (p <0.05, Corrected for Multiple Comparisons).

Baseline Year One

Anterior, superior and posterior corona

radiata Anterior, superior and posterior corona radiata Posterior thalamic radiation Posterior thalamic radiation

Superior and inferior longitudinal fasciculus

Inferior fronto-occipital fasciculus Inferior fronto-occipital fasciculus

Fornix Fornix

Tapetum Tapetum

-- Medial lemniscus

-- Superior, inferior and middle cerebellar

peduncle

-- Cerebral peduncles

(48)

Figure 6. Results of between-group baseline (panel A) and year one (panel B) Tract-Based Spatial Statistics white matter analysis showing pattern of lower fractional anisotropy (red) overlaid on the mean fractional anisotropy skeleton (green) in individuals with Alzheimer’s disease compared to controls (p < 0.05, corrected for multiple comparisons). Images overlaid on the T1-weighted MNI152_T1_1mm standard template provided by Functional MRI of the Brain’s Software Library.

(49)

The between-group TBSS also revealed that individuals with AD had higher MD relative to controls at baseline (Table 5; Figure 7A). At year one, individuals with AD had higher MD relative to controls in similar regions as baseline, but there were also patterns of higher MD in the hippocampal cingulum, not seen at baseline (Figure 7B).

Table 5.

Regions Showing Higher Mean Diffusivity in Individuals with Alzheimer’s Disease Compared to Healthy Controls at i) Baseline and ii) Year One (p <0.05, Corrected for Multiple Comparisons).

Baseline Year One

Genu, body and splenium of the corpus

callosum Genu, body, and splenium of the corpus callosum

Anterior, superior and posterior corona radiata

Posterior thalamic radiation Posterior thalamic radiation Superior and inferior longitudinal

fasciculus Superior and inferior longitudinal fasciculus Superior and inferior fronto-occipital

fasciculus

Fornix Fornix

Tapetum Tapetum

Cerebral peduncles Cerebral peduncles

(50)

Figure 7. Results of between-group baseline (panel A) and year one (panel B) Tract-Based Spatial Statistics white matter analysis showing pattern of mean diffusivity (blue) overlaid on the mean fractional anisotropy skeleton (green) in individuals with

Alzheimer’s disease compared to controls (p < 0.05, corrected for multiple comparisons). Images overlaid on T1-weighted MNI152_T1_1mm standard template provided by Functional MRI of the Brain’s Software Library.

An investigation of microstructural white matter changes in Alzheimer’s disease and healthy aging using diffusion tensor imaging

Supervisory Committee

Abstract

Table of Contents

List of Tables

List of Figures

Acknowledgments

Dedication

Chapter 1: Using Diffusion Tensor Imaging to Identify White

Matter Biomarkers in Alzheimer’s Disease

Chapter 2: An Investigation of Microstructural White Matter

Characteristics in Alzheimer’s Disease and Healthy Aging Using

Diffusion Tensor Imaging

ADNI1, ADNIGO, ADNI2

MRI data

N=1832

ADNI2

MRI data

N=1183

ADNI2

Axial DTI

N=269

ADNI2

Axial DTI + AD + CN

N=113

ADNI2

Axial DTI + AD + Baseline + Year 1

N= 34

ADNI2

Axial DTI + CN + Baseline + Year 1

N= 35*

*2 removed due to insufficient

# of slices