University of Southern California Dissertations and Theses

Three-dimensional functional mapping of the human visual cortex using magnetic resonance imaging

(USC Thesis Other)

Three-dimensional functional mapping of the human visual cortex using magnetic resonance imaging

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content THREE-DIM ENSIONAL FUNCTIONAL MAPPING OF THE HUMAN VISUAL CORTEX USING MAGNETIC RESONANCE IMAGING by Tayi Shen A Thesis Presented to the FACULTY of THE SCHOOL of ENGINEERING UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree MASTER of SCIENCE in BIOMEDICAL ENGINEERING December 1994 Copyright 1994 Tayi Shen This thesis, written by Tayi Shen under the guidance of Faculty Committee and approved by all its members, has been presented to and accepted by the School of Engineering in partial fulfillment of the re quirements for the degree of Master of Science 1 3 - - I S - I 9 9 H ............ I-'acuity C o m m ittee Chairman \ C'tC'jC -iv^* Dedication: To my parents and my beautiful wife for their love and support. Acknowledgment: I would like to thank Dr. Manbir Singh who provided me a lot of guidance and support about this project. I am also grateful to Mr. Tae- Seong Kim, Mr. Deepak Khosla. Mr. Pankaj B Patel and Mr. Hyun Kim for their help for the experiments and analysis The experiments were set up at the LAC/USC Diagnostic Imaging Center Table of Contents Topic Pane I. Introduction 1. Functional MRI 1 2. Objective 2 II. Functional magnetic resonance imaging 1 Nuclear Magnetic Resonance 3 2 Tissue Characterization: and T, 9 3. Brain Activation Metabolism 10 4. Imaging and Pulse Sequencing 13 III. Method 1 MR Imaging 22 2. Test Protocol 24 3. Data analysis 25 4. 3D Representation 4.1 Preprocessing 28 4.2 Surface rendering 29 4.3 3D registration 33 4.4 Cutting away 33 IV. Results and discussion 35 V. Conclusion 39 VI. R eferences 4 1 List of Figures ____________ liti£ ___________________________ b i & L . f igure 1. Fxamples o f magnetic dipoles 4 Figure 2. Nuclei without external magnetic field 5 Figure 3. Nuclei with external magnetic Field 5 Figure 4. Relationship between Larmor frequency and 6 external magnetic field strength Figure 5. Nuclear magnetic resonance 7 f igure 6. Induction and relaxation 8 Figure 7. T, relaxation 10 Figures T, relaxation 10 f igure 9. Metabolic pathway o f the physiological 12 activity in brain f igure 10 Slice selection in z. direction using z gradient 14 Figure 11. Principle of frequency encoding 16 Figure 12. Image acquisition gradients in three dimensions 17 Figure 13. Principle o f spin echo 18 Figure 14. Spin echo pulse sequence 19 Figure 15. Gradient echo pulse sequence 20 Figure 16. The principle o f the subtraction f unctional 2 1 MR imaging Figure 17. f unctional anatomy of the visual cortex 22 f igure 18. Scout sagittal image with 6 selecting planes 23 Figure 19. Scout image with 40 anatomic selecting planes 23 Figure 20. The visual cortex functional mapping 26 f igure 21. Functional anatomy of the midbrain in axial 27 MR image f igure 22. 3D reconstruction o f anatomical images 31 Figure 23. The procedure o f integrated 3D views o f 32 activation and anatomy f igure 24. Anatomical surface rendering after applying 34 a "cutting away" technique Figure 25. The activated region o f visual cortex 36 Figure 26. The activated region o f visual cortex 36 Figure 27. A "cutting plane" applied on the anatomic structure 37 Figure 28. 3D Functional mapping o f visual cortex 37 after applying a "cutting away" technique THREE-DIMENSIONAL FUNCTION MAPPING OF THE HUMAN VISUAL CORTEX USING FUNCTIONAL MAGNETIC RESONANCE IMAGE I. Introduction I. Functional MRI Medical imaging modalities such as magnetic resonance imaging (MRI). position emission tomography (PET), or single- emission computed tomography (SPECT) provide clinicians and investigators with the capability to visualize brain structure and to observe metabolic activities in the brain. Signal intensity enhancement in the human visual and motor task performance have been demonstrated by using different imaging modalities in recent studies (5,6,7,10,1 1,12, 20). MR image data are superior to other modalities because MR images provide greater contrast at the brain surface and at other soft tissue interfaces (14). Further more, MRI is a harmless procedure, while other modalities provide a radiation dose, especially when thin contiguous images are acquired for a 3D reconstruction. Local changes in cerebral blood flow, blood volume, oxygen extraction fraction, and metabolic rate associated with brain activity have been measured with positron emission tomography (5,6). These signal intensity increases appear to be a local increase in oxygenated blood during neuronal activity(5). Without radiation dose, functional 1 magnetic resonance imaging using fast gradient echo and fast echo planar pulse has provided a higher resolution of brain neuroimaging by detecting the tiny change of magnetic field induced by the blood oxygenation. 2. Objective The first objective was to design a visual stimulus to elicit the response of the primary visual cortex The second objective was to image the activity of the visual cortex through changes in the cerebral blood flow and oxygen consumption. Such tomographic mapping of the activated cortical regions by visual stimulation could be imaged using a non-invasive MRI. The final goal was to process the tomographic slices of both functional and anatomic slices to reconstruct a 3D model for visualizing the functional mapping of the primary visual cortex in a 3D fashion Since MR data is true 3D data, we may apply some technique to manipulate the 3D images to get better results visually than conventional 2D methods II. Functional Brain MRI 1. Nuclear Magnetic Resonance Magnetic resonance imaging is based on the principle of nuclear magnetic resonance, which has long been known to physicists as early as ]946( I ) The earliest well application of MR to clinical medicine came in 1971 Since that time, the technology rapidly advanced and has enabled the use of MRI as a standard procedure in medical diagnosis The physics o f magnetic resonance can be described in terms of quantum mechanics. There are about 1,800 recognized combinations of protons and neutrons called nuclides. Most of the nuclides are unstable O f the approximately 280 stable nuclei, only 100 of them possess a very small magnetic moment by their intrinsic mechanical rotation, called spin. We can consider about the hydrogen nucleus as physically spinning about an axis. Since its positive charge is not located at the exact center of the nucleus but is located some distance from the axis, as the nucleus spins, the positive charge moves in an approximately circular path as an electron in a wire. This kind of movement generates a magnetic field and behaves as a permanent magnet dipole (Fig. 1). Only the nucleus containing an odd number of protons or neutrons exhibits magnetic moment, because even numbers of nucleons tend to align such that their spins and magnetization cancel Fig 1 Examples of magnetic dipoles The spinning nucleus behaves as a magnet (Adapted from Huk, Gademann, and Freidmann 1990) Without an external magnetic field, the magnetic field of nuclei is not measurable because the axes of the huge number of tiny magnets are oriented randomly In this case, the sample as a whole does not possess a net measurable magnetic moment M 0(Fig. 2). When the nuclei are exposed to a homogenous static field, the randomly oriented magnetic dipoles line up the magnetic field either parallel or antiparallelfFig. 3). The direction parallel to the applied magnetic field B0 is defined as the z axis o f the three-dimensional coordinate system The energy difference between these two orientations is very small The extremely weak net magnetic moment M( 1 depends on the Electric m agnet Spinning atomic f nucleus \ Permanent magnet 4 external field strength and temperature The measurable signal is greater when the nuclei is exposed in a stronger magnetic field Fig. 2 Nuclei in the absence of an external magnetic field The net magnetic moment is zero due to randomized orientation of the spin axes (Adapted from Huk. Gademann, and Freidmann 1990) 7/ — N — Fig 3. Nuclei in the presence o f an homogeneous external magnetic field B„. The spinning axes are aligned either parallel or antiparallel to the direction of the applied field (Adapted from Huk, Gademann, and Friedmann 1990) 5 Magnetic resonance exists when the excitation frequency is the same with the natural frequency of the system. The frequency of the oscillation depends entirely on the oscillating system itself. The resonance frequency is defined by the equation / = o)/2tt = y/#()/2jr ( 1) which states that for every nuclear magnetic resonance nucleus. excitation can occur only at a certain frequency which is proportional to the magnetic field strength # 0(Fig. 4) The processional frequency / is also termed Larmor frequency, named after British physicist. Sir Joseph Larmor (1857-1942) Frequency MHz 60- 42.6 0 1.5 0.5 Field Strength, Tesla Fig. 4. The resonant Larmor frequency of the hydrogen is linear proportional to field strength. 6 The strength of the magnetic moment depends on the type of nucleus and is described in terms of the gyromagnetic ratio y The hydrogen nucleus, consisting one proton, possesses the largest gyromagnetic constant This fact makes the proton suited for imaging application very well. When applied a RF magnetic excitation pulse, same frequency as Larmor frequency and perpendicular to the direction of magnetic field B(), the net magnetization vector is flipped to a certain angle which increases with the strength and duration of the excitation pulse Fig 5 Nuclear magnetic resonance The nuclei spinning at the same frequency and phase, are induced a precessional motion out of alignment with the main field by applying an electromagnetic RF pulse is the component of net magnetic M on the x-y planet Adapted from Huk, Gademann, and Friedmann 1990) (Fig. 5) C C 7 The transverse magnetisation (Mxy) is a component of the net magnetization vector perpendicular to the mam filed at the Larmor frequency after RF excitation. Since the magnetization parallel to the is relatively small and unmeasurable, the transverse magnetization is the signal to be detected without any interference After the excitation pulse has been turned off, still persists and it induces an alternating voltage in a receiver coil which is perpendicular to the main filed (Fig 6) This signal is called a free induction decay(FlD) signal and is the main role of NMR Fig. 6. The precessional motion of M induces an alternating voltage in the receiver coil after RF pulse excitation Relaxation tends to realign the direction of main field B„ (Adapted from Huk, Gademann, and Friedmann 1990) 8 Hydrogen is of the greatest biological interest among the elements having magnetic nuclei. The reasons are both because it has the most highly magnetic nucleus and because it makes up two-third of the atoms in the living tissue. The hydrogen nucleus is the simplest of all nuclei, consisting of only a single proton. MR imaging of hydrogen is usually referred to as proton image. 2. Tissue Characterization: T, and T, The energy absorbed by RF excitation pulse is gradually released The precessing nuclei release their kinetic energy to the surrounding molecules and their net magnetic axes return to alignment with the field # n The two relaxation mechanisms are chiefly responsible for the superiority on tissue contrast seen in magnetic resonance imaging. The longitudinal or spin-lattice relaxation is also called T, relaxation, and trends to realigns the spins along the original field direction. After s period of T, seconds after the 90 degree RF pulse, Mz has recovered to 63% ( 1/e) of its rest strength(Fig. 7). 9 (ime 3 V y y z M 7 = 0 M; = M x 2 Fig 7 T, Relaxation The recovery process is exponential after end of excitation (Adapted from Huk, Gademann, and Friedmann 1990) The transverse or spin-spin relaxation is also called T 2 relaxation, and is a decay process occurring through the spin-spin interactions (Fig 8) _ _ l/T 2 3 time 1 z X 2 x 3 Fig 8 T, relaxation The dephase process is a rapid exponential decay of transverse magnetization after end of excitation(Adapted from Huk, Gademann, and Friedmann 1990) 10 The rate of transverse dephasing depends on resonance with the spins of neighboring nuclei This form of relaxation takes place without energy loss Localized inhomogeneities in the applied field lead to local differences in the Larmor frequency This effect causes a shorter time constant than T, . called T :* 3. Brain Activation Metabolism Brain activity is supported by energy from the metabolism of glucose to provide adenosine triphoshate(ATP). The brain consumes approximately 20% of cardiac output in resting state. Phosphates must come from glucose and oxygen taken up from the bloodstream continuously for energy supply since there is so little energy stored in the brain in the form of glucose, glycogen or high energy phosphates (Fig. 9). The major use of energy in the brain is to maintain the balance of ions across membranes in an electrochemical disequilibrium that establishes a membrane potential. Additionally, after physiological neural activity, energy is spent internally in the phosphorylation o f second messenger systems and the balance of calcium ion. Thus, when there is a function in synaptic activity there is an increase in glucose uptake and cerebral blood flow into the region of activation( 1,7,17). 11 BLOOD FLOW METABOLISM PHYSIOLOGY Glycogen BBB Glucose — Glucose Lactate «-Pyruvaie Oxygen Oxygen |K + ],, [Ca + + )c Polarized \ Synaptic Potentials Action Potentials Membrane Transport I & Metabolism I Neurotransmitter Uptake Depolarized I [K + Jo . [Ca + + |i ATP t J Phosphocreatine Fig 9 Metabolic pathway of the physiological activity in brain. The membrane potential is depolarized at synapses where neurotransmitter-receptor interactions open ion channels The balance of membrane potential occurs by ion pumping against electrochemiacl gradients ATP provides the energy for this procedure Without oxygen, Glycolytic breakdown of glucose to lactate may be the brain's immediate response to certain stimulation in certain areas (Adapted from Toga 1990). Brain activation studies with PET suggests that much of the metabolism of brain to an increment in cortical activity is by the glycolytic metabolism of glucose to lactate, instead o f the oxidative metabolism of pyruvate to CO ; and water(6) When the cortical region is activated during stimulation, the significant increase of blood flow by 29%, but only 5% increase of oxygen metabolism (5,6) There is a magnetic signal induced when the glycolytic metabolism occurs because oxyhemoglobin is diamagnetic and 12 deoxyhemoglobin is paramagnetic The change of deoxyhemoglobin is the main role of functional magnetic resonance imaging. Paramagnetic deoxyhemoglobin induces an inhomogeneous magnetic field in tissue surrounding blood vessels, causing transverse dephasing and decreased signal intensity in MR images. The decrease of deoxyhemoglobin concentration due to large amount of oxyhemoglobin supply affects the T,* transverse relaxation time constant Tht more the oxyhemoglobin increases the more the T,* increases duo to spin dephasing of nuclei decreases. A longer T :* contributes to increase the T 2*-weighted MR image intensity This effect makes the subtraction image an activation map 4. Imaging and Pulse Sequencing MR images provide a high contrast between white matter, gray matter, and cerebrospinal fluid (CSF). MR is based on multiple tissue parameters( proton density, T t relaxation time, T 2 relaxation time ,and blood flow), unlike CT which only depends on the attenuate coefficient when an X-ray beam has passed through the tissue. As multiple tissue parameters can be manipulated through alterations in the imaging acquisition sequence. Thus, MR has the capacity to provide highly detailed spatial and temporal information concerning brain structure. There are several techniques available for reading the 13 spatial information in the signals for making an image An image can be reconstructed by the principle of computed tomography commonly using the two-dimensional Fourier transform (2DFT) technique. For transverse images, the gradient(Gz ) along the magnetic field functions as the slice selection gradient during the excitation phase. The term gradient means that the magnetic field is altered along a select direction The pulse is given a narrow band frequency spectrum which enable it to excite only in the slice range which the resonance condition is satisfied. Each position along the z direction is mapped to a unique resonance frequency. No signals will be detected outside the slice (Fig. 10). ~ 4 t — Resonance Excitation frequency Slice position Slice thickness Fig. 10.The head is placed in a gradient, which is a narrow frequency range of RF electromagnetic pulses alone the z direction. The z gradient functions as the slice selection and excites the nuclei only in the plane corresponding to the range of resonance frequency (Adapted from Huk, Gademann, and Friedmann 1990). 14 The reason for specifying the thickness of the selection slice is to increase the S/N ratio since the signal from a very thin slice is too weak to reproduce an acceptable image The signal occurring on the selected plane then becomes spatially encoded in the x direction by applying the x grudient((ix) The principle can be illustrated with a phantom consisting two water-filled cylindrical holes that have different locations in the \ direction (f ig I I-a) Without a \ gradient, the samples resonate at the same frequency since they experience the same mam field The free induction decay consists of only one frequency The samples are impossible to be located because the single frequency free induction decay signal can generate only one peak in the frequency domain by using Fourier transformation (Fig I 1 -b ) When the x gradient is turned on, the two samples will experience two different fields in the x direction After the x gradient excitation, the free induction decay will consist of two different frequencies This signal then can deter mine the location o f the samples by using Fourier transformation ( Fig 1 I-c) The MR signal occurring only on the plane excited by z gradient becomes spatially encoded by using x gradient The third dimension is recorded for representing the columns o f the image by phase encoding!Fig. 12) The total imaging acquisition time is determined 15 by the interval between the excitation pulses, called the repetition time(TR) n Time Frequency Frequency Fig. 11. (a), a phantom consisting two cylindrical samples in different x directions, (b), Without the x field gradient(Gx =0), the free induction decay signal consists of a single frequency (c), With a gradient G,. applied, the free induction decay signal consists of two frequencies and the two samples become distinguishable (Adapted from D. D Stark, W. G Bradley, Jr 1992) 16 Fig 12 Acquisition of an image with gradients in three dimensions (Adapted from Huk, Gademann, and Friedmann 1990) There are three commonly used RF pulse sequences: spin-echo, inversion recovery, and gradient echo. The spin-echo technique has several advantages in MR imaging: (!) T 2weighted images are possible by using this sequence, (2) it cancels the problem of magnetic field inhomogeneities, and (3) the time window of the MR signal is easy to be determined. After a 90-degree pulse, the received signal decays with a time constant T 2, the spin-spin relaxation time. In practice, the transverse magnetization decays much faster with a time constant T 2* due to spatial imhomogeneity of the magnetic field. When applied a 180-degree RF pulse at time x ms after the 90-degree pulse, the magnetization can be rephase at t ms later (Fig. 13). That means by applying a 90-1-180 sequence, the magnetization can be received with T2 decay at time 17 2x after the 90-degree pulse (Fig 14) The time interval between the 90-degree pulse and the echo is termed time of echo delay(TF) In practice, instead of a single spin-echo pulse sequence, several 90-T-I80 pulse at intervals of 0 5-5 0 sec are used as the entire sequence for higher S/N ratio This sequence is not suitable for functional MR imaging due to low sensitivity to T,* \ A y x t = T C = 2 t y X Fig. 13. The principle of a spin echo, (a), The phases of the magnetization spins are coherent at time t=0 (b), The transerve magnetization (M ^ ) dephase after the initial 90-degree pulse. At time t=T , by applying a 180-degree RF pulse, the dephased magnetization is turned into reflection position around the y-axis. (c), At time t=2x, the fast precessing spins(F) catch up with the slow ones(S), so the dephasing process converts to rephase one and Mx > increases again (Adapted from D.D. Stark, W. G. Bradley, Jr 1992). The inversion recovery pulse sequence is a technique with the advantage to measure T, . Generally, it provides better anatomical detail and better contrast than T 2-weighted scans between gray-white matter. 18 Transverse magnetization Fig 14 Pulse sequence of spin echo A 90-1-180 RF pulse sequence rephases the magnetization at time t=2x The amplitude of the rephased signal is proportional to e'112 (Adapted from D. D Stark, W G Bradley, Jr 1992) A 180-degree pulse is applied to invert the net magnetization M7 to -M, Since the magnetization is totally longitudinal, there is no transverse component on x-y plane. After turning off the excitation, the vector begins to grow in positive z-direction at rate T, A 90- degree pulse is applied to the tissue after the initial 180-degree pulse This pulse makes the magnetization measurable The amplitude of the signal is related to the T, decay after initial excitation. The gradient echo pulse sequence has the advantage to measure T2*. The amplitude o f the transverse magnetization is related to T2* 19 right after the RF pulse excitation The RF pulse used in gradient echo sequence usually less than 90" Gradient pulses causes the dephasing. At the end of the first gradient pulse, the net transverse magnetization is almost vanished The second gradient causes an echo when the second pulse has been as long as the first one The amplitude of the transverse magnetization at the time TE is proportional to e 1 1 Gradient echo technique is sensitive enough for the use of functional magnetic resonance imaging with a long echo time Gradient-echo image and FLASH (Fast Low Angle Shot) pulse sequences have been used recently for functional studies due to high sensitivity to T ,* (5 ,12,17). Gradient Transverse m agnetization -t/T2 * Fig. 15. Pulse sequence of gradient echo. The first gradient pulse has opposite polarity to the second gradient pulse. After applied two or more gradient pulse, the transverse magnetization is rephase and the amplitude is proportional to etT2* (Adapted from D. D Stark, W. G . Bradley, Jr. 1992) In this project, a gradient echo technique implemented by Philips Medical System, called Fast Field Echo (FFE) pulse sequence, was used for functional magnetic resonance imaging during the visual cortex stimulation. The decrease in paramagnetic deoxyhemoglobin will lead to less transversal dephase and result in increases in signal intensity with T,* weighted sequences (Fig 16) The tiny difference in blood deoxyhemoglobin concentration in activated cortical region can be detected by using this FFE pulse sequence .SI Difference — — > Mxy Stimulation, T2 Control, T2*' TE Fig. 16. The principle of the subtraction functional M R imaging. 21 III. Method Six healthy male volunteers participated in this study Informed consent was obtained by each subject. All experiments are performed at the LAC/USC Diagnostic Imaging Center 1. MR Imaging Imaging was performed on a commercially available 1.5T Philips Gyroscan MRI system with a standard head coil Head fixation was accomplish by using closely fitting foam pads. Sagittal multisection Tl-w eighted images were obtain to localize the visual cortex(Fig 17). Six coronal slices with 6 mm section thickness were taken on the visual cortex region(Fig. 18) FFE(fast field echo, develop by Philips Medical System) pulse sequence with T R - 69 ms, TE= 40 ms, and 256 x256 spatial resolution in a 240 c n r was performed on the system / 39 4 4 Fig. 17 Functional anatomy of the visual cortex (area 17) (Adapted from Huk, Gademann, and Friedmann 1990) Fig. 18. Scout sagittal image with 6 selecting planes. Fig. 19. Scout image with 40 anatomic selecting planes. For 3D representation purpose 40 anatomic images were taken using T, weighted \1 RI to reconstruct the reference structure(Fig. 19). For generating a smooth appearance of the 3D reconstruction the slice distance was made as short as possible. In this study, the thickness of the anatomic image was 4mm. These 40 2D images were then processed to generate an anatomic structure of the subject. During all procedures, the MR room was darkened and the volunteers were asked to keep their eyes closed until the stimulation 2. Test Protocol A large checker board was set in front of the eyes of the subjects for visual stimulation purpose. The checker board was exposed by a flash light controlled from the control panel. In this study, 8 Hz flash visual stimulation was used, because it has been found that the maximal CBF(Cerebral Blood Flow) response in the cortical region was provided with 8 Hz repetition flash photic stimulation(20). All subjects were asked to concentrate on the visual stimulation during the experiment. Cortical functional images were taken using FFE pulse sequence in both the stimulated and the unstimulated state during a single scanning procedure There were 35 images carried out for the study of each coronal plane. The first five images were taken as a control set without any stimulation Then five images were acquired during stimulation period followed by five images without stimulation as an on-off set. Two 24 more on-off sets were performed for confirming the validating of a result. Each procedure for a single slice took about 7 minutes for acquisition during the experiment. Anatomical images were acquired at the same position with stimulation procedure using Tl-w eighted MRI after the on-off visual stimulation procedure. 3. Data Analysis All image data were stored on tapes and then transferred to Sun Sparc 1 workstation( Sun Computers Inc.) at main campus for further processing Subtraction of the control from the stimulus FFE scans results in a map of cortical activity due to the visual stimulation. An average activation image was created by averaging the fifteen activation images during the three stimulation periods. The fifteen baseline images were averaged to generate an average baseline image. After average activation image and average baseline images were created, an average difference image was generated by subtracting the average baseline image from the activation image on a pixel by pixel fashion. The difference images were set by a threshold for displaying purpose Then, the activation images were superimposed on the corresponding coronal anatomic images (Fig. 2 0 ). 25 .The visual cortex functional mapping The activated visual functional image was superimposed on the anatomical reference image la Me) represent the activation response at different selecting planes Fig. 21 shows the c o rresp o n d in g position o f the visual cortex on a coronal plane Primary optic cortex Fig. 21 Primary visual cortex (Adapted from link. Gademann, and Friedmann 1990) E jk H CjB El mm 1 3 ■ 27 4. 3D Representation When appropriate software is developed to display the brain and its substructures in three-dimensional fashion, then students, clinical technicians, and investigators can easily visualize the complex three-dimensional neuroanatomy of the brain. It is important that biomedical applications of 3D imaging are not limited to enhanced display facilities and more precise measurement tools A software system called 3DVIEWNIX was used in this study for three-dimensional representation purpose. This software package was developed for research purpose by the Medical Image Processing Group, Department of Radiology, University of Pennsylvania. 4.1 Preprocessing The head anatomy of the subjects was imaged with 4 mm thickness and T,-weighted pulse sequences. For obtaining a better resolution after 3D reconstruction, all images were interpolated with 1 mm thickness That means 157 slices were generated after interpolation. There are several ways to achieve this requirement. In this study, a 3D cubic interpolation was performed for smoother interpolations. A Gaussian smoothing filter was also used for each slice. It created the filtered image by essentially weighted averaging 28 the cell intensities within a 3x3 neighborhood within the slice for every pixel. The weights given are determined by a 2D Gaussian function. The weights can be controlled by changing the standard deviation of the function selectable on a scale. Tissue segmentation was then performed to create the object information in the form of a structure system from input images The output resolution can be specified in this procedure 4.2 Surface rendering Surface rendering is the most important technique for reconstructing a 3D image. Surface rendering computes a representation of the structure from the image data by using voxel projection technique instead of the commonly used ray-casting algorithm (8,9,15). By this method, only the voxels that potentially contribute to the display areactually involved in the computation. In many practical situations, thicker shell domain cause only more computation and do not change the rendition A shell SHlASN.u) . as a function of the 3D structure representation, associated with a scene SN which is a pair <V,f >, where V is the scene(2D image) domain of SN and / is the density function of SN with a closed thresholding(low - high) interval [l,h] (8,22). In this studies. / and h are always chosen integers and V corresponds to cubic voxels The to is an opacity function, a mapping <o; K->[0,1J. SHL{SNao) ~ (2) where B is the neighborhood function of voxel v m I’ with opacity function m,, (for example, the six neighbors facing to v) is the unit normal vector to the surface passing through voxel v in B shell domain. This function determine the reflective component o f the contribution from v for the rendition That is, V/(v> r’( v , = i ? ^ r (3 ) /tv ) is the voxel density |V/{v)| is the gradient magnitude of the density function is the neighbor opacity code is the boundary likelihood function Given a 2D stacked image set SN. opacity function to. and threshold interval <l,h>, the implemented programs compute a 3D 30 shell for any specified condition of the voxels in the stacked 2D image set. The shell then can be displayed at any resolution at any resolution relative to the resolution of the given image set Every voxel in B potentially transmits as well as reflects light toward the viewpoint depending on the function < oH In general, it is assumed that the intensity of the incident light due to the external light source is the same for all r in B. As a matter of fact, light becomes attenuated as it passes through the semitransparent volume Thus, the intensity at voxels closest to the light source is higher than that at voxels farthest from the source. Fig.22 shows the anatomic structure of the head with a full opacity. Fig 22. 3D reconstruction of anatomical images!40 slices interpolated to 157 slices) The opacity function o is set to I; no detail inside the object ts visible 31 Filtering Surface Rending Interpolation Surface Rending Interpolation Filtering 3D views of Surface Activated Metabolism 3D views of Surface Anatomv Segmentation/ Thresholding Segmentation/ Thresholding 3D Registration Integrated 3D Views of Activation and Anatomv Acquired Functional MR Images Acquired Anatomy MR Images Fig. 23. The procedure of integrated 3D views of activation and anatomy. 32 The 3D activation mapping can be achieved by the same procedure as that for the anatomical structure. The resolution of the activation and anatomy must be the same for merging. 4.3 3D registration Registration is a technique used to minimize the error of merging images from two different modalities or different acquisition from the same imager. The Tl-w eighted anatomic volume was registered with the corresponding functional study by means of a 3D surface-match technique(23). The 3D functional distribution model was created by applying surface rendering technique. This 3D functional mapping was then superimposed on the 3D model of brain surface anatomy(Fig. 23). 4.4 Cutting away Cutting away is a technique to look inside the object by specifying a plane to remove part of the object. A new surface was shown after applied a cutting plane on the anatomic surface(Fig.24). Functional images can also be superimposed on Tt-w eighted anatomic images at any cutting angle when applied a cut-away technique(23). 33 Fig. 24. Anatomical surface rendering after applying a "cutting away" technique 34 IV. Results and discussion Activation was observed after visual stimulation in all subjects as an increase in signal intensity in the known anatomic location of the primary visual cortex in 2D coronal functional images The 2D image set were preprocessed to generate a 3D functional distribution of the activated visual cortex region The 3D anatomic image was also generated after preprocessing An image registration technique was used to map the signal changes onto conventional anatomic images, which were used to create integrated three-dimensional models of brain structure and function. This procedure minimized the position error between these two objects. The spatial distribution of these regions of signal change was superimposed on the 3D brain model of this subject. Figure 25 shows the activated cortical distribution related to the anatomical structure in three-dimension space with full opacity of both the anatomic and functional structure. The activated visual cortex is visible inside the brain when setting half opacity to the anatomic structure (Fig 26) The anatomic structure was fuzzy and some detail information was lost in this method. When applying a cutting away technique, the cutting plane is superimposed on the 3D structure and can be rotated and translated to any angle and position(Fig. 27) , This technique was the most 35 powerful tool to investigate the tissue, since we can visualize the tissue details or functions without a real cut Fig. 25 3D representation of the activated visual cortex. Both functional and anatomic structures were set to full opacity, so no detail inside the brain was shown Only the surface of the registered structure could be seen The activated region was in red color and was located on the known visual cortex Fig. 26 3D representation of the activated visual cortex. The functional and structure was set to full opacity and the anatomic structure was set to half opacity, so the detail functional distribution inside the brain was visible 36 Fig 27. A cutting plane applied on the anatomic structure The cutting plane could be set to any angle and position. Fig 28 3D Functional mapping of visual cortex alter applying a cutting away technique More activation was found in the structure when a slice with thickness 8111111 was cut away This technique make it possible to investigate any where inside the tissue 37 Emerging MRS(magnetic resonance spectroscopy) and PET(positron emission tomography) data, as well as the recent 3D functional magnetic resonance imaging appear to be consistent with the hypothesis of the function of the visual stimulation. Since functional magnetic resonance imaging data is true 3D data set, it has the advantage to let clinicians see how the brain really functions. Surface rendering permits clinicians and investigators to do in vivo inspections of the brain function If functional and anatomic images can be integrated in three-dimensional space and displayed simultaneously, then certain type of metabolic activity or abnormalities can be localized with excellent spatial and temporal resolution through MRI. Some functional image acquisition planes showed poor activation. Possible explanations for the poor activation at some acquisition plane include section selection outside the optimal regional of activated visual cortex. To display a 3D functional mapping involves lots of complex techniques and computations There are many ways to represent the 2D image set in a 3D fashion The results of representation are highly experimental. What constitutes optimal models o f accurate 3D representation is a subject yet to be studied carefully. 38 V. Conclusion This study has demonstrated a powerful technique to investigate the function of human visual cortex in a non-invasive way and in 3D fashion It is easy to realize the relationship between anatomy and functions in the brain by visualizing the integrated 3D model This goal can be accomplished because of the true 3D image data set and the high flexibility of magnetic resonance imaging. Although the activation during visual stimulation has been proven in this study, it is still far away from achieving an optimal 3D model for functions in the brain. Functional and structural magnetic resonance imaging have the capacity to generate huge amounts of information, and techniques for analyzing neuroimaging data are becoming increasingly sophisticated. Non-invasive measurement of the hemodynamic variables may have a significant effect in the diagnosis and management of patient with epilepsy, cerebral neoplasmia, neurodegenerative disorder and ischemia Direct imaging of cortical activation opens new possibilities for presurgical planning, improved specificity in evaluating dementia, and providing quantitative tools for studying neuropsychiatric disorders at a functional level. The techniques for MR image acquisition, 3D image recon struction model, and computational speed are improving everyday. In 39 the future, it should be possible for people to investigate the 3D brain function more accurately in a real time fashion. 40 VI. References 1. M. Singh, Z.-H. Cho, J P. Jones. Foundation of Medical Imaging. Wiley, 1993 2. A.W. Toga. Three-Dimensional Neuroimaging. New York, Raven, 1990. 3 N.C. Andreasen, G. Cohen, G. Harris, et al. Image Processing for the Study of Brain Structure and Function: Problem and Programs J Neuropsychiatry 1992; 4(2): 125-133 4. S Zeki, J.D.G. Watson, C.J. Lueck, et al A Direct Demonstration of Functional Specialization in Human Visual Cortex. J Neurosci. 1991; 11(3) 641-649. 5. Y. Cao, V.L. Towle, D N. Levin, et al. Functional Mapping of Human Motor Cortical Activation with Conventional MR Imaging at 1.5T. JMRI 1993; 3(6):869-875. 6 . P.T. Fox, M E. Raichle, M.A. Mintun, et al. Nonoxidative Glucose Consumption During Focal Physiologic Neural Activity. Science 1988; 241(22):462-464. 7. B.N. Mora, G J Carman and J.M Allman. In Vivo Functional Localization of the Human Visual Cortex Using Positron Emission Tomography and Magnetic Resonance Imaging TINS 1989; 12(8):282-284. 8 . J.K, Udupa and D. Odhner. Shell Rendering. IEEE CG&A 1993; 13(6)58-67 9 J K. Udupa and D Odhner. Fast Visualization, Manipulation, and Analysis of Binary Volumetric Objects. IEEE CG&A 1991; 11(6) 53-62. 10. M. Singh. Toward Proton MR Spectroscopic Imaging of Stimulated Brain Function. IEEE Tran. Nucl. Sci.; 39(4): 1161-1164. 11. G.M. Hauthout, K.A.T. Kirlew, G.J.K.. So, et al. MR Imaging Signal Response to Sustained Stimulation in Human Visual Cortex. JMRI 1994; 4(4):537-543 41 12 L R. Schad, P Won/, M.V. Knopp, et al functional 21) and 31) Magnetic Resonance Imaging of Motor Cortex Stimulation at High Spatial Resolution Using Standard 15 T Imager Magn Reson. Imaging 1994; 12:9-15 13.1) I) Stark. W G Bread ley, Jr Magnetic Resonance Imaging vol I Mosby Year Book, 1992 14 D.N Levin, X Hu and K K fan Surface of the Brain: Three-dimensional MR Images Created with Volume Rendering Radiology 1989. 171 277-280 15 M Levoy Volume Rendering II I I: CG&A 1988; 8(3) 29-37. 16 M L Phelps. I) I- Kuhl Metabolic Mapping of the Brain's Response to Visual Stimulation Studies in Humans Scienee 1981, 211(27): 1445-1448 17 K D Merboldt, W Hanickc and J Prahm functional MR1 of Human Brain Activity at High Spatial Resolution J 1993,139-144 18 Prichard, I) Rothman. 12 Novotny, et al Lactate rise detected by (H NMR in Human Visual Cortex During Physiologic Stimulation Proc Natl Acad Sci. 1991, 8 8 : 5829-5831. 19 D.N. Levin, X Hu, K K I an, et al I he Brain Integrated Three-Dimensional Display of MR and PHT Images Radiology 1989. 172 783-789 20 J W Belliveau, K.K. Kwong, D.N. Kennedy, et al Magnetic Resonance Imaging Mapping of Brain Function: Human Visual Cortex Invest Radiol 1992; 27:S47-S53. 21 P. I'den, J M Maaisog, P Jezzard, et al A comparison of PH!T and MR1 in the Neuroanatomical Localization of Visual Processing. Book of Abstracts 13th SMRM 691 22 J K Udup>a. D Odhner, S Samarasekera. et al The 3DViewnix Software System User Manual. Medical Image Processing Ciroup. University of Pennsylvania, 1993 23 W J. Huk, (i. (iademann, G Kriedmann. MR1 of Central Nervous System Diseases. Springer-Verlag, 1989. 42 INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter free, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, prim bleedthrough, substandard margin*, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note win indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the bade of the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. A B«i> & Hoveif Information Company 300 North Z eeb Road Ann Arbor, M l 48106-1346 USA 313/761-4700 000/521-0600 UMI Number: 1376514 UMI Microfora 1376514 Copyright 1995, by UMI Coapany. All rights reserved. This aicrofora edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Processing And Visualization In Functional Magnetic Resonance Imaging (Fmri) Of The Human Brain

PDF

Respiratory system impedance at the resting breathing frequency range

PDF

Functional water MR spectroscopy of stimulated visual cortex using single voxel

PDF

Auditory brainstem responses (ABR): variable effects of click polarity on auditory brainstem response, analyses of narrow-band ABR's, explanations

PDF

Estimation of upper airway dynamics using neck inductive plethysmography

PDF

Cross-correlation methods for quantification of nonlinear input-output transformations of enural systems using a Poisson random test input

PDF

A physiologic model of granulopoiesis

PDF

Propofol Effects On Eeg And Levels Of Sedation

PDF

Auditory brainstem responses (ABR): quality estimation of auditory brainstem responsses by means of various techniques

PDF

Pulse oximetry failure rates

PDF

Decomposition Of Neuronal Function Using Nonlinear Systems Analysis

PDF

Flourine-19 NMR probe design for noninvasive tumoral pharmacokinetics

PDF

A multimodal screen reader for the visually impaired

PDF

Design of a portable infrared spectrometer: application to the noninvasive measurement of glucose

PDF

A kinetic model of AMPA and NMDA receptors

PDF

Comparison of evacuation and compression for cough assist

PDF

Human Skeletal Muscle Oxygenation And Perfusion: Non-Invasive Measurement By Near-Infrared Spectroscopy

PDF

A computational model of NMDA receptor dependent and independent long-term potentiation in hippocampal pyramidal neurons

PDF

Cellular kinetic models of the antiviral agent (R)-9-(2-phosphonylmethoxypropyl)adenine (PMPA)

PDF

An open ear canal sound delivery system

Asset Metadata

Creator Shen, Tayi (author)

Core Title Three-dimensional functional mapping of the human visual cortex using magnetic resonance imaging

School School of Engineering

Degree Master of Science

Degree Program Biomedical Engineering

Degree Conferral Date 1994-12

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag biology, neuroscience,engineering, biomedical,health sciences, radiology,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Singh, Manbir (committee chair), Khoo, Michael Chee-Kuan. (committee member), Yamashiro, Stanley M. (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c18-4887

Unique identifier UC11357919

Identifier 1376514.pdf (filename),usctheses-c18-4887 (legacy record id)

Legacy Identifier 1376514-0.pdf

Dmrecord 4887

Document Type Thesis

Rights Shen, Tayi

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA