Functional MRI and DTI Fiber Tracking in Patients with Gliomas. (Funktionelle MRT und DTI Fiber Tracking bei Patienten mit Gliomen)

Transcription

1 Aus der Abteilung für Neuroradiologie, der Neurochirurgischen Universitätsklinik, der Albert-Ludwigs-Universität Freiburg im Breisgau Functional MRI and DTI Fiber Tracking in Patients with Gliomas (Funktionelle MRT und DTI Fiber Tracking bei Patienten mit Gliomen) INAUGURAL-DISSERTATION Zur Erlangung des Medizinischen Doktorgrades der Medizinischen Fakultät der Albert-Ludwigs-Universität Freiburg im Breisgau Vorgelegt 2011 Von Thao Nguyen Thanh geboren in Hue, Vietnam

2 Dekan Prof. Dr. Dr. h.c. mult. H. E. Blum 1. Gutachter Prof. Dr. I. Mader 2. Gutachter Prof. Dr. F. Hamzei Jahr der Promotion 2011

3 Contents 1. Introduction White matter fibers... 8 The pyramidal tract... 8 Myelinated axon Brain gliomas Epidemiology WHO classification Progress and prognosis Imaging patterns Functional MRI BOLD technique Post processing: SPM Diffusion Tensor Imaging and Fiber Tracking Anisotropic diffusion FACT Probability Maps Global Tracking Brain plasticity Aims of study Methods Patients Paradigms Functional MRI

4 2.2.2 Diffusion Handedness assessment Motor assessment Data processing Functional MRI Fiber tracking Seed regions FACT Probability Maps Global Tracking Image analysis Statistical analysis Results Motor function Functional MRI Fiber tracking Deviation of fibers Change of FA value Discussion Reorganization of brain function Limitation of fmri Effect of tumor on white matter fibers

5 4.4 Comparison of the tracking abilities of FACT, Probability Maps and Global Tracking FA stop criteria The maximal allowed curvature Branching, merging and connecting fibers Fiber depiction Limitation of tractography Lack of verification Seed points Clinical aspect and outlook Conclusion/Zusammenfassung Conclusion Zusammenfassung References Acknowledgements Curriculum Vitae

6 Tables, Diagram and Figures Table 1.1 WHO classification of glial tumors...11 Table 3.1. Clinical characteristics and description of fmri and DTI...36 Diagram 3.1. Motor impairment of WHO grade II and III gliomas...31 Figure 1.1 Dissection of the corticospinal tract...10 Figure.1.2 Neurovascular coupling...14 Figure 1.3 A 3D-view of Brownian motion...16 Figure 1.4 The relationship between anisotropic diffusion, diffusion ellipsoids, and diffusion tensor...17 Figure 1.5 Principle of FACT...18 Figure 1.6 Probability Maps: merging and connecting fibers...20 Figure 1.7 Principle of the Global Tracking method...21 Figure 3.1 fmri activation of patient # Figure 3.2 fmri activation of patient # Figure 3.3 fmri activation of patient # Figure 3.4 fmri activation of patient # Figure 3.5 Correlation between the ID and grade of motor impairment 37 Figure 3.6 Correlation between the FA asymmetry and grade of motor impairment...38 Figure 3.7 Patient # Figure 3.8 Patient # Figure 3.9 Patient # Figure 3.10 Patient # Figure 3.11 Patient # Figure 3.12 Patient # Figure 3.13 Patient # Figure 3.14 Patient #

7 Figure 3.15 Patient # Figure 3.16 Patient # Figure 3.17 Patient # Figure 3.18 Patient # Figure 3.19 Patient #

8 List of Abbreviations ADC BOLD CBF CBV CNS CST DTI FA FACT fmri Hb HGG ID LGG LTD LTP MEG ODF oxy-hb PET ROI SMA SPM TMS WHO Apparent Diffusion Coefficient Blood Oxygenation Level Dependent Cerebral Blood Flow Cerebral Blood Volume Central Nervous System Corticospinal Tract Diffusion Tensor Imaging Fractional Anisotropy Fiber Assignment by Continuous Tracking Functional Magnetic Resonance Imaging Hemoglobin High Grade Glioma Index of Deviation Low Grade Glioma Long-term Depression Long-term Potentiation Magnetoencephalography Orientation Distribution Function Oxygenated Hemoglobin Positron Emission Tomography Region Of Interest Supplementary Motor Area Statistical Parametric Mapping Transcranial Magnetic Stimulation World Health Organization 6

9 1. Introduction In the treatment of gliomas it is essential to consider the anatomy of eloquent cortex areas and of functionally critical fiber bundles, such as e.g. the corticospinal tract (CST). They can be assessed by functional magnetic resonance imaging (fmri) and by fiber tracking. Fiber tracking can be divided into two groups: local, either deterministic (Fiber Assignment by Continuous Tracking - FACT) or probabilistic (Probability Maps), and global (Global Tracking). FACT (Mori et al. 1999) has been shown to be reliable for glioma and cavernoma resection (Nimsky et al. 2006; Nimsky et al. 2006) but is susceptible to low fractional anisotropy (FA) (Yamada et al. 2003). Probabilistic fiber tracking is less susceptible to these limitations (Kreher et al. 2008), but not clinically proven as far as the FACT. Global Tracking is a newly established tracking method that takes all detectable fibers of the whole brain into account (Kreher et al. 2008; Reisert et al. 2010). This method is time consuming and lacks of clinical proof. An improvement concerning the computational time has been made by Reisert (Reisert, Mader et al. 2010), and so far first clinical applications became possible (Mader et al. 2010). 7

10 1.1 White matter fibers White matter fibers (WMF) contain neuron s axons that link cortical areas with each other and with subcortical centers. WMF are categorized into five groups according to Schmahmann (Schmahmann et al. 2006; Schmahmann et al. 2008) and are identified as follows. 1. Association fibers that connect ipsilateral cortical regions. Local association fibers, or U-fibers, connect adjacent gyri. Neighborhood association fibers connect nearby regions. Long association fibers connect distant areas within the same hemisphere. 2. Striatal fibers that descend to the basal ganglia. They are the Muratoff bundle which contains axons from the limbic system to the striatum and the external capsule that conveys axons from primary sensorimotor cortices to the putamen and axons from the supplementary motor area (SMA) to the caudate nucleus. 3. Commissural fibers that travel to the contralateral hemisphere. They are anterior commissure, corpus callosum and hippocampal commissures. 4. Thalamic fibers are subcortical fibers that travel to the thalamus through the internal capsule (anterior and posterior limbs) and the sagittal stratum. 5. Pontine fibers are subcortical bundle that share the same path with thalamic fibers within the internal capsule and the sagittal stratum. Their destinations are brain stem and spinal cord. The pyramidal tract The pyramidal tract, also known as the corticospinal tract, is responsible for the voluntary movements of the body. By its means the cortex controls the 8

11 subcortical motor centers. It has been described as a brain highway used by a variety of cortifugal fiber traffic (Kuypers 1987). From the motor cortex of the frontal lobe, the pyramidal tract descends through the white matter of the centrum semiovale, and then through the posterior limb of the internal capsule (Jang 2009). It continues through the rostral part of the cerebral peduncle. These fibers pass the rostral portion of pons before entering the pyramids of the medulla oblongata (Schmahmann et al. 2004). Most of the fibers decussate at this level (at the pyramidal decussation) and then form the lateral corticospinal tract of the spinal cord, the other fibers that don t decussate will form the anterior corticospinal cord (ten Donkelaar et al. 2004). All these fibers then synapse with the neurons within spinal anterior horn (Grossman et al. 2003). Myelinated axon The neuron is the functional unit of the brain. It consists of a cell body, an axon and several dendrites. The brain white matter consists mostly of myelinated axons (Paus 2010). The axon may be very long and is cylindrical. The diameter of axon varies according to its function. The myelin sheath of central nervous system (CNS) axons, a proteophospholipid complex, is produced by the oligodendrocytes (Piaton et al. 2010). The main function of myelin sheath is to increase the speed of signal conduction along the axon by establishing a saltatory excitation (Hartline et al. 2007). The highly aligned myelinated fibers restrict water diffusion perpendicular to the fiber orientation, resulting in anisotropic diffusion. Any damage to the myelin sheath can lead to an alteration in water diffusity that can be detected by diffusion imaging techniques (Song et al. 2003; Nair et al. 2005; Concha et al. 2006; Moeller et al. 2007). 9

12 Figure 1.1 Dissection of the corticospinal tract (Ludwig et al. 1956). 1.2 Brain gliomas Epidemiology The incidence of primary brain tumors is about per persons (DeAngelis 2001). The average age is 54 years, males are more common than females (Wrensch et al. 2002). The incidence rate in children < 15 year-old is 2.6 per children with the most common tumors being astrocytomas ( 41.7%), medulloblastomas (18.1%) and ependymomas (10.4) (Kaatsch et al. 2001). About 15-25% percents of gliomas are low grade gliomas (LGG). They usually affect young adults (Walker et al. 2003). Two thirds of LGG are supratentorial (1/3 frontal, 1/3 temporal). One third is infratentorial (brainstem, cerebellum, and pons) (Osborn 1993). LGG is often situated within eloquent regions (Duffau et al. 2004). 10

13 1.2.2 WHO classification According to the 2000 WHO Classification of nervous system tumors (Kleihues et al. 2002), glial tumors are classified into two main groups: astrocytic and oligodendroglial (Table 1). Astrocytomas are categorized into 3 main groups: (1) Diffuse astrocytoma (WHO grade II) which consists of fibrillary astrocytoma, protoplasmic astrocytoma and gemistocytic astrocytoma. (2) Anaplastic astrocytoma (WHO grade III). (3) Glioblastoma (WHO grade IV), which consists of giant cell glioblastoma and gliosarcoma. Table1.1 WHO classification of glial tumors Astrocytic tumors Oligodendroglial tumors Diffuse astrocytoma Oligodendroglioma Fibrillary astrocytoma Anaplastic oligoastrocytoma Protoplasmic astrocytoma Gemistocytic astrocytoma Anaplastic astrocytoma Mixed gliomas Glioblastoma Oligoastrocytoma Giant cell glioblastoma Anaplastic oligoastrocytoma Gliosarcoma Pilocytic astrocytoma Pleomorphic xanthoastrocytoma Subependymal giant cell astrocytoma 11

14 Progress and prognosis Operation is the treatment of choice for LGG (Duffau 2009). Untreated grade II gliomas have a constant growth over time with the average size increase 4.1 mm in diameter per year (Mandonnet et al. 2003). The overall 5-year survival ranges from 27% to 85% of patients (Wessels et al. 2003). Patients who underwent gross tumor resection likely to have a longer survival period than those who underwent subtotal tumor resection (Johannesen et al. 2003) Imaging patterns LGG are infiltrative tumors with slow growing pattern. Despite circumscribed appearance on imaging studies they usually infiltrate the neighbouring tissue (tumor s cells are found beyond the borders of abnormal signal). They have homogeneous pattern without contrast enhancement (Osborn et al. 2004). Computed tomography WHO grade II gliomas have low density on native CT in comparison to brain parenchyma. Tumors have blurred boundary. Hemorrhage and cystic change are seldom. Calcification occurs in 20%. There is no enhancement after contrast administration (Osborn 1993). Magnetic resonance imaging WHO grade II gliomas are hypo-/isointense on T1 and iso-/hyperintense on T2. Surrounding edema is seen on T2 and FLAIR. Contrast enhancement is rare. Decrease of ADC can bee seen on diffusion MRI (Osborn 1993; Grossman and Yousem 2003). MR Spectroscopy High Cholin and Cholin/Creatine ratio, low NAA-concentration. High Myo- Inositol/Creatine ratio (0.82 +/- 0.25) (Osborn, Blaser et al. 2004). 12

15 1.3 Functional MRI Blood-Oxygenation-Level-Dependent (BOLD) technique Functional magnetic resonance imaging is a modern, well-developed technique of cognitive science. The most important advantages of fmri are the ability to investigate the brain function without application of X-ray and the ability to be repeated. The BOLD technique using the level of vascular oxygen as an intrinsic contrast is a great development (Ogawa et al. 1990; Ogawa et al. 1990). The principle of this technique bases on the difference between the magnetic properties of deoxygenated hemoglobin (Hb) and oxygenated hemoglobin (oxy-hb). This difference can be as much as 20% for completely oxygenated and completely deoxygenated blood" (Pauling et al. 1936). Hb, which contains unpaired electrons, is paramagnetic, therefore reduces the local magnetic homogeneity, which will result in the reduction of MR signal. On the other hand, Oxy-Hb, which contains no unpaired electrons, is diamagnetic, which means that it doesn t interfere much with the local magnetic field (Pauling and Coryell 1936). There is a strong relationship between neuronal activity and energy metabolism (Kennedy et al. 1976). When activated, neurons consume more oxygen than in the rest state (about 5%) (Fox et al. 1986). Due to the delay of vascular system response to the neuronal activation, the level of Hb increases leading to the reduction of MR signal (the initial dip) (Narayan et al. 1995). The delayed response of the vascular system to the neuronal activation results in a chain of changes in the local vessels to supply oxygen and energy to the activated areas. There is an increase in regional blood volume (Frostig et al. 1990). This volume s increase peaks s after stimulation s onset (Fox and Raichle 1986; Narayan, Esfahani et al. 1995). As well as the cerebral blood 13

16 volume (CBV), the cerebral blood flow (CBF) also increases. This increase is much higher than the increase of oxygen metabolism and spread over distances of 3-5 mm (Fox and Raichle 1986; Narayan, Esfahani et al. 1995; Malonek et al. 1996; Hoge et al. 1999). The increase of CBF is so strong that it does not only compensate the reduction of intravascular oxygen level, but also oversupplies this area and the vicinity and washes out the Hb. The result is the increase of MR signal. Simultaneous with the CBF increase, the arterial diameter also increases (Ngai et al. 1995). After stimulation is removed, CBV returns to the normal state at a lower rate than that of CBF (Jones et al. 1998). This delayed CBV normalization, as well as the prolonged oxygenation consumption in brain tissue, result in a reduction of vascular Oxy-Hb concentration, producing a post stimulation signal undershoot (Schroeter et al. 2006). Figure.1.2 Neurovascular coupling. After a short initial dip, the fmri signal increases and reaches the maximum level after 5-6 s. The signal reduces to the baseline after about 10s (Stippich 2007). 14

17 1.3.2 Post processing: Statistical parametric mapping (SPM) Functional MR images are processed using SPM which is free software package widely used in neuroscience. It was developed by the Wellcome Department of Imaging Neuroscience at University College London ( Its methodology and concept was first described by Friston (Friston et al. 1990; Friston et al. 1991). It consists of several options to analyze and test hypothesis about brain function. It is a voxel-based approach in which the images are (1) realigned to minimize the differences between successive scans using an affine rigid-body transformation. The realigned images are (2) spatially normalized into a standard space using Tailarach coordinates (Tailarach et al. 1988) and then (3) spatially smoothed. A General Linear Model is used to describe the data in terms of experimental effects. To test experimental hypothesis, different classical statistical inferences are used. SPM 94 was the first version of the SPM software which was written mainly by Friston in The latest version is SPM8 which was released in April

18 1.4 Diffusion Tensor Imaging and Fiber Tracking DTI is a MRI technique to measure the anisotropic diffusion of water molecules in tissue in order to produce neural tract images. It provides information about directional structure of white matter fiber Anisotropic diffusion In a homogenous environment, the thermodynamic effects cause particles to diffuse randomly and equally in all directions. This type of motion was first observed in 1827 by the British botanist Robert Brown, hence called Brownian motion. In living tissues such as muscle, however, there are structures that can restraint the molecular diffusion (Tanner 1979; Scollan et al. 1998). The white matter is composed of highly aligned fibers which can cause restriction of molecular diffusion. As a result, water molecules tend to diffuse along fiber s axis rather than perpendicularly to it (Moseley et al. 1990). Figure 1.3 A 3D-view of Brownian motion. Source Wikipedia.org 16

19 Figure 1.4 The relationship between diffusion (upper row), diffusion ellipsoids (middle row), and diffusion tensor (bottom row). In an isotropic condition (column a), water diffuses equally in all directions. The diffusion ellipsoid is therefore spherical and can be represented by one diffusion constant, D. In anisotropic environment (column b, c), water diffusion has directionality and therefore the diffusion ellipsoid is elongated. In column b, the anisotropic diffusion is mainly aligned along z axis and can be represented by three diffusion constants, λ 1, λ 2 and λ 3. In column c, the anisotropic diffusion is not aligned to any main axis and is represented by nine diffusion constants (Mori et al. 1999). 17

20 1.4.2 FACT Fiber assignment using continuous tracking (FACT) was introduced in 1999 by Mori and colleagues. This technique is proved to be superior to its previous method using a discrete number field which can lead to a deviation of actual fiber (Mori, Crain et al. 1999). Given that water is restricted in the direction perpendicular to the axons and diffuses preferentially parallel to them, this phenomenon can be mathematically represented by an ellipsoid characterized by three diffusion constant along its three orthogonal axes (Figure 1.5 a and b). The tracking propagates on the basis of the direction of the eigenvector associate with the largest FA value. Figure 1.5 Principle of FACT. A and b: anisotropic diffusion of water and the ellipsoid with three eigenvector on its three principal axes; c and d: tracking with discrete (c) and continuous (d) number field. The actual fibers are indicated by curved arrows whereas average fiber direction in voxel is indicated by open arrows; e: tracking is continued as long as nearby vectors are strongly aligned. When vector orientation becomes random (f), tracking is stopped (Mori et al. 1999). 18

21 FACT is proven to be reliable in resection of gliomas and cavernomas (Nimsky, Ganslandt et al. 2006; Nimsky, Ganslandt et al. 2006). However, it seems to be susceptible to fiber crossing in which the crossing fibers can lead to less anisotropic or even isotropic conditions (Basser et al. 2000). When FACT runs through such areas, the tracking process can be terminated due to its lack of ability to distinguish those crossing fibers. The same problem arises when using FACT in individuals who have brain abnormalities such as brain tumors which can cause fiber deviation. FACT is also proved to be vulnerable to the low FA-areas caused by infiltrative tumors such as low grade gliomas (Mori et al. 2002) Probability Maps Probability Maps are a new approach presented by Parker in 2003 (Parker et al. 2003). Different to the deterministic approach of FACT, a kind of mathematical connectivity from a seed region to a voxel is calculated. By running the streamline process repeatedly using Mote Carlo methods, multiple trajectories are built. The probability is calculated by the frequency such random walks pass through the chosen voxel. To overcome the fiber crossing problem, Kreher (Kreher, Schnell et al. 2008) suggested the connecting map. They are fibers which connect two seed regions. Two visiting maps starting from those regions are calculated. These two visiting maps are then combined to produce a connecting map that includes all voxels that may connect to both seed regions. Connected area is determined when two fibers (or group of fibers) face opposite directions. Merge area is determined when two fibers meet each other and face the same direction (Figure 1.8). 19

22 Figure 1.6 Probability Maps: merging and connecting fibers. Yellow area is defined as connected area as two groups of fibers face opposite directions. Magenta area is determined as merged area as these fibers meet and face the same direction (Kreher et al. 2008). Probabilistic fiber tracking has been applied in optic neuritis, multiple sclerosis, stroke recovery, epilepsy, Tourette syndrome as well as in surgical planning for deep brain stimulation in lower leg stump pain (Ciccarelli et al. 2005; Owen et al. 2007; Rodrigo et al. 2008; Hu et al. 2009; Makki et al. 2009; Pannek et al. 2009). 20

23 Global Tracking The principal of Global Tracking was first introduced by Mangin (Mangin et al. 2002). Its method was further developed by different authors (Kreher, Mader et al. 2008; Fillard et al. 2009). The basic concept is, in contrast to the conventional walking methods, several line elements representing water diffusion are constructed simultaneously through the whole volume. These elements are then bound together under specific conditions to form the fibers. This novel method has been said to be superior to the old walking methods as they can overcome the problems of crossing and merging fibers. However, the main limitation of Global Tracking that prevents it from being clinically useful is the extensive computational time. Some efforts have been made to improve its practicality (Reisert et al. 2010) and its first clinical application was in patients with epilepsy (Mader, Anastasopoulos et al. 2010). Figure 1.7 Principle of the Global Tracking method. The target density curve changes over temperature. It is relatively flat in a very high temperature (a). As a result, the cylinders are almost random. The target density is sharper in lower temperature (b) which results in more structured but still variable cylinder configurations. The target density is very narrow in a temperature nearly zero (c). Therefore only samples that are inside a local extremity are drawn. Hence, the cylinder configurations are very structured and only small changes are possible (Kreher et al. 2008). 21

24 1.5 Brain plasticity In contrast to patients with acute stroke, patients with slow growing lesions such as low grade gliomas have no or only mild neurological deficit, and no or only mild disability in daily life. In comparison to only 30% of stroke patients recovering to professional life (Varona et al. 2004), 90% of patients with LGG are able to comeback to normal life one year after operation (Duffau 2007; Duffau et al. 2007). A recent study on patients undergoing medial frontal operation shows an increased activation in the supplementary motor area (SMA) of contralateral hemisphere in patients with postoperative deficit (Krainik et al. 2001). This indicates the brain ability to compensate the functional damage by a plastic change to the SMA areas on the healthy hemisphere. Another study on patients with SMA LGG shows an overactivation of the contralateral SMA (Krainik et al. 2004). There is a correlation between the recruitment of contralateral hemisphere with the percentage of preoperative SMA activation removed. One study on language function in patients with Broca area LGG points out a reorganization of language area at ventral and dorsal premotor cortices (Benzagmout et al. 2007). Postoperative studies confirm that resection of Broca area in these patients does not result in any language disability. The reason of this good clinical outcome seems to lie in the better and more effective reorganization of the brain in slow growing lesions. Regardless of many researches on brain plasticity, the mechanism of long term brain plasticity remains unclear. 22

25 1.6 Aims of the study The aims of this work were to compare the detection of the CST by the three tracking methods (FACT, Probability Maps and Global Tracking) in human gliomas. Additionally, the effect of localization and mass effects of gliomas on brain motor function was evaluated. 23

26 2. Methods 2.1. Patients Patients were recruited from the Outpatients Clinic of the Department of Stereotactic Surgery, Freiburg University Medical Center. The inclusion criteria are (1) patients (male and female) aging from year old with unilateral low grade gliomas (WHO grade II) or anaplastic astrocytomas (WHO grade III) in rolandic area. The exclusion criteria are (1) patients with pacemakers, vascular clips, metal cardiac valves (2) patients with claustrophobia (3) patients with mental problems or too ill, who are unable to perform the motor task (4) patients who have received gross resection of tumors. Thirteen patients (3 females and 10 males, mean age: 43 years ± 10) with 5 WHO grade II and 8 WHO grade III tumors affecting the motor system participated in this study. They were successively recruited from our outpatient s department and prospectively included in the study. Attributed to the localization of the tumor in an eloquent brain region, a stereotactic biopsy had been performed for the determination of the histology. All patients had a history of epilepsy during their course of disease. The main duration of disease was 93 months ± 63. Written informed consent was obtained from all patients. The study was approved by the local ethics committee ( 24

27 2.2. Paradigms Functional MRI Functional MRI was performed at whole body 3 Tesla scanner (TIM Trio, Siemens, Erlangen, Germany) by using a 4 channel head coil with single shot EPI sequence, TR 2610 ms, TE 30 ms, image dimensions 64x64x42, voxel size 3x3x3 mm. Motor paradigm A block-related design consists of passive movements of wrists using a pressure-driven arm splint with movement's frequency of 1 Hz. The passive movement was chosen as an identical stimulation was aimed for every patient. There were four blocks of passive hand movements alternating with five blocks of rest. Each block lasted 25 seconds. There was no gap between consecutive blocks. Visual instructions were given through MR-compatible mirrors. Passive motor task Block Design (duration of each block = 25s) Time [s] Diffusion Diffusion measurements using 61 diffusion directions at b = 0 s/mm 2 and 1000 s/mm 2, TR ms, TE 94 ms, 68 slices, voxel size 2x2x2 mm. High resolution T1 weighted 3D anatomical images were obtained by a MPRage sequence, TR 2200 ms, TE 2.15 ms, voxel size 1x1x1 mm. 25

28 2.3. Assessment of handedness Handedness was accessed using Edinburgh-Inventory (Oldfield 1971). It consists of ten items of daily living activations such as writing, drawing, throwing, scissors, toothbrush and spoon using etc. The handedness index was calculated as follows: HI= (R-L)/ (R+L) HI: handedness index, R and L are the numbers of acts performed by right and left hands, respectively. Patient is right-handed when HI > 0. This simple and practical method has been proven to be reliable with high test-retest reliability (Ransil et al. 1994; Busch et al. 2009) Motor assessment Motor assessment was performed using Fugl-Meyer test, arm section (Fugl- Meyer et al. 1975), where the following score was used: 0=no, 1=slight, 2=moderate, 3=marked, 4=severe motor impairment Data processing Functional MRI All images were corrected by a custom program to exclude the motion-related effects. FMRI analysis was performed off-line using SPM 8 (Wellcome Department of Imaging Neuroscience, Institute of Neurology, University College London, England). Anatomical images were co-registered to functional images. Time-series images were smoothed using Gaussian smoothing kernel of 8x8x8 FWHM. No normalization was done. The T1 images were segmented into grey matter (GM), white matter (WM) and 26

29 corticospinal fluid (CSF). This information was taken to create a brain mask being later used for the definition of the tracking area (WM). An fmri model was specified with a basis function of a canonical of Hemodynamic Response Function (HRF), no derivatives. T contrast (first level specification) was calculated for passive movement versus baseline for the stimulation of motor system which has been established in Freiburg Brain Imaging Laboratory Fiber tracking High angular resolution diffusion imaging (HARDI) (Hirsch et al. 2003; Gorczewski et al. 2009) was used for all three tracking methods. Fiber tracking was performed by in-house software being available at Seed regions The seed region was taken from the global maximum of the statistical t-map and extended by 7 mm in all directions using Matlab, MathWorks, Massachusetts, United States, The target region was placed manually at the known CST area of anterior pons similar to method used by Kim et al. (Kim et al. 2008). FACT FACT was calculated from the global maximum to the pons. It was running on the orientation distribution function (ODF) created by the constant solid angle approach (Aganj et al. 2009). The cutoff in the spherical harmonic 27

30 expansion was set to 4 and no further regularization was applied. The stop criterion of FACT was set to fractional anisotropy level of FA>0.1 and the allowed maximal curvature was 90 degree between two consecutive steps. Probability Maps Two visiting maps were calculated at threshold FA>0.1. A connected map was created by combining two visiting maps according to (Kreher, Schnell et al. 2008). No revisiting was allowed. Threshold was set where the least fibers appeared. The depicted fibers were displayed in 3 orthogonal sections. Global Tracking The parameters of Global Tracking were chosen as suggested in (Reisert, Mader et al. 2010). The iteration parameters are: start temperature 0.1, stop temperature: 0.001, 3x10 8 iterations. The model parameters were: cylinder s width 1 mm and length 3 mm. The weight of a cylinder was set to a ¼ of the DWI-signal deviation. The domain of reconstruction was obtained by selecting all voxels with a FA>

31 2.6. Image analysis Image analysis of fmri and three tracking methods was performed focusing on the cortical activation and the detection and deviation of CST by three tracking methods. Tracking was considered successful if fibers connecting two seed regions were depicted. Any termination of fibers in between was considered unsuccessful. The degree of fiber s deviation was scored using an index of deviation (ID) which was calculated as follows: ID= (a-b)/ (a+b). a: distance from the CST to the midline in [mm] on the healthy side. b: distance from the CST to the midline in [mm] on the pathologic side. The ID varies between -1.0 and +1.0, where a positive value indicates deviation of CST on the pathologic side, and a value of 0 indicates symmetrical CST. The FA values of voxels within detected tracts on both pathologic and healthy sides were calculated using a Matlab-implemented tool. The FA asymmetry was calculated as suggested by (Stinear et al. 2007): FA asymmetry = (FA healthy FA pathologic)/(fa healthy + FA pathologic). 29

32 FA pathologic: the most noticeable change of FA value within CST on the pathologic side. FA healthy: the FA value within CST at corresponding location on the pathologic side. FA asymmetry varies between -1.0 and +1.0, where a positive value indicates reduced FA in the pathologic CST, and a value of 0 indicates symmetrical FA in the CST. 2.7 Statistical analysis Statistical analysis was conducted using SPSS 18.0 (SPSS Inc, Chicago, USA). A Spearman s rho correlation coefficient was calculated for grade of motor impairment (Fugl Meyer test, arm score) with deviation of fibers, for FA asymmetry with motor impairment (Fugl Meyer test, arm score) and for the grade of deviation with the histology of tumor. FA values on the healthy and the pathologic side of the midbrain and cerebral peduncle were tested by using a two-sided t-test. 30

33 3. Results Subjects of this chapter are the results of fmri and tractography experiment. A brief description of the results is presented in Table Motor function Diagram 3.1. Motor impairment of WHO grade II (low grade) and III (high grade glioma). 50% of WHO grade III had marked motor impairment. According to Fugl-Meyer test, six patients (cases # 2, 4, 5, 8, 11 and 12) didn t suffer from weakness. Half of them (cases # 2, 4 and 5) were WHO grade II, while the other half (cases # 8, 11 and 12) were WHO grade III. Two 31

34 patients (cases # 1 and 3) had slight movement impairment. Both were WHO grade II. One WHO grade III (case # 10) had a moderate motor impairment. Four (cases # 6, 7, 9 and 13) had marked motor impairment. All of them were WHO grade III. There was no statistically significant difference between low grade and high grade gliomas regarding the grade of motor impairment. However, high grade tumor tends to induce severe impairment given that all patients with marked motor impairment were HGG (50% of HGGs). 3.2 Functional MRI Cortical activation was redistributed around the tumor in one patient (patient #2), Figure 3.1, who had a WHO grade II glioma and one patient (patient #10), Figure 3.2, with a WHO grade III glioma. Patient #2 had no weakness, while patient #10 had a moderate motor impairment. Activation within tumor was found in one patient (patient #7), Figure 3.3, who had a WHO grade III tumor which was accompanied by marked motor impairment. Activation of the motor cortex in both hemispheres was found in one patient (patient #13), Figure 3.4, who had a WHO grade III tumor accompanied by marked motor impairment. Perilesional activation was also seen. In other patients (patients #1, 3, 4, 5, 6, 8, 9, 11 and 12) a normal hand presentation was found. Additionally, activation of SMA was found in most patients. 32

35 Figure 3.1 Patient #2. WHO grade II glioma. redistributed around tumor. Activation was found Figure 3.2 Patient #10. WHO grade III tumor. Perilesional activation was found. 33

36 Figure 3.3 Patient #7. WHO grade III tumor. Activation persists within tumor. Figure 3.4 Patient # 13. WHO grade III tumor. Activation was found at precentral gyri bilaterally. Perilesional activation was also seen. 34

37 3.3 Fiber tracking FACT was not successful in four patients (patients #1, 6, 7 and 10). In all cases, FACT was terminated in regions where FA was markedly reduced. The termination of FACT was accompanied in two patients (patients #6 and 7) by a marked, in one patient (patient #10) by a moderate and in one patient (patient #1) by a slight motor impairment. Global Tracking was successful in 12/13 cases. It was terminated in one case (patient #7) with a huge fronto-parietal tumor infiltrating the cortex and the underlying white matter. FA value within depicted tract was dramatically reduced. The patient suffered from marked motor impairment. Probability Maps of motor fibers could be obtained in every case, even in cases with large tumors and mass effects. In one case (patient # 7), however, where FACT and Global Tracking were not successful, only Probability Maps with a low Probability could be found. FA value within the depicted tract was very low (0.2 ± 0,04). 35

38 Table 3.1 Clinical characteristics and description of fmri and DTI Patient number WHO grade. II - III 1 II Tumor localization Fugl Meyer Test Score 0-4 fmri of pathologic hemisphere FACT Probability Maps Global Tracking Index of deviation FA of CST on the pathologic side FA of CST on the healthy side FA asymmetry L-frontotemporal 1 M1 intact No Yes Yes 0, (z=48) 0.48 (z=48) 0,33 2 II R-parietal 0 M1 relocated around tumor Yes Yes Yes 0, (z=52) 0.44 (z=52) -0,23 3 II L-temporal 1 M1 intact Activation of ipsilateral SMA Yes Yes Yes 0, (z=31) 0.64 (z=31) -0,08 4 II L-frontal 0 5 II L-frontal 0 M1 intact Activation of ipsilateral SMA Yes Yes Yes 0, (z=37) 0.67 (z=37) -0,08 M1 deviated, but intact Activation of ipsilateral SMA Yes Yes Yes 0, (z=58) 0.50 (z=58) -0,13 6 III R-frontotemporal 3 M1 intact Activation of ipsilateral SMA No Yes Yes 0, (z=43) 0.55 (z=43) 0,37 7 III L-parietal 3 Activations persist within tumor No Yes No 1, (z=40) 0.50 (z=40) 0,45 8 III L-frontal 0 M1 intact Activation of ipsilateral SMA Yes Yes Yes 0, (z=52) 0.41 (z=52) 0,17 9 III L-frontaltemporal 3 M1 intact Yes Yes Yes 0, (z=47) 0.38 (z=47) 0,29 10 III 11 III R-parietaltemporal 2 R-parietooccipital 0 No activation at corrected Activation around tumor at uncorrected No Yes Yes 0, (z=55) 0.4 (z=55) 0,23 M1 deviated, but intact Activation of ipsilateral SMA Yes Yes Yes 0, (z=56) 0,48 (z=56) -0,02 12 III R-parietooccipital 0 13 III L-frontal 3 M1 deviated, but intact Activation of ipsilateral SMA Yes Yes Yes 0, (z=45) 0.39 (z=45) 0,18 Peritumural activation Activation of contralateral M1 Activation of SMA bilaterally Yes Yes Yes 0, (z=50) 0.48 (z=50) 0,37 36

39 3.4 Deviation of fibers The ID was calculated based on Global Tracking. Only in case #7, Global Tracking was not successful in a region with very low FA that might suggest a strongly deviated or disrupted tract, a ID of 1 was given. Significant deviation (ID>0.1) was found in 7 cases (# 1, 3, 4, 6, 9, 10 and 13). Normal or mildly deviation (ID< 0.1) was found in the other cases (# 2, 5, 8, 11 and 12). In case # 7, the fibers found by FACT and Global Tracking were prematurely terminated. The deviation of the tract was correlated with the grade of motor impairment (Spearman s rho r=0.7, p=0.007), Figure 3.5. However, there was no correlation between the grade of deviation and the histology of tumor. Figure 3.5 Correlation between the index of deviation and grade of motor impairment (Fugl- Meyer test), (Spearman s rho r=0.7, p=0.007). 37

40 3.5 Change of FA value The FA values within the depicted tracts were reduced in eight patients (patients # 1, 6, 7, 8, 9, 10, 12 and 13). One (patient #1) had WHO grade II glioma. The others had WHO grade III tumor. The FA asymmetry was correlated with the motor impairment (Spearman s rho r=0.82, p=0.001), Figure 3.6. FA values at z-coordinates of 26 ± 2 (along the z-axis of the scanner) were low in the region of the cerebral peduncle in 8 patients (# 1, 2, 4, 5, 6, 8 and 10), although only in 1 patient (#6) an infiltration of the midbrain could have been suspected. A two-sided t-test, however, did not reveal any difference between the tumor and the healthy side with a FA of 0.47 (SD ± 0.14) for the healthy side and a FA of 0.37 (SD ± 0.15) for the tumor side at the midbrain level, p=0.08. This means that the reduction of FA in the midbrain is rather related to a general degradation of image quality than to an infiltration by the pathology. In fact, images at the midbrain did show most distortions, even though a distortion correction had been applied (Zaitsev et al. 2004). Figure 3.6 Correlation between the FA asymmetry and grade of motor impairment (Fugl Meyer test), (Spearman s rho r=0.82, p=0.001). 38

41 Figures show details of three tracking methods (FACT, Probability Maps and Global Tracking) in thirteen patients. a, b are FACT, showing a projection of fibers; c, d are Probability Maps, showing a particular section with fibers; e, f are Global Tracking, showing a projection of fibers; g, h are FA values of depicted fibers on the right (R) and left (L) hand side. 39

42 Figure 3.7 Patient #1 had a huge, diffuse tumor in the left frontal lobe. FACT was terminated on the left side in a region with reduced FA from z=60 ( ) until z= 45 ( ). Probability Maps and Global Tracking were successful. ID =

43 Figure 3.8 Patient #2 had a big tumor within the right precentral gyrus. No motor impairment was present. On fmri during right hand movement the activated areas were relocated around the tumor (see Figure 3.1). All tracking methods were successful. There was no significant change of FA within depicted tract compared to the healthy side. ID =

44 Figure 3.9 Patient #3 had a well defined tumor within the left temporal lobe. All tracking methods were successful. The CST was slightly deviated (ID = 0.14) which is accompanied by a slight motor impairment. There was no significant change of FA within depicted tract compared to the healthy side. 42

45 Figure 3.10 Patient #4 had a well defined tumor in the left frontal lobe. FMRI activations were found in the left M1 and left SMA during right hand movement. All tracking methods were successful. There was no motor impairment. DTI showed an ID of There was no change of FA within depicted tract compared to the healthy side. 43

46 Figure 3.11 Patient #5 had a tumor in the left frontal lobe which extends to the precentral gyrus. The precentral gyrus was slightly deviated backwardly, but was still identifiable. FMRI during right hand movement still showed activation of the left M1 but the cortical activation was slightly reshaped around the tumor (not shown). The patient didn t have motor impairment. There was no change of FA within depicted tract compared to the healthy side. ID =

47 Figure 3.12 Patient #6 had a large, heterogeneous WHO grad III tumor in the right fronto-temporal lobe, accompanied by marked motor impairment. On structural images the precentral gyrus was well preserved. FMRI showed well activated M1 and SMA on the lesioned hemisphere (not shown). Probability Maps and Global Tracking were successful. FACT was terminated in a low FA circumstance. ID =

48 Figure 3.13 Patient #7 had a WHO grad III tumor in the left parietal lobe which extends to the precentral cortex. On structural images the left precentral gyrus was strongly involved. FMRI during left hand movement showed multiple activated foci within the tumor (see Figure 3.3). FACT and Global Tracking were not successful. FA within depicted tract was reduced remarkably from z=57 ( ) to z=38 ( ). ID = 1. 46

49 Figure 3.14 Patient #8 had a WHO grad III tumor in the left frontal lobe. The patient suffered from slight motor impairment. On structural images the precentral gyrus was well preserved. Activations were found within the left M1 and left SMA during right hand movement. All tracking methods were successful. FA within depicted tract was clearly reduced, especially at z=52 ( ). ID =

50 Figure 3.15 Patient #9 had a left fronto-temporal WHO grad III tumor. The precentral gyrus was well preserved and fmri showed normal activation of M1 (not shown). All tracking methods were successful. FA was reduced within depicted tract on the lesioned hemisphere, especially at z=48 ( ). The patient had marked motor impairment. ID =

51 Figure 3.16 Patient #10 had a right parieto-temporal tumor involving the precentral gyrus. No activation was seen on fmri at corrected threshold (FWE = 0.05). When an uncorrected threshold was used, there were activated areas arround tumor (see Figure 3.2). Probability Maps and Global Tracking were successful. FACT was terminated within an area with low FA. ID =

52 Figure 3.17 Patient #11 had a WHO grad III tumor in the right parieto-temporal lobe with mass effect. The central sulcus and precentral gyrus were compressed and deviated anteriorly. FMRI during left hand movement showed activation of right M1 and right SMA. The CST was not deviated (ID = 0.00). The patient had no motor impairment. No change of FA was present within the depicted fibers. 50

53 Figure 3.18 Patient # 12 had a WHO grad III tumor within the right parieto-temporal lobe including the postcentral gyrus. The precentral gyrus was deviated anteriorly. FMRI showed normal activation of M1. Small activation within the right SMA was also found. All tracking methods were successful. FA within the right CST was reduced, especially in region from z=49 ( ) to z=45 ( ). The patient had no motor impairment. ID =

54 Figure 3.19 Patient #13 had WHO grade III tumor within the left frontal lobe. The precentral gyrus was directly involved. FMRI during right hand movement showed activation of the M1 and SMA bilaterally (see Figure 3.4). All tracking methods were successful. FA was reduced within left CST from z=59 ( ) to z= 49 ( ). ID =

55 4. Discussion 4.1. Reorganization of brain function The absence of neurological impairment in individuals with slow-growing lesions such as LGG is widely recorded. The hypothesis of compensatory recruitment of ipsilateral or contralateral networks has been suggested by many authors (Chollet et al. 1991; Weiller et al. 1992; Lewine et al. 1994; Luders et al. 1997). There are two main theories about mechanisms of plasticity. The first is about unmasking of the existing, but latent, horizontal connections (Sanes et al. 2000). The second is the modulation of synaptic efficacy such as long-term potentiation (LTP) or long-term depression (LTD). Neurotransmitter system required in LTP and LTD include activation of N-methyl-D-asparate receptors and down regulation of the inhibitory γ-aminobutyric acid system (Hess et al. 1994; Hess et al. 1996; Hess et al. 1996). The fundamental concept of those theories, as suggested by many authors, is that the cortex consists of multiple overlapping cortical representations (Donoghue et al. 1992; Schieber et al. 1993). These areas are functionally connected through an intensive bidirectional horizontal network (Huntley et al. 1991). However, the brain responds to different stimuli in different ways. As discussed in previous chapters, plasticity is more effectively in slow-growing lesions than in acute insults. A lot of studies have been performed with the purpose to identify the brain s functional compensation in both stroke and gliomas, in both human and animal. 53

56 The effect of brain tumor on the motor system has been investigated intensively during the last decades. Using magnetic source imaging (MSI) Schiffbauer and colleagues have shown that functional activity can persist within tumor (Schiffbauer et al. 2001; Schiffbauer et al. 2002). Similar results were found in magnetoencephalography (MEG) studies on patients with gliomas (Ganslandt et al. 1997; Ganslandt et al. 2004) and in fmri study on patients with cerebral tumors (Mueller et al. 1996). One of the first fmri studies about brain plasticity was reported by Yousry in 1995 (Yousry et al. 1995). It has shown that there is difference in size of motor cortical representation between patients with tumors when compared to that in healthy volunteers. Those findings were proved by intrasurgical stimulation. The motor cortical activation was diffuse and redistributed around the tumors, instead of being spot-like and located near the posterior bank of precentral gyrus. The possible explanation for those observations, according to the authors, is the plastic reorganization of the cortical representation induced by tumors (Yousry, Schmid et al. 1995). Similar results were reported in another fmri study in patients with gliomas, in which most of the tumors showed activation within or immediately adjacent to the periphery of tumors (Righini et al. 1996). Beside local compensatory mechanism, plasticity can occur in remote areas. A positron emission tomography (PET) study showed activations within premotor cortex during voluntary finger movements, along with activation within primary motor cortex and parietal somatosensory cortex (Seitz et al. 1995). Another area involved in remote compensatory mechanism is SMA (Atlas et al. 1996; Wunderlich et al. 1998). Another study which included 110 patients with cerebral tumors was conducted in 2002 by Krings. It revealed an 54

57 increase in activation of ipsilateral SMA with increase of motor deficit, whereas decrease in activation of primary motor cortex was noted. The increase in activation of SMA was interpreted as an indicator of the lesioninduced reorganization of the brain (Krings et al. 2002). Compensatory plasticity can occur even in the contralateral hemisphere. When compared patients who have perirolandic tumor to healthy volunteers, Yoshiura found that the activation within motor cortex of the healthy hemisphere is much higher in patients than that in the control group. This was thought to reflect the plasticity in order to compensate the damaged function (Yoshiura et al. 1997). Similar results were reported by Fandino. FMRI results were validated by intraoperatvie cortical stimulation and the agreement was up to 82% (Fandino et al. 1999). A study with transcranial magnetic stimulation (TMS) also showed similar results (Caramia et al. 1998). Four basic patterns of brain reorganization were described which could also be found in our patients. First, motor function can persist within the tumor due to its infiltrative behavior, as seen in case # 7, where motor activation could be demonstrated within the tumor, however accompanied by a marked motor impairment. Second, functional areas can be redistributed around the tumor, as seen in cases # 2 and # 10. Third, a functional network can be recruited within the ipsilateral hemisphere, especially SMA as seen in 7 cases (# 3-6, 8, 11, and 12). Fourth, networks of the contralateral hemisphere are recruited, as seen in case #

58 4.2. Limitation of fmri FMRI can provide valuable information concerning cortical representation of motor system. However, its results should be interpreted carefully. The first problem is the anatomic distortion caused by tumor that sometime makes it difficult to identify cortical landmarks which are important for fmri interpretation. A merely displaced primary motor area should not be mistaken as peritumoral redistribution. Furthermore, any anatomic changes caused by tumor can result in errors of segmentation of structural image (T1) which may lead to mis-coregistration between time-series images and structural images and errors of normalization process from individual space to Tailarach space. Different ways have been suggested to overcome this problem. Most common are the unified segmentation suggested by Ashburner (Ashburner et al. 2005) (which was used in this study) and the cost function masking suggested by Brett (Brett et al. 2001). Furthermore, BOLD signal is dependent on the neurovascular coupling, which is determined by the response of vascular system to the neuronal activation (Frostig, Lieke et al. 1990; Narayan, Esfahani et al. 1995; Ngai, Meno et al. 1995; Malonek and Grinvald 1996; Jones, Schirmer et al. 1998; Hoge, Atkinson et al. 1999; Schroeter, Kupka et al. 2006). In patients with brain tumors, especially gliomas, the tumor-induced changes of vascular auto regulation can result in loss of cortical activation within and around the tumors (Holodny et al. 1999; Schreiber et al. 2000; Ulmer et al. 2003). Holodny in 1999 presented a case, in which fmri was done preoperatively in a patient with a glioblastoma multiforme (Holodny, Schulder et al. 1999). The fmri result showed a reduction of activation in the sensorimotor area, which confirmed function normally by intraoperative bipolar direct stimulation. In 56