ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2017

Digital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Medicine 1394

Diffusion tensor magneticresonance imaging of the brain

Tractography analysis with application in healthyindividuals and patients

JOHANNA MÅRTENSSON

ISSN 1651-6206ISBN 978-91-513-0147-1urn:nbn:se:uu:diva-332596

Dissertation presented at Uppsala University to be publicly examined in Gustavianum,Auditorium Minus, Akademigatan 3, Uppsala, Wednesday, 20 December 2017 at 09:15 forthe degree of Doctor of Philosophy (Faculty of Medicine). The examination will be conductedin English. Faculty examiner: Associate professor Tim Dyrby (Danish Research Centre forMagnetic Resonance, Copenhagen University Hospital, Hvidovre, Denmark).

AbstractMårtensson, J. 2017. Diffusion tensor magnetic resonance imaging of the brain. Tractographyanalysis with application in healthy individuals and patients. Digital ComprehensiveSummaries of Uppsala Dissertations from the Faculty of Medicine 1394. 62 pp. Uppsala: ActaUniversitatis Upsaliensis. ISBN 978-91-513-0147-1.

In study 1, thirty-eight healthy controls were used for optimization of the method. Fifteenpatients with progressive supranuclear palsy and an equal number of age-matched healthycontrols underwent diffusion tensor MRI and were then investigated and compared groupwise.

It was shown that tractography analyses may preferably be performed regionally, such asalong the tracts or in different segments of the tracts. Normalization of white matter tracts canbe performed using anatomical landmarks.

In study 2, 104 males and 153 females in the age interval 13 to 84 years of age participatedas healthy individuals in order to investigate age-related changes with diffusion tensor MRI.

It was shown that spatially differences in age-related changes exist between subdividedsegments within white matter tracts. The aging processes within the CB and the IFO varyregionally.

In study 3, 38 human brains were used for investigation of the white matter tract inferiorlongitudinal fasciculus (ILF) and its subcomponents. Of these, white matter anatomicaldissection was performed in 14 post-mortem normal human brains. The remaining 24 brainswere investigated in vivo with diffusion tensor MRI in healthy individuals.

It was validated that fibers of the ILF in the occipito-temporal region have a clear, constant anddetailed organisation. The anatomical connectivity pattern, and quantitative differences betweenthe ILF subcomponents, confirmed a pivotal role of the ILF.

In study 4, 12 patients with iNPH were included in the study and examined with diffusiontensor at three time points. For comparison, 12 healthy controls, matched by gender and agewere also included. Controls were examined with MRI only once.

It was shown that DTI measures differ significantly between patients with iNPH and healthycontrols. DTI measures of the CC, the CST and the SLF, correlated to changes in clinicalsymptoms after shunt surgery.

Deeper knowledge about functions of the brain increases possibilities to take advantages fromDTI analyses with tractography.

Keywords: Magnetic Resonance Imaging, Diffusion Tensor Imaging, Tractography, whitematter tracts

Johanna Mårtensson, Department of Surgical Sciences, Radiology, Akademiska sjukhuset,Uppsala University, SE-75185 Uppsala, Sweden.

© Johanna Mårtensson 2017

ISSN 1651-6206ISBN 978-91-513-0147-1urn:nbn:se:uu:diva-332596 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-332596)

To my Daughter - My Hero With all my Love

List of Papers

This thesis is based on the following papers, which are referred to in the text by their Roman numerals.

I Mårtensson, J., Nilsson, M., Ståhlberg, F., Sundgren, P., Nilsson,

C., van Westen, D., Larsson E-M., Lätt, J. (2013). Spatial analy-sis of diffusion tensor tractography statistics along the inferior fronto-occipital fasciculus with application in progressive supra-nuclear palsy. Magnetic Resonance Materials in Physics, Biol-ogy and Medicine, 26(6): 527-37

II Mårtensson, J., Lätt, J., Åhs, F., Fredrikson, M., Söderlund, H.,

Schiöth, H.B., Kok, J., Kremer, B., van Westen, D., Larsson, E-M., Nilsson, M. (2018). Diffusion tensor imaging and tractog-raphy of the white matter in normal aging: The rate-of-change differs between segments within tracts. Magnetic Resonance Imaging, 45(27): 113-119

III Latini, F., Mårtensson, J., Larsson, E-M., Fredrikson, M., Åhs,

F., Hjortberg, M., Aldskogius, H., Ryttlefors, M. (2017). Seg-mentation of the inferior longitudinal fasciculus in the human brain: A white matter dissection and diffusion tensor tractog-raphy study. Brain Research, 1675(15): 102-115

IV Mårtensson, J., Lätt, J., van Westen, D., Kremer, B., Fahlström,

M., Laurell, K., Virhammar, J., Larsson, E-M. (2017). Pre-oper-ative DTI of corticospinal tract, corpus callosum and superior longitudinal fasciculus in patients with idiopathic normal pres-sure hydrocephalus correlates with changes in clinical findings after shunt surgery. Manuscript

Reprints were made with permission from the respective publishers.

Contents

Introduction ................................................................................................... 11Magnetic resonance imaging .................................................................... 11Diffusion ................................................................................................... 13

Pulsed gradient spin echo sequence ..................................................... 14Artefacts and errors ............................................................................. 16Diffusion in tissue ................................................................................ 16Diffusion tensor ................................................................................... 18Diffusion measures .............................................................................. 19

Tractography ............................................................................................ 20Deterministic tractography .................................................................. 20Clinical applications of tractography ................................................... 22Diffusion in white matter ..................................................................... 22Investigated white matter tracts ........................................................... 24

Aims .............................................................................................................. 32General aim .............................................................................................. 32Specific aims ............................................................................................ 32

Materials and methods .................................................................................. 33Study 1 ...................................................................................................... 33

Subjects ................................................................................................ 33MR imaging ......................................................................................... 33Data analysis ........................................................................................ 33

Study II ..................................................................................................... 35Subjects ................................................................................................ 35MR Imaging ......................................................................................... 36Data analysis ........................................................................................ 36

Study III .................................................................................................... 37Subjects ................................................................................................ 37Anatomical dissection .......................................................................... 38MR Imaging ......................................................................................... 38Data analysis ........................................................................................ 38

Study IV ................................................................................................... 39Subjects ................................................................................................ 39MR imaging ......................................................................................... 39Data analysis ........................................................................................ 40

Results ........................................................................................................... 42

Discussion ..................................................................................................... 43

Conclusions ................................................................................................... 49General conclusion ................................................................................... 49Specific conclusions ................................................................................. 49

Future aspects ................................................................................................ 51

Acknowledgements ....................................................................................... 53

Sammanfattning på svenska .......................................................................... 57

References ..................................................................................................... 60

Abbreviations

AAC Apparent area coefficient AD Axial diffusivity ADC Apparent diffusion coefficient AI Asymmetry index ANOVA Analysis of variance CB Cingulum bundle CC Corpus callosum CSF Cerebrospinal fluid CST Corticospinal tract D Diffusion constant DTI Diffusion tensor imaging DWI Diffusion weighted imaging EPI Echo planar imaging FA Fractional anisotropy FACT Fiber assignment by continuous tracking algorithm FLIRT FMRIB´s linear image registration tool FSL FMRIB software library IFO Inferior frontooccipital fasciculus ILF Inferior longitudinal fasciculus MD Mean diffusivity MMSE Mini mental state examination MNI Montreal Neurological institute MRI Magnetic resonance imaging ODF Orientation density function PGSE Pulsed gradient spin echo PNS Peripheral nerve stimulation PSP Progressive supranuclear palsy QUTE Quantitative tractography evaluation RD Radial diffusivity ROI Region of interest SLF Superior fronto occipital fasciculus TE Echo time TR Repetition time

11

Introduction

Magnetic resonance imaging Magnetic resonance imaging (MRI) is an imaging technique used to visualize for example human tissues in vivo, both anatomically and functionally. The technique has been developed and improved continuously since its introduc-tion. These efforts have improved the technique further, and new methods continue to be introduced. MRI technology is based on the action of particles with spin angular momentum, especially the hydrogen nucleus, in a magnetic field. The magnetization precesses with a frequency that is proportional to the outer magnetic field, which was described in 1946, and the idea to use this physical phenomenon to generate images by performing manipulations of the magnetic field around hydrogen protons was introduced and described in 1973 [1-3].

The precessing frequency is referred to as the resonance frequency, also named as the Larmor frequency after the Irish mathematician and physicist Joseph Larmor (1857–1942). The gyromagnetic ratio, γ, is defined as the ratio of the magnetic moment to its angular momentum, and it couples the magnetic flux density, B, to the resulting frequency, f, according to

Equation 1 𝑓 = 𝛾𝐵

The hydrogen nucleus, i.e. a single proton, is the most common nucleus to

be used in clinical applications and has its γ defined as 42 MHz/T. Clinical MR scanners typically have static magnetic fields, B0, with a strength of 1.5 or 3 tesla (T). The scanners used in this thesis had B0 strengths of 3 T, which corresponds to a resonance frequency of 127 MHz. The precessing motion of the total magnetization, M, when contributions from of all nuclei are taken together, is described by

Equation 2 %&%'= 𝛾(𝑀𝑥𝐵)

12

By adding another magnetic field with a lower magnitude, B1, the direction of M gradually changes. B1 varies with time, and because of this the MR signal can be generated.

The resonance frequency depends on B. By introducing spatial variations in B, i.e. magnetic field gradients, a controlled way of selecting locations in which the protons can undergo excitation is possible. With the use of magnetic field gradients, a spatial dependency of frequency and phase of excited protons is also made possible. Magnetic field gradients are applied together with ex-citation pulses in a planned and certain way so that it becomes possible to encode certain groups of protons, to excite and spatially resolve them, and in such a way that an MR image is finally produced.

After protons have undergone excitation, they return to their resting states. This process is called spin relaxation, and consists of two relaxation processes, independent of each other. One of the relaxation processes is called the longi-tudinal relaxation, describing the recovery of the longitudinal magnetization, and the other one is called the transverse relaxation, describing the decay of the transverse magnetization.

The longitudinal relaxation process is a recovery during which the longitu-dinal magnetization, Mz, increases along B0 after the excitation pulse has been applied. The recovery of Mz after the excitation is described by

Equation 3 𝑀𝑧 𝑡 = 𝑀./(1 − 𝛽𝑒4'/67)

where β is the fraction of Mz0 to be recovered and T1 is the longitudinal

relaxation time in ms. The excitation pulse makes all the protons precess in phase. Spatial and dynamic variations in the magnetic environment lead to different rates of phase dispersion, and the protons lose their synchronized phases during the transversal relaxation time. The transverse magnetization remaining at time, t, after the excitation is given by

Equation 4 𝑀89 𝑡 = 𝑀89/𝑒4'/6:

where Mxy is the remaining transverse magnetization, Mxy0 is the initial

transverse magnetization, i.e. when all the protons are in phase, and T2 is the transverse relaxation time in ms.

13

Diffusion Since the middle of the 1980s, diffusion-weighted magnetic resonance imag-ing (DWI-MRI) has been increasingly used to investigate tissue functions in various organs [4]. Diffusion weighting has been utilized extensively in neu-roimaging during the last decades after it was shown that ischemic strokes could be detected earlier with DWI than was possible using only T1- or T2-weighted images [5].

Diffusion is a physical, macroscopic measure, known as Brownian motion when the term is used to define for free molecules. It is described as the motion of particles, for example water molecules in fluid water, and is proportional to the diffusion time, 𝑡;, the molecule sizes, and the surrounding environment. If the water is cold, the particles move slower than if the water is warm, since the particles move around with a velocity depending on their thermal energy. The motion itself is statistically well described by a diffusion coefficient, D, which depends on the size of the molecules, the temperature and the viscosity of the medium [6]. The distance between their initial and final positions de-pends on the velocity of the particles, and in which manner they interact with surrounding molecules. The transfer seems random because it is impossible to predict where the molecules will end up. A schematic representation of self-diffusion due to thermal energy in an environment consisting of surrounding molecules can be seen in Figure 1, where the lines between the molecules represent the paths of the molecules motions. The process of molecular motion is described by Fick’s second law, applied for situations of self-diffusion due to thermal energy. The probability P(x,t) to move the relative displacement x during the time t is described as in Equation 5. When t = 0, P(x,t) is a delta function, and when t > 0, P(x,t) is a Gaussian function.

Equation 5 𝜕𝑃/𝜕𝑡 = 𝐷(𝜕@𝑃)/(𝜕@𝑥)

Equation 5 describes Fick’s law in one dimension. Expanded to three di-

mensions, Fick’s law becomes as shown in Equation 6.

Equation 6 ABA'= 𝐷𝛻@𝑃

14

Figure 1. Water molecules move due to their thermal energy. An environment con-sisting of surrounding molecules affects the paths of the molecules motions. In hu-man tissue, the measured diffusion reflects the apparent environment with a certain amount of hindrance and restriction from, for example, cells and macromolecules.

Pulsed gradient spin echo sequence The design of the diffusion measurements used in this thesis was based on the pulsed gradient spin echo sequence (PGSE), introduced by Stejskal and Tan-ner in 1965 [7]. To obtain diffusion weighting to the MRI images, bipolar diffusion gradients are added to the pulse sequence used. These gradients are equal in strength and in spin echo sequences, they are applied before and after the refocusing 180° pulse. The degree of diffusion weighting is specified by the b value, which can be controlled by manipulating the diffusion gradient blocks. The b value is given by Equation 7.

Equation 7 𝑏 = 𝛾@𝐺@𝛿@ ∆ − HI

where b is the diffusion weighting, γ is the gyromagnetic constant, G is the amplitude of the gradient, δ is the time duration during which the diffusion gradient G is applied, and Δ is the time duration between the two applied dif-fusion gradient blocks. A schematic representation of the PGSE sequence is given in Figure 2.

15

In a single shot spin echo sequence, the entire raw data set is acquired after one excitation pulse. With only partial k space coverage and the rest of the k space half Fourier reconstructed, this method achieves a high speed that is of importance when measuring diffusion in multiple directions. The diffusion signal strength at a specified b value is described by Equation 8, where S is the diffusion weighted MR signal, b is the b value, and D is the self-diffusion constant of tissue.

Equation 8 𝑆 𝑏 = 𝑆(0)𝑒(4L;)

Single shot techniques can be used for echo planar imaging (EPI). Instead

of acquiring one line in k space at a phase encoding step, the complete k space is acquired during one repetition time when EPI is used for sampling.

Pulse sequences always require compromises, and so does this one. The compromises for speed in the single shot spin echo EPI are the sensitivity to susceptibility-induced image distortion, low resolution, and high artefact lev-els.

Figure 2. A schematic representation of the pulsed field gradient spin echo sequence which was introduced by Stejskal and Tanner. The radio frequency (RF) pulses shown are the 90° excitation pulse, the 180 ° inversion pulse and the echo, during which the signal is sampled and the read out encoding gradient (Gread) is active. The gradient for encoding slice (Gslice) is active during the excitation and the inversion pulses, and the phase encoding gradient (Gphase) is applied before the readout. The two diffusion encoding gradients (Gdiff) are equal and applied before and after the in-version pulse, respectively.

16

Artefacts and errors There are some challenges when implementing and analyzing the results from diffusion tensor imaging (DTI), and pulse sequences always require compro-mises. The image quality that comes with the single shot spin echo EPI may be geometrically warped because of the magnetic field inhomogeneity or the gradient-induced eddy currents, which can be seen as distorted images [8]. Corrections for eddy current and motions need to be applied in order to reduce such distortions. It is recommended that acquired images undergo manual in-spection to check for image artifacts or the occurrence of other errors. The gradient switches of the diffusion sensitive pulse sequences are strong and rapid, which may induce peripheral nerve stimulation, so called PNS.

Diffusion in tissue There are three major diffusional processes: the free diffusion, hindered dif-fusion, and restricted diffusion [9]. These processes can be further divided into isotropic and anisotropic diffusion [5].

In tissue, the measured diffusion depends on many factors. The microstruc-ture consists of barriers and obstacles such as cell membranes and myelin sheaths that slow down the diffusion. Barriers can also introduce an upper limit to their overall mean-square displacement so that the diffusion gets re-stricted. Restrictions and hindrance of the diffusion influences the measured diffusion signal, but in which way is hard to find out since it is a complicated issue. Figure 1 and Figure 3 represent schematic sketches of limiting and af-fecting environments in human tissue. In Figure 1, molecules are contributing to the molecules path between the cells, and in Figure 3, myelin sheaths of the axons are acting as barriers making the water molecules move along the di-rection of the axons. One thing that we can say for sure, is that the measured diffusion reflects the apparent environment with a certain amount of hindrance and restriction around the water molecules that move along an applied gradi-ent.

17

Figure 3. Principal sketch of the diffusion of the water molecules in white matter. Myelin sheaths of the axons are acting as barriers, making most of the water mole-cules move in parallel to the direction of the axons.

The square of the mean displacement, 𝑥@, and tD is linear, and free diffusion is described by Equation 9, where 𝑥 is the mean displacement, n is the meas-urement dimensions, D is the self-diffusion constant in tissue, and tD is the diffusion time. For one dimension, n is equal to 2, for two dimensions, n is equal to 4, and for three dimensions, n is equal to 6, and so on.

Equation 9 𝑥@ = 2𝑛𝐷𝑡;

It is the apparent displacement of water molecules in a specific situation

that is of interest in diffusion-weighted MRI rather than the D itself. Further, the quantitative value of D is specific in different media, which means that D cannot be quantified in the real environment of human tissues. It is rather the apparent diffusion that is measured in the tissue, and D is therefore replaced by the more reasonable, apparent diffusion coefficient, ADC.

18

When the water molecules move with the same equal possibility of dis-placement in all directions, the diffusion is isotropic. This is the situation when no oriented barriers preferentially affect the diffusion in one direction over another. In a situation in which barriers are present equally in all directions, the diffusion is hindered identically in all orientations, meaning that the diffu-sion is still considered isotropic [10]. If the displacement of the molecules is not identical in all directions, the diffusion is anisotropic. When the measured diffusivity is anisotropic, such as in the white matter, the diffusion may be better described by the diffusion tensor [11]. To describe diffusion anisotropy, the diffusion tensor must therefore be introduced. See also a schematic over-view of the tensor and tractography in Figure 4.

Diffusion tensor The diffusion tensor describes the current diffusion with nine coefficients such as described in Equation 1.

Equation 10 ∈ 𝐷 =𝐷88 𝐷89 𝐷8.𝐷98 𝐷99 𝐷9.𝐷.8 𝐷.9 𝐷..

but since 𝐷PQ = 𝐷QP, the diffusion tensor is completely described by six co-efficients. The diagonal elements 𝐷88, 𝐷99, 𝐷.. are associated with the eigen-values of the diffusion tensor matrix and can therefore be replaced by λ1, λ2, and λ3, according to

Equation 11 𝐷 =𝜆@ 𝐷89 𝐷8.𝐷89 𝜆@ 𝐷9.𝐷.8 𝐷.9 𝜆I

∈S∈@∈I

DTI is an MRI technique that allows the diffusion tensor to be estimated in each voxel by acquiring a series of diffusion MR images in different directions and fitting the tensor model to the data. Even though the lower threshold for solving the diffusion tensor is six gradient directions, a higher number of di-rections are used in DTI to provide robust estimates of the diffusion parame-ters.

19

Diffusion measures DTI can be used to study the microstructural development, the maturation, and the aging of the cerebral white matter across the life span [12]. The parameters that can be derived from the tensor model include fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD), which reflect various processes in the microstructure and the structural integ-rity within the white matter of the brain.

A scalar invariant known as the trace diffusion can be computed as

Equation 12 𝑇𝑟𝑎𝑐𝑒(𝐷) = 𝐷𝑥𝑥 + 𝐷𝑦𝑦 + 𝐷𝑧𝑧

And then further used to calculate the MD, which is simply the averaged

diffusion and is calculated as

Equation 13 𝑀𝐷 = 𝑇𝑟(𝐷)/3

An ellipsoid is always defined by its three eigenvalues, and so is the diffu-

sion tensor ellipsoid. AD is the first eigenvalue, which is the largest eigenvalue describing the predominant direction of the tensor ellipsoid, i.e. the AD = λ1. The RD is the mean of the other two eigenvalues, defining the cross section of the tensor ellipsoid, i.e. the RD = (λ2+ λ3)/2.

The FA presents the fraction of diffusion that is present in the predominant direction of diffusion. The eigenvalues of the tensor ellipsoid help defining the FA value according to

Equation 14 𝐹𝐴 = I@

]74]:^ ]:4]

:^ ]_4]:

]7:^]::^]_:

Degeneration of white matter can, for example, be indicated by decreased

FA, and age-related differences in FA have been reported in distinct region-specific patterns of numerous structures in older compared to younger adults [13]. In addition, chronic ischemic white matter lesions, which may be symp-tomatic or asymptomatic, are often present in patients with neurodegenerative dementia [14]. Further, microstructural changes that could potentially be re-lated to clinical findings have been found in white matter close to the ventri-cles in patients with idiopathic normal pressure hydrocephalus (iNPH) [15]. Figure 4 illustrates the anisotropy and how it is associated to the diffusion tensor.

20

Figure 4. Principal sketch of anisotropy and the diffusion tensor. The tensor is de-fined by the three eigenvalues, l1 l2, and l3, among which l1 is the largest one. The FA value is a measure of the degree of anisotropy. It is calculated from the three ei-genvalues and has values between 0 and 1, where 0 is isotropic diffusion and 1 is to-tal anisotropy in one direction (along l1). In the intact white matter, the anisotropy is high due to the myelin barriers, but FA decreases if the axons are structural damaged or degenerated myelin has resulted in losses of barriers.

Tractography DTI is a sensitive method for detecting alterations in tissue microstructure and architecture. It also enables tractography, a reconstruction method for visual-ization and quantitative measures of the white matter tracts.

Deterministic streamline tractography is one of the most commonly used tractography techniques in the clinical setting. With this technique streamlines are reconstructed by exploiting the principal diffusion direction of each voxel in order to connect the trajectories. A detailed explanation of the various as-pects of this technique is given in next paragraph. The resulting three-dimen-sional representation visualizes the shape and topology of the tracts.

Deterministic tractography Based on the statement that water molecules to a greater extent move along the axons than they move across the axons, tractography algorithms rely on one fundamental assumption: that when several axons are aligned along a common axis, the diffusion of water molecules will be hindered to a greater extent across this axis than along it, see Figure 3. This assumption means that even though we measure diffusion in many different directions, we expect the algorithm to reconstruct streamlines along the axonal directions. Streamline tractography algorithms integrate fiber orientations voxel-wise, and couple

21

these from voxel to voxel to construct a streamline. A streamline can be de-fined as a line along which the flow of a moving fluid is least turbulent. The tangent of a streamline is always parallel to the vector field, which in this case is equal to the direction of the fiber. The tangent is equal to the first eigenvec-tor of the diffusion tensor such as described in Equation 15.

Equation 15 𝒕(𝑠) = 𝑒S(𝒓(𝑠))

where r(s) is the current [x, y, z] location at the distance s along the stream-line. Further, the evolution of a streamline can be described according to the differential equation [16] in Equation 16.

Equation 16 %𝒓(c)%c

= 𝜀S(𝒓(𝑠))

All tractography algorithms have in common that they aim to track paths where the diffusion is least hindered, i.e. maximized. Strategies used for achieving this differ between algorithms, and many different tractography techniques have been developed over time [17].

Probabilistic tractography Another method to perform tractography is with the probabilistic technique. The first step when performing this technique is to build up a function for characterization of the key measurement, i.e. the fiber orientations. The func-tion is called an orientation density function (ODF) and describes the proba-bility density in relation to the true fiber orientation [17]. Instead of quitting the streamline reconstruction where a possible error takes place, the probabil-istic tractography aims to develop a full representation of the uncertainty that is associated with formulated statements of the algorithm. Tractography re-sults aim to be presented with a statement of the confidence level at which the tract represents the least hindrance to diffusion from one region to another, given the certain model and the data [17]. Probabilistic tractography is thereby able to track regions of high uncertainty where the deterministic tractography techniques fail, because the thresholds reached force the reconstruction to quit. This difference offers the probabilistic tractography a higher possibility to track more difficult regions, i.e. full tracts visible even if a region of the tract has a high uncertainty [18]. The user is able to do more quantifications and comparisons of the confidence, which makes it a flexible method when com-pared to the deterministic tractography method. On the other hand, this tech-nique requires that conclusions from the results must be drawn with lower

22

confidence for all anatomical locations within and beyond this uncertain re-gion.

Clinical applications of tractography The tractography technique is often used to extract and isolate white matter tracts according to a priori neuroanatomical knowledge about the fiber bun-dles in the brain. Initially, trajectories are created as streamlines between voxels with anatomical criterias set as thresholds. Regions of interest (ROIs) are drawn for the streamlines to pass through in order to define the actual tract of interest. In this manner, major white matter tracts can be visualized and analyzed. The technique can be evaluated and validated by comparing the re-sults obtained visually and qualitatively with those obtained in post-mortem dissection studies [19, 20]. For neurosurgeons, it has become more widely used as a tool for preoperative planning of surgery and the visualization and localization of white matter tracts [21].

Diffusion in white matter Water is the major component of the white matter, located both intracellular and extracellular. White matter also contains tight packed axons that make the extracellular space small in comparison to the intracellular space, see Figure 5 for a principal sketch illustrating an axon and its myelin sheaths [17].

When diffusional molecules are studied in anatomical tissues, sampled dif-fusion reflects many different local environments because of the different types and numbers of barriers, for example, the diffusion of water molecules in neural tissue is affected by the neuroanatomy itself due to several factors. The influences on the diffusion processes due to cell membranes and macro-molecules make the final displacement different from what it would have been if the diffusion had been free and incessant [22]. Intracellular water molecules may be hindered by the cytoskeleton and restricted by the cellular membranes, while the biological cells within the tissue may hinder extracellular water mol-ecules. By understanding these underlying microstructural processes, it may be possible to obtain important and valuable information about the biological microstructure by observing the diffusion of water molecules under certain circumstances. The measured signal from a diffusion MRI experiment con-tains collected contributions from the motions of water molecules in all the various environments, and it becomes possible to obtain valuable information about the biological microstructure.

23

Figure 5. A schematic sketch of some elements of the white matter, which represents a neural cell body with its short dendrites and one long axon. The axon is sur-rounded by myelin sheaths, each one of them wrapped around the axon in layers. These are called internodes and are separated from each other by small amyelinated regions called the Nodes of Ranvier.

The anisotropic phenomena in white matter were observed for the first time in 1990 during an MRI examination of the central nervous system in a cat brain. In this experiment, the displacement of the water molecules appeared larger along the axons than across the axons, i.e. the water molecules moved further along the axons than across them during the same amount of diffusion time [5].

It should be kept in mind that diffusion anisotropy is not a specific and unique property of white matter, since it is a physical phenomenon that can occur in other tissues as well, such as muscle tissue, kidney, myocardium, or in liquid crystals. Still, it has undoubtedly provided the basis for a large num-ber of novel measurements with many possible applications regarding inves-tigations of white matter tracts and the neuroanatomy[23].

24

Investigated white matter tracts Corpus callosum The corpus callosum (CC) is a C-shaped inter-hemispheric bundle of commis-sural fibers located beneath the cerebral cortex in the brains of the placental mammals. It connects the left and right cerebral hemispheres. This is the larg-est white matter tract in the human brain, responsible for inter-hemispherical communication. It can be subdivided into four regions: the rostrum, the genu, the body, and the splenium [24]. Structural defects in the CC, arising during development, may result in neuro-physical dysfunctions, and defects of the CC have been observed in neuropsychiatric disorders [25].

The maturation of the CC occurs differently in different regions. The mat-uration of the genu occurs earliest, hereafter the development continues pos-teriorly along the body to the splenium. Rostrum is the last region to mature. The FA can be used to study the age peaks of maturation; since the FA in-creases during the development, when the myelin is formed, and decreases during aging [14, 25]. When the FA value is used as an age peak for matura-tion, three different age peaks are obtained for the corpus callosum, depending on which region is measured. In a study by Lebel et al., the age peak in the genu corpus callosum was reached at the age of 20 y, the peak age of the splenium corpus callosum was reached at the age of 25 y, and the body corpus callosum reached its age peak at the age of 35 y. When the MD value was analyzed in the same study, the age peaks within the corpus callosum when the minimum values of the MD were reached were obtained at the same age for the genu, splenium, and the body of the corpus callosum, at approximately 30 y. Still, the rates of changes and the quantitative values of the MD were different in the three regions of the corpus callosum [14].

Figure 6. Example of the corpus callosum (CC) from a sagittal view; exctracted us-ing anatomical dissection in a post-mortem human brain (to the left) and DTI trac-tography overlaid on a morphological T2 weighted image (to the right). In this post mortem image, only half of the CC is shown, since it is the midsagittal slice pre-sented.

25

Inferior fronto-occipital fasciculus The inferior fronto-occipital fasciculus (IFO) belongs to the association fibers and may be affected in neurodegenerative disorders. Several previous studies have suggested that the IFO plays an important role in reading, attention, and visual processing [26]. The IFO is crucial for the ventral intra-hemispheric transfer of information between the frontal cortex and the occipital, temporal, and parietal cortices because it connects the frontal lobe, the temporal lobe, and the occipital lobe [27]. This is a major white matter tract, with termina-tions to different regions of the cortex, and may be affected to a varying extent in different segments [28]. According to Lebel et al., this tract reaches its age peak of the FA at approximately the age of 20 y, and the minimum value of the MD occurred about 5 years later at the age of 25 y [14].

Figure 7. Example of the large major tract inferior frontooccipital fasciculus (IFO) from a sagittal view; exctracted using anatomical dissection in a post-mortem human brain (to the left) and DTI tractography overlaid on a morphological T2 weighted image (to the right).

26

Cingulum bundle The cingulum bundle (CB) belongs to the limbic fibers and connects the frontal lobe, the parietal lobe, and the temporal lobe in the form of an arch, which extends from the rostral subcallosal area anteriorly and then follows the curved superior surface of the CC on the sagittal plane. Already in 1995, it was suggested that the cingulate cortex might be subdivided into the anterior, mid, posterior, and retrosplenial cortices [29]. After the development of the DTI tractography technique, it was further suggested that the CB is a complex network that contains both long fibers and short association fibers [26]. Jones et al. presented a subdivision of the CB corresponding to the parahippocampal, retrosplenial, and subgenual portions, producing a topographic arrangement of the CB fibers [30]. Previous studies have demonstrated that the CB is in-volved in core processes such as executive function, decision-making, and emotion processing [24]. The CB has been analyzed in association with inves-tigations of conditions such as normal aging, mild cognitive impairment, Alz-heimer’s disease, schizophrenia, and depression [30]. In a study by Lebel et al., the age peak of the FA and the minimum MD were reached at approxi-mately the same age of 40 y [14].

Figure 8. Example of the cingulate bundle (CB) from a sagittal view; exctracted us-ing anatomical dissection in a post-mortem human brain (to the left) and DTI trac-tography overlaid on a morphological T2 weighted image (to the right).

27

Inferior longitudinal fasciculus Several studies have mapped the subdivisions of the inferior longitudinal fas-ciculus (ILF) with the aim of linking them to different specific functions [31-34]. This tract connects the occipital cortex with the temporal lobe. Its func-tions have not been completely mapped, but there are suggestions that the ILF can support parallel and bidirectional segregated networks between the occip-ital cortex and the temporal regions. Such a support might result in positive or negative visual symptoms [35]. Neuropsychological syndromes have been as-sociated with defects of this tract, including visual agnosia and amnesia [35]. Other studies have suggested that the ILF is implicated in autism, Asperger’s syndrome, schizophrenia, and alexia [36, 37]. The ILF is one of the major occipital-temporal association tracts. According to Lebel et al. this tract reaches its minimum value of MD quite early, at 25 y, and the age peak of FA already at 20 y [14].

Figure 9. Example of the inferior longitudinal fasciculus (ILF) from a sagittal view; exctracted using anatomical dissection in a post-mortem human brain (to the left) and DTI tractography overlaid on a morphological T2 weighted image (to the right). The tract is pointed out on the post mortem image and marked with blue markers.

28

Corticospinal tract The corticospinal tract (CST) is a descending tract, primarily concerned with motor functions, extending from the cerebral motor cortex, through the ventral midbrain and the pons, down to the medulla oblongata. Here, the CST fibers collect into a compromised bundle, forming a pyramid that has given the CST its other name "the pyramidal tract" [38]. The CST is divided into two parts: anterior and lateral. The anterior corticospinal tract is primarily responsible for gross movement of the trunk and the proximal musculature. The lateral corticospinal tract mostly mediates fine motor movements of the upper and the lower limbs. The CST may be affected in neurological diseases such as iNPH, in which higher FA and AD values have been observed in several stud-ies [39, 40]. The corticospinal tract is a projection fiber in which the age peak of FA and the minimum of MD occur later than in the other projection fibers. The MD reaches its minimum at about the age of 25 y, and a few years later, the FA peak is reached [14].

Figure 10. Example of the corticospinal tract (CST) from a sagittal view; exctracted using anatomical dissection in a post-mortem human brain (to the left) and DTI trac-tography overlaid on a morphological T2 weighted image (to the right). The tract distributes from the brainstem to the motor cortex and the sensory cortex.

29

Superior longitudinal fasciculus The superior longitudinal fasciculus (SLF) is one of the association tracts and can be subdivided into three segments numbered SLF I, SLF II, and SLF III. It connects the frontal, the occipital, the parietal, and the temporal lobes. The SLF is located lateral to the centrum semiovale and distributes from the frontal lobe through the posterior end of the lateral sulcus, either to the occipital lobe or around the putamen to the anterior portions of the temporal lobe. This tract is involved with functions such as visual attention, working memory, spatial memory, information, and transferring of soma sensory information such as language articulation. Several studies have suggested a connection between defects of this tract and language performances like naming difficulties in the elderly [41]. According to Lebel et al., this tract reaches its minimum value of MD at the latest time-point among all association fibers, at the age of 35 y. Its age peak of FA occurred earlier, at approximately 25 to 30 y [14].

Figure 11. Example of the large tract superior longitudinal fasciculus (SLF) from a sagittal view; exctracted using anatomical dissection in a post-mortem human brain (to the left) and DTI tractography overlaid on a morphological T2 weighted image (to the right).

30

Analyses of white matter tracts In many tractography studies, averaged values are analyzed globally over whole tracts [42]. This might lead to a loss of information about regional de-viations in the diffusion parameters, since it ignores the rich anatomical vari-ations in the DTI measures along the tracts, thereby reducing the effectiveness of the tractography technique [14, 43]. In paper I, we presented a method for the analysis of DTI parameters along white matter tracts that showed consid-erable heterogeneity in the absolute values of FA and MD along the tracts.

In aging, subdivision of tracts could help understanding whether matura-tion and aging affect the whole tract equally, or whether there is also a spatial pattern of variation within tracts [43]. This information can then be considered in other studies in which aging effects might otherwise act as confounders. In paper II, we investigated whether age-related effects on DTI parameters are homogenous or not within white matter tracts by studying a large number of healthy controls. We focused on the IFO and the CB. The complex composi-tion of the CB, with an apparent continuity of fibers albeit partly composed of numerous short association fibers, makes it likely that not all parts of the tract exhibit the same aging pattern [30, 44]. We divided the tracts into different segments, enabled tractography analysis of the different parts of the tracts sep-arately, and compared them with each other.

In paper III, the anatomical results from white matter dissection of post-mortem normal human hemispheres, conducted to define the course of the ILF and its subcomponents, were investigated in healthy controls in vivo using DTI and deterministic tractography. The complex hierarchical organization of the ILF fibers may reflect the activity of several functional networks, as demonstrated by non-invasive studies and direct cortical/subcortical stimula-tion [31-34]. The subcomponents of the ILF were studied regarding the ana-tomical trajectories, connectivity patterns, and quantitative DTI measures.

In paper IV, we investigated white matter tracts in the vicinity of the dilated ventricles in iNPH patients and compared the results with clinical findings. Patients with iNPH have increased cerebrospinal fluid (CSF) pulsation with ventricular enlargement that may cause changes in the diffusion parameter values of neighboring white matter tracts. DTI might be able to detect altered microstructure within the white matter in these patients [39]. The symptoms of iNPH may be improved by shunt surgery, but preoperative prediction of the outcome is difficult [45, 46]. Paper IV aimed to add information about how the white matter tracts are affected and recovered in conjunction with enlarged ventricles in these patients. Tractography was used to evaluate the CC, the CST, and the SLF. Gait disturbance test, the mini mental state examination (MMSE) cognitive test, and clinical assessment were performed.

In all of the subjects studied in this thesis, MRI of the brain has been per-formed at one of two different sites using the similar Philips Achieva 3 T MRI scanners. DTI has been performed using a single-shot spin echo sequence with

31

EPI, 60 contiguous slices, voxel size 2x2x2 mm3, diffusion-weighting factors of b = 800 - 1000 s/mm2, and with diffusion encoding along 48 different di-rections. Motion and eddy current corrections of the data have been per-formed, and diffusion parameter maps have been calculated using the Diffu-sion Toolkit and TrackVis (www.trackvis.org) in Paper I and Paper III, in an in-house developed software (Matlab) in Paper II, and in Olea Sphere V3.0 (Olea Medical, La Ciotat, France) in Paper IV. Experienced radiologists have visually evaluated the morphological MRI images of the subjects. In Figure 12, an overview of all white matter tracts analysed in this thesis is presented.

Figure 12. An overview of all the white matter tracts that were extracted in this the-sis. For notice, the SLF was extracted in the opposite hemisphere in comparison to the other tracts. The SLF is shown as light blue in the back. Corpus callosum is shown in red, the IFO in green, the ILF in orange, the CB in pink, and the CST in dark blue.

32

Aims

General aim The aim of this thesis was to optimize analysis methods for diffusion meas-urements in the white matter of the human brain, to make these clinically ap-plicable, and to validate these through application in healthy volunteers and patients.

Specific aims

Study I The purpose of the first study was to develop a method for analysis of diffu-sion parameters along white matter tracts, using spatial normalization based on anatomical landmarks, and to test the applicability of the method in the IFO in patients with progressive supranuclear palsy (PSP) and healthy controls.

Study II The purpose of the second study was to investigate differences in age-related rates-of-changes between segments within white matter tracts, and to study the spatially varying pattern of the aging process within the CB and the IFO.

Study III The purpose of the third study was to establish whether a constant organization of fibers exists among the ILF subcomponents, and to acquire anatomical and quantitative information about each subcomponent obtained from anatomical dissections and in-vivo DTI with tractography.

Study IV The purpose of the fourth study was to investigate DTI parameters in white matter tracts in the vicinity of the dilated ventricles in iNPH patients, relate the results to clinical findings, compare them with those obtained in matched healthy controls, and to correlate these results with clinical findings in the pa-tients.

33

Materials and methods

Study 1 Subjects Thirty-eight healthy controls underwent diffusion tensor MRI and were ran-domly divided into two groups. These groups were analyzed using a devel-oped framework for analysis of diffusion parameters along white matter tracts, and the results were compared between the two groups to optimize the statis-tical sensitivity of the analysis method.

Fifteen PSP patients, recruited from the Memory Clinic and neurology de-partment of Skåne University Hospital, Lund, Sweden, and an equal number of age-matched healthy controls also underwent diffusion tensor MRI and were investigated using the in-house developed framework. The results were compared groupwise.

The statistical sensitivity, optimized in the first step, was defined as a threshold for differences between the patients and the healthy controls. The study was approved by the local ethics committee, and all participants gave informed consent prior to the experiment.

MR imaging The DTI was acquired using slightly different parameters in the different groups. For the 38 healthy controls, the following parameters were used: rep-etition time/echo time (TR/TE) 6.626/77 ms/ms, acquisition time of 5 min 46 s, and diffusion weighting factor (b) of 1000 s/mm2. For the PSP patients and the age-matched healthy controls, the following parameters were used: TR/TE 7.881/90 ms, acquisition time of 6 min 49 s, and a diffusion weighting factor (b) of 800 s/mm2.

Data analysis Diffusion tensor imaging The diffusion tensor MRI data were retrospectively corrected for motion and eddy currents using FSL-FLIRT, offering a global optimization method for a robust affine registration of DTI images of the brain [47].

34

Tractography (parameters) Tensor-based streamline tractography was performed and analyzed using the Diffusion Toolkit and TrackVis. The whole brain was seeded from two ran-domly positioned seeds in each voxel. Streamlines were calculated with an FA threshold of 0.1 and an angular threshold of 45°. The whole brain tractography was visualized in TrackVis, where navigation was possible in the coronal, transversal, and sagittal planes. TrackVis was used for defining the anatomic regions, and three ROIs were drawn in the coronal plane to extract the IFO. Anatomical landmarks along the IFO were needed to perform the spatial nor-malization. These were drawn in TrackVis, exactly in the same way as the ROIs, but of only one single voxel element size each.

The IFO was extracted using three ROIs, all of them drawn in coronal planes as cross sections to the direction of the IFO. The first ROI was defined around the anterior floor of the external capsule, with the inferior border where the temporal and frontal branches of the external capsule joined. The second ROI was drawn in a coronal plane in the occipital lobe, with the inferior border in the lingual and fusiform gyrus and the superior border where the splenium joined at the midsagittal line. The anatomical position of the third ROI was chosen as a third of the distance between the frontal and occipital ROIs, along the antero-posterior direction. The anatomical landmarks were drawn in the center of the bundle of streamlines in the same coronal plane as the ROIs.

Algorithm The quantitative tractography evaluation algorithm (QuTE) was constructed in three steps: at first a mean track was generated, then diffusion parameters were projected onto the mean track, and finally the mean tracks were spatially normalized.

Cross sections along the mean track were defined by streamline points with a distance less than 1 mm from a plane with its origin in mi and its normal n = mi–mi-1. To initiate this procedure, the mid-point of the streamline with me-dian length was chosen to define the first cross section, in which the first point m-1 was defined as the center-of-mass in the cross section with its normal vec-tor n defined as its direction. Iteration was then performed by calculations of points mi, where i = 0 to N. After the iteration was terminated, the initial point mi, with i = -1, was discarded and points were relabeled as mj, with j = N–i. Now, the calculation of the mean track could be performed by calculate points from mj=N in the opposite direction. The cross sections never included more than one point per streamline, which always was selected as the point closest to mi. Points with a distance from mi larger than 20 mm were excluded to prevent spurious streamlines from distorting the results. The maximal diame-ter of a cross section was therefore 40 mm. The parameter values were ob-tained by averaging the values at all points in each cross section, yielding vec-tors used as the projected parameter values onto the mean track.

35

The apparent area coefficient (AAC) was calculated by determining the area of a mask at each cross section. First, each point in a cross section was represented by a circle with a radius of 0.5 mm. A radius less than 0.5 mm would be too small, while a radius larger than 0.5 mm would be able to blur the cross section. The smallest possible area that could be obtained was πr2, with r = 0.5 mm. The calculated area of the mask represented the AAC value of each cross section along the mean track.

The anatomical landmarks were used in the normalization of the parameter vectors. The segments of the parameter vectors were linearly interpolated be-tween two anatomical landmarks to a specific segment length. The interpola-tion procedure resulted in vectors of 100 elements, per mean track and subject, which were used for group analysis. Asymmetries of the IFO were hemispher-ically investigated by calculating the asymmetry index (AI) for each analyzed parameter (y) in each position (x) along the IFO as described in Equation 17.

Equation 17 𝐴𝐼(𝑦(𝑥)) = 2 9fgh49ij9fgh^9ij

Statistical analysis From the initial optimization of the statistical sensitivity, a cluster size of two and a local significance level of α = 0.05 were defined as suitable values. These thresholds were applied in the comparison between the PSP patients and the age-matched healthy controls. To reach a statistical power of 0.8 in some of the analyzed parameters, a minimum number of 25 subjects needed to be included in each group.

Study II Subjects Healthy individuals in the age interval 13 to 84 years of age, with no known history of neurological or psychiatric disease or brain injury, were included in this study. In total, 104 males and 153 females participated. All participants had been enrolled as controls in previous DTI studies, either at Lund Univer-sity Hospital or at Uppsala University Hospital. Informed consent had been obtained from all participants aged 15 or older, and from the parents of indi-viduals aged younger than 15 years. All previous studies in which these par-ticipants had been included were approved by the local ethics committees. This study was a multicenter study with retrospective analysis of the cohorts, and was approved by the ethics committee in Uppsala, Sweden, where the project was carried out.

36

MR Imaging DTI was performed with identical protocols and acquisition parameters at both sites. The TE/TR was 77/6626 ms/ms, and the diffusion-weighting factor b = 1000 s/mm2.

Data analysis Diffusion Tensor Imaging The diffusion tensor MRI data were retrospectively corrected for motion and eddy currents using ElastiX, [48]. The diffusion parameter maps were calcu-lated in Matlab, using a linear least squares tensor fitting procedure.

Tractography After the deterministic tractography had been calculated, the FMRIB Software Library (FSL) was used for drawing ROIs in the Montreal Neurological Insti-tute (MNI) template space. The ROIs were then projected back to the native space, where diffusion parameters were acquired, using a nonlinear image reg-istration tool in the FSL software. By registering the FMRIB58 FA template to the FA maps, warp fields were used for the back projections. The ROIs were defined according to previous neurological studies in which white matter tract extraction had been described [26, 49, 50].

The CB was extracted using four paramedian ROIs. The first one was lo-cated superior to the genu of corpus callosum. The second ROI was located superior to the mid body of the corpus callosum. The third was located supe-rior to the splenium, and the fourth was located posteroinferiorly to the sple-nium.

The IFO was extracted using three ROIs. The first one was located in the frontal deep white matter in the lateral part of the frontal lobe. The second ROI was located in the occipital deep white matter parieto-occipitally and lat-eral of the lateral ventricle. The third was located in the external capsule, spe-cifically in the deeper sub insular white matter. The IFO is compressed in the region of the external capsule and the temporal stem, but diverges in the frontal and the occipital lobes.

Segments of white matter tracts New ROIs were drawn to define the borders between the different segments within the tracts: the anterior, the posterior, and the inferior. The CB was di-vided into an anterior, a posterior, and an inferior segment. The IFO was di-vided into three segments of approximately equal lengths, with the first and the last ROIs were used as outer borders. The middle segment of the IFO cap-tured the part of the tract that is compressed compared to the other parts of the IFO. The mean values of the diffusion parameters FA, MD, AD, and RD were calculated within each segment.

37

Statistical analysis By performing multiple linear regressions, acquired values were adjusted for confounders according to Equation 18, where y represents the dependent DTI parameter, x1 the tract volume, x2 the gender, and x3 the site of examination

Equation 18 y= 0 + 1 x1 + 2 x2 + 3 x3

Tract size was defined as the number of voxel containing at least one re-

constructed streamline, multiplied with the voxel volume. The results from the multiple linear regressions were used to correct for the three covariates: tract size, site, and gender. Outliers were also excluded by performing an outlier rejection, in which all data points with residuals above three standard devia-tions were defined as outliers.

To prepare for the statistical analyze, all subjects were divided into five groups depending on their ages. The age intervals were chosen based on the white matter maturation and age peaks of the CB and the IFO [14]. The age categories along with the number of included subjects used are defined in Ta-ble 1.

Table 1. Age categories used for the statistical analysis.

Age intervals min age max age number youngest 13 18 23 young adults 19 40 37 middle aged 41 60 23 elderly 61 84 36

Comparisons of age-related changes were performed between the segments of each tract. The null hypothesis of all age groups to have equal means was tested for by using the analysis of variance (ANOVA), yielding the probability of equal differences between the segments. If a result shows present differ-ences between any segments, this would mean that a spatial heterogeneity is present within the tracts.

Study III Subjects The ILF was investigated in 38 human brains in total. Of these, white matter dissection was performed in 14 post-mortem normal human brains to define the distribution of the ILF and its subcomponents. The remaining 24 brains were investigated in right-handed healthy volunteers using in vivo diffusion

38

tensor MRI and tractography to investigate the ILF with regard to its course, distribution, and subcomponents. The healthy volunteers had been enrolled as healthy controls in previous DTI studies at Uppsala University Hospital. In-formed consent had been obtained from all participants, and the study was approved by the local ethics committee.

Anatomical dissection To establish whether the ILF could be divided into subcomponents according to a constant organization of the fibers, anatomical fiber dissections of the post-mortem human brains were accomplished, first of the lateral surface and then of the ventral surface. The in vivo dissection using tractography was then performed by following the same step-by-step procedure in order to acquire anatomical and quantitative information on each sub-component. The quanti-tative and anatomical results were then discussed in the light of previously published studies regarding functions of the ILF.

Lateral surface The most lateral and superficial portion of the ILF was identified infero-lateral in respect to the temporal portion of the arcuate fasciculus and medial to the ventral occipital fasciculus within the occipital lobe. Further, the ILF stem was identified as the branches originating from the dorsolateral occipital cortex.

Ventral surface After removing the U-fibers, a ventral portion of the ILF was identified. In all dissected brains, it was composed of fibers originating from two distinct areas: the fusiform gyrus and the lingual gyrus.

MR Imaging DTI was performed with TE/TR 77/6626 ms/ms, and the diffusion-weighting factor b = 1000 s/m2.

Data analysis Diffusion tensor imaging The diffusion tensor MRI data were retrospectively corrected for motion and eddy currents using ElastiX, [48]. The diffusion parameter maps were calcu-lated in Matlab, using a linear least squares tensor fitting procedure.

39

Tractography Diffusion tensor tractography was used for virtual in vivo dissection to repro-duce the anatomical findings from the white matter dissection of the post-mor-tem brains. The ILF was extracted following the same step-by-step anatomical criteria as used for the white matter dissections. Initially, generous large ROIs were localized around the anatomic regions for the expected distribution of the ILF. Defining and extracting other, well known, white matter tracts could extract these from the results to create a final, pure and isolated, result of streamlines belonging to the ILF. The DTI parameters were analyzed as mean values of the streamlines in each segment of each subject.

Statistical analysis The statistical analyses included the 24 healthy volunteers. The entire ILF vol-ume was significantly lateralized to the right, and a substantial inter-hemi-spheric symmetry was obtained. There was no statistically significant differ-ence between the DTI parameters analyzed and the subcomponents of the ILF.

Study IV Subjects In total, 12 patients with iNPH were included in the study. All patients had clinical symptoms that included a typical gait disturbance, with or without cognitive impairment, and urgency incontinence. Wide ventricles and tight high convexity sulci had been observed on MRI images in all patients. For comparison, 12 healthy controls, matched by gender and age, and with no self-reported history of neurological or psychiatric disease or brain injury were also included. Controls were examined with MRI only once. The subjects were divided into three groups: the group of healthy controls, the patient group before shunt surgery defined as pre-operative in the article, and the same pa-tient group after shunt surgery defined as post-operative in the article. The study was approved by the local ethics committee in Uppsala, Sweden, and informed consent had been obtained from all participants.

MR imaging Patients diagnosed with iNPH had MR examinations at the time of clinical testing and before and 3 months after the shunt surgery. An equal number of healthy controls, matched by age and gender, underwent MR examination once. Morphological MRI scans of all participants were performed and eval-uated visually by an experienced neuroradiologist.

40

Data analysis Diffusion tensor imaging DTI of the brain was performed with TE/TR 77/6626 ms/ms, and the diffu-sion-weighting factor b = 1000 s/mm2. Motion and eddy current correction of the data was performed retrospectively, and diffusion parameter maps were calculated using Olea Sphere V3.0 (Olea Medical, La Ciotat, France).

Tractography Tractography was used for extracting white matter tracts that were of interest in these patients in relation to their symptoms. The streamline reconstruction was performed using the fiber assignment by continuous tracking algorithm (FACT). This is an algorithm that selects a ROI and uses it as a seed region, from which reconstructed streamlines are propagating in both directions until a bound of the termination criteria is reached. All ROIs were manually se-lected in native space in each subject. The T2-weighted b0 images of the DTI data and the FA maps were used for navigation. The CST, CC, and SLF were extracted.

The ROIs used to extract the white matter tracts were anatomically defined according to information about white matter anatomy described in previous neurological studies [26].

Only the segment of CST that was located lateral to the ventricles was in-cluded in the performed analysis, since the enlarged lateral ventricles predom-inantly affect the straight part of the CST adjacent to the body of the lateral ventricles. The segment was extracted using three including ROIs. The CC was extracted by segmenting the whole tract using one large ROI at the mid-sagittal line, covering all voxels that correspond to the distribution of the CC. The SLF was extracted using three including ROIs, combined with some ROIs that were used to exclude erroneous streamlines. A detailed analysis including the FA, the MD, the axial diffusivity (AD), and the radial diffusivity (RD) was performed. DTI parameters were extracted as mean values of each of the three white matter tracts.

Statistical analyses The DTI measures of the extracted white matter tracts CC, CST, and SLF were compared groupwise for each tract between the healthy controls and the pa-tients.

Differences between pre- and postoperative values were calculated for the DTI parameters in each tract. The difference was positive if values were higher postoperatively.

Pre-operative DTI measures were correlated with the clinical findings be-fore shunt surgery as well as with the changes in clinical symptoms after shunt surgery.

41

One-way analysis of variance (ANOVA) was used as a statistical test for the comparisons of DTI parameters between the groups. A paired t-test was used for comparison between pre-operative and post-operative DTI results. Correlations between pre-operative DTI results and clinical symptoms, as well as between pre-operative DTI results and the changes in clinical symptoms after shunt surgery, were investigated using the two-tailed Spearman´s corre-lation.

42

Results

Study I The FA values were lower in PSP patients than in healthy controls around the middle part of the IFO. The AAC was lower in the PSP patients than in the healthy controls in most parts of the left IFO. A significantly larger hemi-spheric asymmetry of the AAC was found in the PSP patients in comparison to the healthy controls in some regions.

Study II Age-related patterns of changes were found in the CB and the IFO, with the exception of the middle segment of the IFO, in which no significant effects of age could be observed. The FA and the RD had an overall well consistent pattern. A combination of the FA and the RD appeared to a sensitive measure of age-related patterns and differences.

Study III We confirmed that the main structure of the ILF is composed of three constant components reflecting the occipital terminations: the fusiform, the lingual, and the dorsolateral-occipital. ILF volume was significantly lateralized to the right. The distribution and the connectivity of the ILF subcomponents suggest that they play a pivot role in integrating information between different brain regions. The ILF stem was identified and isolated in three dimensions, includ-ing its origins, distributions, and terminations. From the results, the subcom-ponents could be described.

Study IV Differences in DTI parameters were found between healthy controls and pa-tients. The smallest differences were in FA values before and after shunt sur-gery. Using the Spearman´s Rho test, significant correlations between pre-operative DTI and clinical symptoms before shunt surgery were found. Some preoperative DTI measures correlated to the changes in clinical symptoms af-ter shunt surgery. Presence of white matter lesions did not differ significantly between healthy controls and patients.

43

Discussion

Tractography algorithms DTI is the leading method for mapping the neuroanatomic architecture by re-constructing tissue structures in vivo. The technique has matured, and many ideas and expectations have been come forth during the last decade.

Consequently, many software packages and solutions have been devel-oped. Several algorithms have been developed to reconstruct white matter tracts. All tractography algorithms have in common that they aim to compute trajectories, or streamlines, through data where most of the white matter tracts run in parallel. Due to the many options regarding underlying algorithms and methods, quantitative as well as qualitative results differ to some extent be-tween the different software programs [51]. Thus, the tractography algorithm chosen determines which software is be suitable for a given purpose, and the chosen techniques then need to be used consequently. In this thesis, focus has been on the tensor-derived algorithms, even though the probabilistic tractog-raphy method, in which peaks of the orientation distribution functions are used for reconstruction, was mentioned in the introduction for clarification. Deter-ministic tractography, i.e. tensor derived, quits the tracking procedure when thresholds are exceeded. Probabilistic tractography launches tracking with dif-ferent probabilities, which actually means that streamlines are obtained even in regions where no white matter tracts are present anatomically. This requires a further step of investigation to assess the probability to the presence of white matter tracts.

Parameters It is important to keep in mind what is different between analyzed groups. Diffusion MRI data are model based, and the measured data consist of several assumptions about underlying processes and structures. These kinds of model assumptions represent simplifications of reality and neglect certain aspects of the true mechanisms of the data [30]. If the numbers of reconstructed stream-lines are different, the conclusion could be drawn that something is so differ-ent between the two groups of tracts compared that there must have been some impact on the number of streamlines. The reasons why the numbers of recon-structed streamlines differ are not so easy to specify in clinical applications such as in this thesis. A conclusion that can be drawn is that there must have been differences between the tracts in the analyzed groups. Often, this can be information enough when tractography is used in clinical applications or in

44

comparisons between patients and healthy controls. Further, it might some-times be possible to detect such differences by visual study of the trac-tographies, such as in the patients with iNPH that were analysed in study IV. Another example is when the reconstructed tract volume is used as a quanti-tative measure; such as we did in study III. The parameter is described by the number of voxels through which reconstructed streamlines pass. The measure cannot tell us the absolute volume of the analyzed white matter tract, but it can give us information about the spatial regions capacity to reconstruct streamlines in relation to other regions in the same subject, regardless of the reasons behind. A resembling analyse was performed for two dimensions in study I, where the AAC value obtained information about the spatial spread of the tract by working as an appropriate measure of the area of the cross-sections along a tract. In applications of tractography in different groups, such as patients and healthy controls, or healthy controls of different ages, answers need to be found successively, step by step. With the use of standardized meth-ods and consequent measures, interesting information can be obtained both in research and in the clinical setting.

Comparisons of diffusion properties Many pathological studies show reductions of anisotropy, which can be due to the loss of diffusion barriers or the addition of more isotropic water. It has been shown that a myelin reduction of 100% only results in a FA reduction of 15% under certain circumstances [23]. It may be possible that small changes of anisotropy can reflect larger pathological differences than previously thought. [23]. On the other hand, the anisotropy can be affected by other fac-tors than a specific pathological property. For example, one factor can affect the largest eigenvalue, while another factor has effect on the third eigenvalue. Both factors affect the value of FA. As an example, several cellular changes may underlie increases of FA and decreases of MD in the healthy population during development such as axonal packing, membrane permeability, axon diameter, proliferations, changes in water content, and myelination [52]. In-vestigations are needed that provide more detailed insight regarding eigenval-ues. Findings from studies of animal models indicate that the RD is more sen-sitive to myelination changes, while the AD is more sensitive to axonal deg-radation [52]. For many white matter tracts, increases of FA seem to be asso-ciated with decreases in RD, according to a study by Simmonds et al. [53]. Findings regarding AD are more variable, which was suggested by Lebel et al. to be partly due to tract-specific or timing-specific changes in the axonal morphology [52]. Conclusions from DTI parameter evaluations need to be drawn after taking several findings into account, such as different DTI param-eter measures together with visual reviews of all subjects.

Because of the large biological and experimental variabilities, the value of one DTI parameter within a subject at a given time-point may not be informa-tive on its own, especially if the subject is a healthy control, and it is therefore

45

suggested to be interpreted in a context of large cohorts and even better if it is repeated over multiple time-points.

Prevented changes in DTI parameters might differ between different sub-jects due to different processes going on, affecting the white matter in different ways. For example, FA is often increased in white matter tracts close to the ventricles in patients with iNPH, even though it is often decreased in neuro-degenerative diseases. As a consequence, resulting changes of FA can be small and no difference between healthy controls and the patients might be found. If white matter lesions are present close to the analysed white matter tracts, these can contribute to the measured values. In such cases, the FA val-ues should decrease and the MD value increase, and possibly affect the results. In study IV, the results could have been affected in an opposite way, since the FA values were higher in the patients. The results would therefore be less de-tectable in the presence of white matter lesions. In study IV, the obtained white matter lesions did not differ significantly between the healthy controls and the patients. Detected white matter lesions were not found close to the CC, nor to the SLF, and effects on these white matter tracts were hereby not expected. Some of the white matter lesions were found close to the ventricles, and CST might therefore be affected from these. The CST had increased FA values and if the white matter lesions have been affected the results, these would have contributed with a decrease in the measured value. The distribution of the an-alysed white matter tracts in relation to the ventricles are visualized in figure 1 a-i in the manuscript of study IV.

Importance of mapping age and healthy controls Understanding the typical and healthy brain provides a baseline from which to detect and characterize brain abnormalities associated with various neuro-logical and psychiatric disorders, symptoms, and diseases. More knowledge about maturation, development, and aging in human neural tissue is needed. The technique of diffusion MRI is nowadays well suited to study the develop-ment of the white matter, although it is limited by its lack of specificity and other standardized methodological concerns [52]. It is of high advantage to use data from healthy controls to learn more about the human neural tissue and about the architecture of white matter tracts. Healthy controls have there-fore been studied, or included for comparison purposes, in all studies that were performed in this thesis. Findings and new knowledge can be of high im-portance in clinical studies with patient groups included, and there is still more to discover. For example, asymmetry of age-related changes in white matter tracts have been observed, but the reason why remains unclear [52]. In study II, we investigated age-related changes in different segments of white matter tracts, and found these changes to act heterogenic. A deeper knowledge of such heterogeneous processes as we found in study II may increase the accu-racy in interrogations of tractography results.

46

In the pursuit of reducing inter-subject variability, group studies are often chosen that are as homogenous as possible by synchronizing factors like hand-edness, gender, etc. The variability between the subjects can itself contain im-portant information since there may be an enormous variability in character-istics like cognitive and behavioral performance within a population. This var-iability can establish couplings between DTI parameters and subject abilities [52].

Hardware and protocol method Available protocols and sequence choices represents a primarily limitation to the image quality that can be achieved, together with the limitations from the risk of PNS. To obtain the highest quality of the results and ensure the accu-racy of the conclusions, all data used in this thesis are acquired on the same type of scanner with an equal field strength. Several aspects of the pulse se-quence protocol might affect image quality and the parameters that are derived from the data. Researchers must design a protocol while keeping in mind de-sired analysis methods, acquisition time limits, and the technical capabilities of the scanner. To minimize the amount of different error sources from the pulse sequences used, we applied equal DTI sequences and protocols from similar scanners used in the multicenter study II.

Enhanced diffusion sensitivity can be reached with the use of increased b values higher than 2000 ms, but of course such an enhancement requires a compromise. The diffusion sensitivity is increased at the cost of SNR, since the minimum TE threshold gets longer. As high a SNR as possible is needed to calculate diffusion parameter maps of sufficient quality, and to obtain higher SNR, individual directions can be acquired multiple times and aver-aged. For clinically applicable DTI, several limitations are present. The acqui-sition times are limited and can therefore not be used as a compromise, for example, for SNR loss due to increased b values. Therefore, an important fo-cus in the clinical setting should be to find a stable DTI protocol that is robust and sensitive enough to obtain the desired results. b values higher than 2000 s/mm2 can be recommended if certain circumstances are present, such as try-ing to solve crossing fibers with constraint spherical deconvolution, and mul-tiple non-zero b values are necessary to compute certain diffusion metrics like kurtosis. For the studies in this thesis, where tractography of major white mat-ter tracts was performed and MRI examinations applied to patient cohorts and groups of healthy controls, we chose the best methods for our aims. These were considered to be an already well known and used sequence, with a rea-sonable acquisition time, and use of deterministic tractography with algo-rithms that offered sufficient robustness. The tensor-based tractography over-simplifies the human brain by ignoring crossing fibers, but is still sufficient for many purposes, and absolutely sufficient for group comparisons such as performed in this thesis [30, 54]. One important factor to increase the statisti-cal power in our data was a high SNR. In order to obtain a sufficient SNR, we

47

chose suitable b values of 800s/mm2 and 1000s/mm2, and we did not exceed 48 gradient directions.

Analyses along tracts DTI parameters are often analyzed over entire white matter tracts, with the assumption that the whole tract is acting in unison [43]. This might be a suf-ficient method of tractography if, for example, the white matter tracts are vis-ually evaluated regarding their shapes, or in group analyses of major tracts, bearing in mind that regional lesions can drown in the averaged values, and regional effects are not detectable to the same extent. To investigate functions or changes within white matter tracts, an alternative is to divide the tracts into smaller sections for individual analyses, either to compare values between sec-tions or to analyze only the interesting section that is expected to be affected. Different functions and cortex terminations have been observed within many major white matter tracts, and one might take this into account during inves-tigations of DTI parameter values [30, 35, 43, 55]. From study II, we could report different rates-of-changes with age in the CB and the IFO. The ILF can be divided into smaller tracts that connect to different regions of the cortex and seem to have different functions, as was observed in our study III.

Decreases of FA have been observed in the CST after shunt surgery in iNPH patients in several studies [39, 56], and the importance of adding DTI as a noninvasive diagnostic tool is emphasized in a review made by Siasois et al. [57]. The SLF has a high risk of compression by enlargement due to its location superior to the lateral ventricles. It is difficult to interpret DTI param-eters measured within SLF in iNPH patients due to their deviating distribution and anatomy, since it is hard to predict from where findings can be derived. Differences may occur because of affected functions due to the disease or the present symptoms, but they may also occur because of a different path of the trajectory [30]. Regardless of which reasons lie behind the deviating results in SLF in the patients with iNPH, when compared to healthy controls, the differ-ences are present, and we can therefore draw the conclusion that the SLF seems to be somehow affected in this patient group.

We observed a visually detectable deviation in the trajectory in some pa-tients. The SLF was in the shape of a wave, i.e. undulating, in four of the 12 tractographies in the iNPH patients. The undultations may be associated with the CSF pulsations since it followed the shape of the superior border of the enlarged ventricles in these cases [45]. Even though this was an interesting finding, indicating that further studies of this white matter tract in this patient group could be very interesting, it was not included into the manuscript due to the low number of obtained undulating SLF tracts.

High intensity white matter lesions can act as confounders on the tractog-raphy results, as discussed in study IV. First, these may lead to decreases of the FA value and increases of the MD value. Secondly, these can affect the final results if they appear differently in compared groups, which might be the

48

case if patient groups are compared to healthy controls since an increased pres-ence of these is often observed in patients with neurodegenerative diseases. All subjects in this thesis were investigated for white matter lesions in order to minimize this effect. In study IV, the Fazekas grades were compared be-tween the healthy controls and the patients but no significant differences were found, indicating that the white matter lesions would not affect the final results in that project.

49

Conclusions

General conclusion Optimized DTI analyses of white matter tracts can preferably be performed regionally by dividing the tracts into segments. This may improve the utility of this method in clinical applications.

Specific conclusions

Study I Tractography analyses may preferably be performed regionally, such as along the tracts or in different segments of the tracts.

Normalization of white matter tracts can be performed using anatomical landmarks. Together with regional analyses of the white matter tracts, this method could offer promising advantages in studies of patient groups with or without cerebral atrophy.

Study II Spatial differences in age-related changes exist between subdivided segments within white matter tracts. The aging processes within the CB and the IFO vary regionally.

Study III The fibers of the ILF in the occipito-temporal region have a clear, constant and detailed organisation.

The anatomical connectivity pattern, and quantitative differences between the ILF subcomponents, confirm a pivotal role of the ILF in connecting highly specialized visual areas to anterior temporal areas that are important for emo-tional, semantic and visual memory processes.

50

Study IV DTI measures differ significantly between patients with iNPH and healthy controls.

DTI measures of the CC, the CST and the SLF, correlate to changes in clinical symptoms after shunt surgery. DTI could potentially become a sup-portive tool for the prediction of clinical outcomes from shunt surgery in pa-tients with iNPH. Larger studies with higher numbers of participants are needed in order to conclude more specifically how quantitative measures of DTI might be usable as a biomarker in iNPH.

51

Future aspects

General aspects The number of DTI studies has increased enormously during the last 15 years. Many trajectories have been identified and evaluated and their development, maturation, and aging have been observed. Interest in the differences within white matter tracts has increased, and many studies report findings suggesting that some white matter tract regions act differently from other regions, as well as that segments of subdivided tracts differ in how they mature, act and change with age. These important findings may lead to a greater interest in investigat-ing the major white matter tracts with regard to their subcomponents and cou-pling the results to different functions and cortex terminations.

Specific aspects New, more exacting methods to analyze DTI data could yield more insights to the architecture within voxels, which in turn might contribute to higher level of knowledge. These methods are valuable to yield more knowledge on a bi-omedical level, which in turn can be of great benefit in the development of clinical applications.

Future technical developments together with the introduction of higher field strengths and higher SNR will contribute further to extend the applica-tions of DTI to clinical studies. Yielding a higher SNR leads to a flexibility that can be used, for example, to apply increased gradient directions or a higher spatial resolution. Regarding these concerns, the 7T scanner can be an interesting option to evaluate white matter tracts in a more detailed manner. The future may provide a greater number of cross-sectional evaluations. With multimodalities such as PET-MRI scanners, new possibilities are presented regarding the analysis of different techniques within the same examination.

Standardized methods In many cerebral diseases, it is of interest to assess changes in the microstruc-tural integrity of white matter tracts as a biomarker for diagnosis and for as-sessment of treatment response. Evaluations of DTI with tractography can be performed using different methods, and it is desirable to establish standardized methods that can be reproduced in different centers and using different MRI scanners. DTI with tractography is a promising imaging method that poten-

52

tially could be very useful for diagnosis and follow-up in many cerebral dis-orders if standardized measurements in specific white matter tracts can be per-formed.

Diffusion MRI has provided an invaluable contribution and approach to the study of white matter. Tractography has provided extremely valuable insights into the neuroanatomy, which in turn has led to the discovery of new white matter fiber bundles as well as a refinement of what was already known. The inter-individual variability in the connectivity of human brains should be seen as an important discovery regarding the existence of a clear organization.

One thing can easily be said with accuracy, we measure a signal change in the presence of an applied gradient that is sensitive to motion. The challenging task following from this statement is to discover what factors lie behind the measured signal changes.

We must continue to investigate white matter tracts, examine how they function, and find out how they are built up. There exists a lot more to learn about white matter anatomy. The more we learn, the more we understand. From there, we can predict what we can expect. Knowledge about neuroana-tomy and neural function provides a way to evaluate tractography results with greater certainty. This improved knowledge will aid in distinguishing our find-ings between obtained results from the clinical setting from those obtained from technical circumstances.

53

Acknowledgements

This journey began at a train from Groningen to Amsterdam in 2011. A thought, a plan, and a number of creative solutions from creative people, ended up in a collaboration between the three universities; Uppsala Univer-sity, University Medical Center of Groningen, and Lund University. I gratefully acknowledge Uppsala university for financial support of this PhD project within the U4 Ageing Brain Network. I also thank the U4 for very well-arranged Summer schools around the universities in Europe. This work would not have been existed without my main supervisor Elna-Marie Larsson, the most amazing person I have ever met. You are always friendly, caretaking, honest, effective and funny; simply everything a super-visor needs to be, and much more. You were also the one who brought me to Uppsala. I will always be grateful for all support and help from you and An-ders; sleepovers, dinners, and good advices. I followed you from south to north, the future will tell if I follow back south someday.

I also would like to gratefully thank… My assistant supervisors Jimmy Lätt and Danielle van Westen who made the beginning of this journey possible.

Freddy Ståhlberg and all members of the MR Physics group at Lund Univer-sity for introducing me to the world of MRI and research.

Collegies, friends and my roommates in the basement of Radio Physics, Tea Sordia, Suaad Meerkhan, Anders Nilsson and Adnan Bibic. Thank you for talks about nothing and everything, nice company during fika, lunches and friday beers. Those were the times...

All co-authors for good collaboration and exciting projects.

Researchers who contributed with patients and healthy controls for examina-tion with MRI. My supervisors of course, and also Christer Nilsson, Fredrik

54

Åhs, Mats Fredrikson, Hedvig Söderlund, Helgi Schiöth, Johan Virham-mar, Katarina Laurell, and everybody else involved. Without you, I would still collect data…

Francesco Latini for everything you taught me about human brains and neu-roanatomy. Thank you for introducing me to the exciting world of neuroana-tomy “IRL” and sharing your knowledge with me at the department of Medi-cal Cell Biology. I will never forget our bumpy driving through Uppsala’s old city streets with brains and formalin spilling around in the back seat of my car. That car is now sold... Mats Hjortberg (1961 – 2016) at the department of Medical Cell Biology, Uppsala University for giving me the fantastic possibilities to learn a lot from neuroanatomic dissections of post mortem brains. I am honored, may you rest in peace. I am also deeply grateful for all admirable individuals who have generously donated their brains for the benefit of research and science. Britt-Mari Bolinder for being a wonderful college and my mainstay at work. So many people just need you so much. Thank you for sharing your experi-ences and enormous knowledge, and for supporting me at my jousting tourna-ments in the weekends!

Christl Richter Frohm for all help with administrative tasks during the years and for showing me your lovely sewing creations. You are an outstanding tal-ent.

Håkan Pettersson for lots of valuable help with many things. Everything gets so much easier with your assistance.

The painter of the front cover image, Daria “Dasha” Petrachkova Mårtens-son, for sharing your fantastic ability to create amazing paintings, always be-ing helpful and never giving up until everybody are satisfied. May you always keep creating! Also, thanks to Bella, for helping me with the design of the front cover image; you are mentioned further below.

The staff at the surgical sciences who work hard with administration, Siv An-dersson, Karin Johansson, Elin Eriksson. You do a great work.

Everybody in Röntgenkorridoren for talks and laughters, and for teaching me the importance of fika and lunch. Erik Thorse´n, Håkan Lönnestav, Per-Erik Åslund, Niklas Blomqvist, Mitham Jabbar, Åke Marjamäki, Per Al-brektsson, Emanuel Hillberg, Lars Jangland, Sara Bruce, Anna Är-lebrandt, Karolina Strand. From now on, I will do my best to stop working in time for lunch in the fikarum at 12. J

55

All the staff at medical physics, and all the staff at the department of radi-ology, at the University hospital in Uppsala. Combining the research with clin-ical work makes everything even more fun. Tomas Bjerner, for the spread of positive energy! Colleagues and friends at work for nice company and making the overall life at work exiting. Especially Elin Lundström, Andreas Frick, Jonas Eng-ström, among others for pleasant lunches and nice talks now and then.

Martin Ingelsson, Jelmer Kok, Berry Kremer, Remco Renken and every-body else involved in the U4 team, for good advices and for making trips to the Netherlands even more fun. I thank the PhD students within the U4 Ageing Brain Network for all the fun we have had. The secret room at UMCG is very nice! Mark Lubberink, Alexander Englund, Markus Fahlström and Karin Åberg for relieving and supporting me in making the necessary priorities. Al-lowing me the dedicated time in the last intense stage made this work com-plete.

Arvid Morell. If it wasn´t for you, I wouldn´t be where I am today. Random things just happen sometimes. Thank you for nice conference company!

My research collaborators in projects out of this thesis who have been un-derstandable and patiently waited for research support. You know who you are. I am looking forward for many interesting projects to come.

My beloved partner in life, coach, and my best friend Niklas Sternbrink for your never-ending patience, positive energy and support during the last in-tense time. I cannot make a list of everything you have done for me, so I simply just thank you for everything.

My mother Lotta Mårtensson for always being so proud of me. You have always believed, encouraged, and supported me. You are a fighter who have shown me that nothing is impossible and that you can achieve what you want in life, if only you believe in it hard enough, and I love you for that.

My father Nils-Inge Mårtensson. This journey would never have been possi-ble without you. You helped us build up our new lives in Uppsala. You came up with all our stuff, and built us a stable during 4 days in a way only you can. You have taught me all I know about a lot of practical things outside the MR examination room. You are one of a kind and I appreciate and love you for that.

56

My daughter Bella Mårtensson. There are no words strong enough to explain my never-ending love. I carry with me the deepest appreciations for being your mother. You have always been asking questions and shown interest for my work, and never complained on me working at strange time-points. You have always been standing next to me with your love and support and you have given me the most valuable, interesting and fun breaks from writing. You are the nicest and strongest human being I have ever met and I admire you for being you. Thank you from my deepest heart for everything you taught me about life and for all the lovely pastries that kept me going. I love you into eternity, and I will always support and believe in you.

And, of course, thanks to my caring dear friends and everyone else who I forgot to mention here!

57

Sammanfattning på svenska

Diffusion tensor magnetic resonance imaging of the brain: tractography analysis with application in healthy individuals and patients.

Bakgrund, material & metoder Diffusion tensor imaging MRI (MR-DTI) är en MR-teknik som introducera-des år 1994. Tekniken baseras på beräkning av den så kallade diffusionsten-sorn i varje voxel. Detta görs genom att samla in en serie MR-bilder där man mäter diffusionen, d.v.s. vattenmolekylernas spontana rörelser, i olika rikt-ningar, och därefter anpassa en tensormodell till datan. När uppmätt diffusion av vattenmolekyler är oberoende av i vilken riktning mätningen utförts, är dif-fusionen isotrop. Som en effekt av att fibrerna i nervbanorna är tätt ihop-pack-ade och löper intill varandra blir diffusionen mer hindrad vinkelrät mot nerv-banorna än längs nervbanorna. Diffusionen är då anisotrop. I vit hjärnvävnad beskrivs diffusion bäst av en tensormodell.

DTI är en känslig metod för att påvisa förändringar i vävnadens mikro-struktur och anatomiska arkitektur. Metoden möjliggör även traktografi; en rekonstruktionsmetod för att visa förloppet av banorna och göra beräkningar av diffusionen i nervbanor. En av de vanligaste traktografiteknikerna är deter-ministisk traktografi, i vilken banornas sannolika förlopp rekonstrueras genom att följa den huvudsakliga diffusionsriktningen från voxel till voxel och där-med koppla ihop trådar, s.k. streamlines, så att en tredimensionell representat-ion av nervbanornas utformning kan visas.

I många sjukdomar i hjärnan är det av intresse att studera mikrostrukturella

förändringar i nervbanor, att finna biomarkörer för diagnostik och bedöma be-handlingsvar. Kvantitativa utvärderingar av DTI och traktografi med hjälp av standardiserade och reproducerbara metoder är önskvärda. Metoden kan po-tentiellt få stor användning både vid diagnostik och vid uppföljning vid flera olika sjukdomar.

DTI och traktografi kan även användas för att studera mikrostrukturell ut-

veckling, mognad och åldrande. Påverkan och förändringar i den vita hjärn-vävnaden kan påvisas och åldersrelaterade skillnader i olika områden detek-teras med hjälp av olika mätvärden som kan kvantifieras vid DTI- och trakto-grafi-analys. Skillnader i dessa mätvärden har rapporterats i vissa regioner i

58

hjärnan vid jämförelser av strukturer mellan äldre och yngre friska frivilliga individer. Man har hittat kroniska vitvävnadslesioner, symtomatiska eller asymtomatiska, hos patienter med neurodegenerativ demens. Vidare kan mik-rostrukturella förändringar relateras till kliniska fynd.

Många DTI- och traktografistudier analyserar globala medelvärden i hela

nervbanor. Denna typ av analys kan leda till falskt negativa fynd och utebliven information om regionala skillnader i diffusionsegenskaper i nervbanorna.

I artikel 1, presenterar vi en metod för analys av DTI parametrar längs nervbanor. Vid studium av åldrande, kan uppdelning av rekonstruerade nerv-banor i mindre delar, s.k. segment, underlätta vid undersökning av om mognad och åldrande påverkar hela nervbanan på samma sätt eller om det existerar en spatiell variation inte bara mellan olika nervbanor, utan även inom samma nervbana. Denna information kan man sedan ta hänsyn till i kommande stu-dier, i vilka ålderseffekter annars kunde påverka slutresultatet.

I artikel 2, utredde vi om åldersrelaterade effekter på DTI parametrar är homogena inom nervbanor genom att studera ett stort antal friska frivilliga personer. Vi fokuserade på två nervbanor; Inferior Fronto-occipital fasciculus (IFO) och cingulum bundle (CB). Vi delade upp nervbanorna i segment, vilket möjliggjorde traktografianalys av de olika segmenten separat, varefter vi jäm-förde resultaten i de olika segmenten med varandra.

I artikel 3, utförde vi anatomisk dissektion av friska hjärnor post-mortem. Under dissektionen identifierades nervbanan Inferior Longitudinal fasciculus (ILF) och dess delkomponenter samt utbredningen av dessa olika banor som tillsammans bildar ILF. Samma nervbana, ILF, utreddes och analyserades i friska frivilliga personer genom avbildning med DTI och traktografi. ILF är en av de stora associationsbanorna och det finns tidigare studier som kartlagt dess uppdelning i mindre banor samt dess specifika funktionerVi studerade diffusionsegenskaperna i egenskap av olika mätvärden och parametrar i del-komponenterna av ILF.

I artikel 4, undersökte vi nervbanor som förlöper i närheten av vidgade ventriklar i patienter med iNPH och jämförde resultaten med kliniska fynd. Dessa patienter har ökade pulsationer från cerebrospinalvätska (CSF) med för-storade ventriklar som kan orsaka förändringar i diffusionsegenskaperna i när-liggande vit hjärnvävnad. DTI kan vara en användbar metod för att påvisa förändrad mikrostruktur i den vita hjärnvävnaden hos dessa patienter. Symto-men vid iNPH kan minskas av shuntoperation, men den postoperativa progno-sen är svår att bedöma preoperativt. I den här studien, var syftet att med DTI finna mer information om hur nervbanorna påverkats av förstorade ventriklar och av shuntoperation i dessa patienter. Traktografi användes för att utvärdera Corpus callosum (CC), Corticospinal tract (CST), och Superior Longitudinal fasciculus. De kliniska testerna omfattade gångtestet "gait disturbance test", det kognitiva testet "mini mental state examination" och en noggrann klinisk bedömning gjordes. Resultaten från DTI jämfördes mellan patienter med

59

iNPH och friska frivilliga kontroller. Resultat från pre-operativ DTI jämfördes med både pre-operativa och post-operativa kliniska fynd.

Samtliga MR-undersökningar av hjärna i dessa projekt har utförts på

Philips Achieva 3T MR. DTI har utförts med en singleshot spin eko sekvens (SSSE) med ekoplanar imaging (EPI). Olika programvaror har använts för ef-terrekonstruktion och analys av bilderna. Alla morfologiska MR-bilder har granskats av radiolog.

Resultat Artikel 1. Gruppjämförelser av diffusionsparametrar i Inferior fronto occipital fasciculus (IFO), i båda hemisfärer separat, mellan PSP patienter och ålders-matchade friska frivilliga kontroller utfördes för alla positioner längs med hela nervbanan. Skillnader mellan grupperna observerades i vissa regioner och asymmetri i nervbanans tvärsnittsarea mellan de båda hjärnhalvorna var sig-nifikant större i PSP patienter jämfört med friska frivilliga kontroller.

Artikel 2. Åldersrelaterade förändringar skilde sig både mellan olika nerv-banor och inom samma nervbanor. Olika åldersmönster observerades i alla segment av de analyserade nervbanorna, för alla analyserade diffusionspara-metrar. Åldrande påverkar inte vit hjärnvävnad uniformt och homogent, utan varierar regionalt; både mellan olika och inom samma nervbanor.

Artikel 3. Vi bekräftade att huvudstrukturen av ILF består av tre kompo-nenter som indelas efter de occipitala regioner där når cortex; fusiform, lingual och dorsolateral-occipital. Konnektiviteten och fördelningen som observera-des hos ILF och dess subkomponenter spelar en betydande roll för dess funkt-ioner.

Artikel 4. Några av diffusionsparametrarna skilde sig signifikant mellan friska frivilliga och patienter. Korrelationer påvisades mellan pre-operativ DTI och kliniska fynd före shuntoperation. Korrelationer påvisades även mel-lan pre-operativ DTI och kliniska fynd efter shuntoperation.

Konklusion Styrkan i denna studie är deltagarna; stora kohorter, omsorgsfullt utvalda och väl utredda deltagare. Homogeniteten i datainsamlingen är en styrka då all data är insamlad med samma typ av MR-kamera. Vidare, har metoderna till-lämpats på olika patientkohorter och matchade friska kontroller. DTI och trak-tografi har använts tillsammans med andra tekniker som t.ex. anatomisk dis-sektion och kliniska tester. Många traktografitekniker och olika kvantitativa analysmetoder har använts i forskningsprojekt runt om i världen. För att eta-blera standardiserade metoder för kliniska applikationer, behöver vi valide-ringar i både friska frivilliga och patienter. Resultaten från studierna i denna avhandling kan förhoppningsvis bidraga till utveckling och validering av me-toder samt till djupare kunskaper och leder oss därmed ytterligare ett steg mot detta mål.

60

References

1. Purcell, E.M., H.C. Torrey, and R.V. Pound, Resonance Absorption by Nuclear Magnetic Moments in Solid. Physical Review, 1946. 69(1-2): p. 37-38.

2. Bloch, F., Nuclear Induction. Physical Review, 1946. 70(7-8): p. 460-474. 3. Lauterbur, P.C., Image Formation by Induced Local Interactions: Examples

Employing Nuclear Magnetic Resonance. Nature, 1973. 242: p. 190-191. 4. Chilla, G.S., et al., Diffusion weighted magnetic resonance imaging and its recent

trend-a survey. Quant Imaging Med Surg, 2015. 5(3): p. 407-422. 5. Moseley, M.E., et al., Diffusion-weighted MR imaging of acute stroke:

correlation with T2-weighted and magnetic susceptibility-enhanced MR imaging in cats. AJNR Am J Neuroradiol, 1990. 11(3): p. 423-429.

6. Le Bihan, D., et al., Direct and fast detection of neuronal activation in the human brain with diffusion MRI. Proc Natl Acad Sci U S A, 2006. 103(21): p. 8263-8268.

7. Stejskal EO, T.J., Spin diffusion measurements: Spin echoes in the presence of a time-dependent field gradient. J Chem Physics 1965. 42: p. 288-292.

8. Le Bihan, D., et al., Artifacts and pitfalls in diffusion MRI. J Magn Reson Imaging, 2006. 24(3): p. 478-488.

9. Le Bihan, D. and A. Karni, Applications of magnetic resonance imaging to the study of human brain function. Curr Opin Neurobiol, 1995. 5(2): p. 231-237.

10. Beaulieu, C., The basis of anisotropic water diffusion in the nervous system - a technical review. NMR Biomed, 2002. 15(7-8): p. 435-455.

11. Basser, P.J. and D.K. Jones, Diffusion-tensor MRI: theory, experimental design and data analysis - a technical review. NMR Biomed, 2002. 15(7-8): p. 456-467.

12. Wang, Q., et al., Fiber modeling and clustering based on neuroanatomical features. Med Image Comput Comput Assist Interv, 2011. 14(Pt 2): p. 17-24.

13. Fieremans, E., J.H. Jensen, and J.A. Helpern, White matter characterization with diffusional kurtosis imaging. Neuroimage, 2011. 58(1): p. 177-188.

14. Lebel, C., et al., Diffusion tensor imaging of white matter tract evolution over the lifespan. Neuroimage, 2012. 60(1): p. 340-352.

15. Lebel, C., et al., Microstructural maturation of the human brain from childhood to adulthood. Neuroimage, 2008. 40(3): p. 1044-1055.

16. Basser, P.J., et al., In vivo fiber tractography using DT-MRI data. Magn Reson Med, 2000. 44(4): p. 625-632.

17. Johansen-Berg, H., Behrens T, Diffusion MRI From quantitative measurement to in vivo neuroanatomy. 2009: Elsevier.

18. Tournier, J.D., S. Mori, and A. Leemans, Diffusion tensor imaging and beyond. Magn Reson Med, 2011. 65(6): p. 1532-1556.

19. Lawes, I.N., et al., Atlas-based segmentation of white matter tracts of the human brain using diffusion tensor tractography and comparison with classical dissection. Neuroimage, 2008. 39(1): p. 62-79.

20. Assaf, Y., et al., The role of diffusion MRI in neuroscience. NMR Biomed, 2017. epub.

61

21. Jeurissen, B., et al., Diffusion MRI fiber tractography of the brain. NMR Biomed, 2017.

22. Tanner J.E., S.E.O., Restricted Self-Diffusion of Protons in Colloidal Systems by the Pulsed-Gradient, Spin-Echo Mathos. The Journal of Chemical Physics, 1968. 49(4): p. 1768-1777.

23. Beaulieu, C., The Biological Basis of Diffusion Anisotropy, in Diffusion MRI - From quantitative measurement to in vivo neuroanatomy. 2009, Elsevier. p. 105-126.

24. Wu, Y., et al., Segmentation of the Cingulum Bundle in the Human Brain: A New Perspective Based on DSI Tractography and Fiber Dissection Study. Front Neuroanat, 2016. 10: p. 84.

25. Fenlon, L.R. and L.J. Richards, Contralateral targeting of the corpus callosum in normal and pathological brain function. Trends Neurosci, 2015. 38(5): p. 264-272.

26. Catani, M. and M. Thiebaut de Schotten, A diffusion tensor imaging tractography atlas for virtual in vivo dissections. Cortex, 2008. 44(8): p. 1105-1132.

27. Hau, J., et al., Cortical Terminations of the Inferior Fronto-Occipital and Uncinate Fasciculi: Anatomical Stem-Based Virtual Dissection. Front Neuroanat, 2016. 10: p. 58.

28. Sarubbo, S., et al., Frontal terminations for the inferior fronto-occipital fascicle: anatomical dissection, DTI study and functional considerations on a multi-component bundle. Brain Struct Funct, 2013. 218(1): p. 21-37.

29. Vogt, B.A., et al., Human cingulate cortex: surface features, flat maps, and cytoarchitecture. J Comp Neurol, 1995. 359(3): p. 490-506.

30. Jones, D.K., T.R. Knosche, and R. Turner, White matter integrity, fiber count, and other fallacies: the do's and don'ts of diffusion MRI. Neuroimage, 2013. 73: p. 239-254.

31. Duffau, H., The anatomo-functional connectivity of language revisited. New insights provided by electrostimulation and tractography. Neuropsychologia, 2008. 46(4): p. 927-934.

32. Mandonnet, E., P. Gatignol, and H. Duffau, Evidence for an occipito-temporal tract underlying visual recognition in picture naming. Clin Neurol Neurosurg, 2009. 111(7): p. 601-605.

33. Sarubbo, S., et al., Towards a functional atlas of human white matter. Hum Brain Mapp, 2015. 36(8): p. 3117-3136.

34. Zemmoura, I., et al., New insights into the neural network mediating reading processes provided by cortico-subcortical electrical mapping. Hum Brain Mapp, 2015. 36(6): p. 2215-2230.

35. Latini, F., New insights in the limbic modulation of visual inputs: the role of the inferior longitudinal fasciculus and the Li-Am bundle. Neurosurg Rev, 2015. 38(1): p. 179-189; discussion 189-190.

36. Ashtari, M., et al., Disruption of white matter integrity in the inferior longitudinal fasciculus in adolescents with schizophrenia as revealed by fiber tractography. Arch Gen Psychiatry, 2007. 64(11): p. 1270-1280.

37. Epelbaum, S., et al., Pure alexia as a disconnection syndrome: new diffusion imaging evidence for an old concept. Cortex, 2008. 44(8): p. 962-974.

38. Jang, S.H., A review of corticospinal tract location at corona radiata and posterior limb of the internal capsule in human brain. NeuroRehabilitation, 2009. 24(3): p. 279-283.

39. Jurcoane, A., et al., Directional diffusion of corticospinal tract supports therapy decisions in idiopathic normal-pressure hydrocephalus. Neuroradiology, 2014. 56(1): p. 5-13.

62

40. Scheel, M., et al., Diffusion tensor imaging in hydrocephalus--findings before and after shunt surgery. Acta Neurochir (Wien), 2012. 154(9): p. 1699-1706.

41. Madhavan, K.M., et al., Superior longitudinal fasciculus and language functioning in healthy aging. Brain Res, 2014. 1562: p. 11-22.

42. Odonnell LJ, W.C., Golby AJ., Tract-based morphometry for white matter group analysis. Neuroimage 2009. 45: p. 832–844.

43. Colby, J.B., et al., Along-tract statistics allow for enhanced tractography analysis. Neuroimage, 2012. 59(4): p. 3227-3242.

44. Brickman, A.M., et al., Testing the white matter retrogenesis hypothesis of cognitive aging. Neurobiol Aging, 2012. 33(8): p. 1699-1715.

45. Virhammar, J., et al., Preoperative prognostic value of MRI findings in 108 patients with idiopathic normal pressure hydrocephalus. AJNR Am J Neuroradiol, 2014. 35(12): p. 2311-2318.

46. Klinge, P.M., Idiopathic normal pressure hydrocephalus--neurosurgical management of dementia! Med Health R I, 2012. 95(3): p. 86-7.

47. Jenkinson, M. and S. Smith, A global optimisation method for robust affine registration of brain images. Med Image Anal, 2001. 5(2): p. 143-156.

48. Klein, S., et al., elastix: a toolbox for intensity-based medical image registration. IEEE Trans Med Imaging, 2010. 29(1): p. 196-205.

49. Catani, M., et al., Virtual in vivo interactive dissection of white matter fasciculi in the human brain. Neuroimage, 2002. 17(1): p. 77-94.

50. Wakana, S., et al., Fiber tract-based atlas of human white matter anatomy. Radiology, 2004. 230(1): p. 77-87.

51. Christidi, F., et al., Fiber tracking: A qualitative and quantitative comparison between four different software tools on the reconstruction of major white matter tracts. Eur J Radiol Open, 2016. 3: p. 153-161.

52. Lebel, C., S. Treit, and C. Beaulieu, A review of diffusion MRI of typical white matter development from early childhood to young adulthood. NMR Biomed, 2017.

53. Simmonds, D.J., et al., Developmental stages and sex differences of white matter and behavioral development through adolescence: a longitudinal diffusion tensor imaging (DTI) study. Neuroimage, 2014. 92: p. 356-368.

54. Jeurissen, B., et al., Investigating the prevalence of complex fiber configurations in white matter tissue with diffusion magnetic resonance imaging. Hum Brain Mapp, 2013. 34(11): p. 2747-2766.

55. Chen, Z., et al., Effect of normal aging on the structure of marginal division of neostriatum as measured by MR phase imaging and diffusion tensor imaging. J Magn Reson Imaging, 2017. 45(5): p. 1343-1351.

56. Nicot, B., et al., Effect of surgery on periventricular white matter in normal pressure hydrocephalus patients: comparison of two methods of DTI analysis. Acta Radiol, 2014. 55(5): p. 614-621.

57. Siasios, I., et al., The role of diffusion tensor imaging and fractional anisotropy in the evaluation of patients with idiopathic normal pressure hydrocephalus: a literature review. Neurosurg Focus, 2016. 41(3): p. E12.

Acta Universitatis UpsaliensisDigital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Medicine 1394

Editor: The Dean of the Faculty of Medicine

A doctoral dissertation from the Faculty of Medicine, UppsalaUniversity, is usually a summary of a number of papers. A fewcopies of the complete dissertation are kept at major Swedishresearch libraries, while the summary alone is distributedinternationally through the series Digital ComprehensiveSummaries of Uppsala Dissertations from the Faculty ofMedicine. (Prior to January, 2005, the series was publishedunder the title “Comprehensive Summaries of UppsalaDissertations from the Faculty of Medicine”.)

Distribution: publications.uu.seurn:nbn:se:uu:diva-332596

ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2017