fmri97.no S Y N O P S I S FIRST NORWEGIAN SYMPOSIUM ON FUNCTIONAL MAGNETIC RESONANCE IMAGING OF THE BRAIN Bergen fmri group BERGEN, NORWAY

Transcription

1 fmri97.no S Y N O P S I S Bergen fmri group FIRST NORWEGIAN SYMPOSIUM ON FUNCTIONAL MAGNETIC RESONANCE IMAGING OF THE BRAIN BERGEN, NORWAY MAY 15-16, LATEX2e - al970507

2 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, fmri97.no Functional magnetic resonance imaging (fmri) of the brain is the application of MRI techniques to investigate cerebral physiology while preserving anatomic specificity. As such, fmri is becoming a tool both in clinical neuromedicine and in cognitive neuroscience. The purpose of the symposium, which is initiated by the fmri group in Bergen, is to present some of the latest theory and experimental techniques in fmri of the brain. The symposium is intended to be a good starting point for clinicians and researchers who are planning to set up clinical fmri examinations or neurocognitive experiments on a clinical scanner. The symposium is relevant to neurologists, neurosurgeons, neuropsychologists, psychiatrists, and radiologists as well as researchers from physics, statistics, neurobiology, and experimental psychology. The symposium will also facilitate communication between researchers who already have experience with fmri. Symposium committee: Lars Ersland, PhD Department of Clinical Engineering, Haukeland University Hospital Alf-Inge Smievoll, MD Department of Radiology, Haukeland University Hospital Roger Barndon, R.T. Department of Radiology, Haukeland University Hospital Lars Thomassen, MD Department of Neurology, Haukeland University Hospital Kenneth Hugdahl, PhD Department of Biological and Medical Psychology, University of Bergen Håkan Sundberg, PhD Department of Biological and Medical Psychology, University of Bergen Arvid Lundervold, MD PhD Department of Physiology, University of Bergen Symposium secretariat: Torill Myrdal / fmri97.no Department of Clinical Engineering Haukeland University Hospital N-5021 Haukeland, Norway Phone: Fax: Torill.Myrdal@medtek.haukeland.no WWW: The symposium is supported by the Research Council of Norway (NFR-MH), University of Bergen, Haukeland University Hospital, Nycomed Imaging AS, and Medical Equipment Aps.

3 PROGRAM. First Norwegian Symposium on Functional Magnetic Resonance Imaging of the Brain Radisson SAS Hotel Norge, Bergen, Norway, May 15-16, 1997 Wednesday, 14th Registration Thursday, 15th Registration Chairman: Kenneth Hugdahl Welcome Dean of Faculty of Medicine, Prof. Jon Lekven Head of Department of Radiology, Prof. John Ludvig Larsen Contrast Mechanisms in fmri Prof. Peter Bandettini Coffee break Fast Imaging Techniques with Applications to fmri - non-epi Dr. Claudia Österle Break Fast Imaging Techniques with Applications to fmri - EPI Dr. Claudia Österle Lunch Chairman: Arvid Lundervold Spatial and Temporal Resolution Prof. Peter Bandettini Whole Brain fmri Data Analysis Prof. Robert Cox Coffee break Towards Real-Time fmri Sessions Prof. Robert Cox Break Software Demonstration Prof. Robert Cox et al Symposium dinner

4 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, Friday, 16th -. Chairman: Håkan Sundberg Applications of Functional MRI to Human Brain Mapping Prof. Jeffrey Binder Coffee break Functional MRI Studies of Language Processing Prof. Jeffrey Binder Break Functional Brain Mapping Using 3D MRI, EEG, and Dipole Source Analysis Prof. Helmut Buchner Lunch Chairman: Lars Thomassen Stroke Imaging in Man and Animal Dr. Olav Haraldseth Coffee break Perspectives on Contrast Agents in Functional MRI Studies Dr. Olav Haraldseth Panel discussion Adjourn

5 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, Contents 1 Contrast Mechanisms in fmri (Peter Bandettini) Blood Volume Blood Perfusion Blood Oxygenation Embedded Contrast General Pulse Sequence Modulations Fast Imaging Techniques with Applications to fmri - Non-EPI (Claudia Österle) Basics of Image Generation in MR Non - EPI - Techniques FLASH Improving the Temporal Resolution: Reduced k-space Techniques Improving the Volume Coverage: MUSIC and PRESTO RARE Functional Spectroscopy Fast Imaging Techniques with Applications to fmri - EPI (Claudia Österle) EPI Sequence Examples of BOLD Sensitive EPI Studies Comparison EPI - FLASH for fmri Spatial and Temporal Resolution (Peter Bandettini) Temporal Resolution Spatial Resolution Whole Brain fmri Data Analysis (Robert Cox) The Problem The Signal: Pattern Matching The Signal: Pattern Hunting The Signal: Multiple Comparisons and Spatial Models The Noise: Subject Motion The Noise: Physiological Fluctuations The Noise: Scanner Fluctuations The Noise: Variance and Covariance Practical Points Towards Real-Time fmri Sessions (Robert Cox) The Goal The Rationale Software Issues The Methods

6 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, The Progress The Plans Applications of Functional MRI to Human Brain Mapping (Jeffrey Binder) 21 8 Functional MRI Studies of Language Processing (Jeffrey Binder) 22 9 Functional Brain Mapping Using 3D-MRI, EEG and Dipole Source Analysis (Helmut Buchner) Somatosensory Evoked Potentials Visual Evoked Potentials Epileptic Spikes Stroke Imaging in Man and Animal (Olav Haraldseth) Perspectives on Contrast Agents in Functional MRI Studies (Olav Haraldseth) 27 A INVITED SPEAKERS 28 B PARTICIPANTS (registered before May 7th) 29

7 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, Contrast Mechanisms in fmri (Peter Bandettini) An important goal in fmri research is the characterization of the coupling between neuronal activity, metabolism, hemodynamic changes, and subsequent changes in the MRI signal. The utility of fmri as a brain mapping tool is directly related to the degree to which the relationship between MRI signal changes and underlying neuronal activation is established. Research on these topics should increase the utility and interpretability of fmri and other hemodynamic - based brain imaging methods. Several types of cerebrovascular information can be mapped using MRI. The tomographic information that can be obtained include: a) maps of cerebral blood volume (1, 2, 3, 4) and cerebral perfusion (5, 6, 7, 8, 9, 10, 11), and b) maps of changes in blood volume (12), perfusion (7, 8, 9, 13, 14, 15), and oxygenation (13, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26). Below is a description of how these various hemodynamic properties are selectively detected using fmri. 1.1 Blood Volume A technique developed by Belliveau and Rosen et al. (1, 2, 3) utilizes the susceptibility contrast produced by intravascular paramagnetic contrast agents and the high speed imaging capabilities of echo planar imaging (EPI) to create maps of human cerebral blood volume (CBV). A bolus of paramagnetic contrast agent is injected (the technique is slightly invasive) and T2 or T2* - weighted images are obtained at the rate of about one image per second using echo-planar imaging (EPI) (7, 27, 28, 29). As the contrast agent passes through the microvasculature, susceptibility gradients (magnetic field distortions) are transiently produced. These gradients, which last the amount of time that it takes for the bolus to pass through the cerebral vasculature, cause intravoxel dephasing, resulting in a signal attenuation which is linearly proportional to the concentration of contrast agent (1, 2, 30), which, in turn is a function of blood volume. Changes in blood volume that occur during hemodynamic stresses or during brain activation can then be created by subtraction of two maps: one during a resting state and one during a hemodynamic stress or neuronal activation (12). The use of this method marked the first time that hemodynamic changes accompanying human brain activation were mapped with MRI. 1.2 Blood Perfusion An array of new techniques now exist for mapping cerebral blood perfusion in humans. The MRI techniques are similar to those applied in other modalities such as positron emission tomography (PET) and single photon emission computed tomography (SPECT) in that they all involve arterial spin labeling. The MRI based techniques hold considerable promise of high spatial resolution without the requirement of contrast agent injections. They use the fundamental idea of magnetically tagging arterial blood outside the imaging plane, and then allowing flow of the tagged blood into the imaging plane. The RF tagging pulse is usually a 180=B0 pulse that inverts the magnetization. Generally, these techniques can be subdivided into those which use continuous arterial spin labeling, which involves continuously inverting blood flowing into the slice (5, 31), and those which use pulsed arterial spin labeling, periodically inverting a block of arterial blood and measuring the arrival of that blood into the imaging slice. Examples these techniques are: 1) echo planar imaging with signal targeting and alternating RF, (EPISTAR), which involves alternately inverting slabs of magnetization above and below the imaging slice (7, 14), and 2) flow-sensitive alternating inversion recovery, (FAIR), which involves the alternation between slice selective and non slice selective inversion. The

8 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, latter was introduced by Kwong et al. (9, 15, 32) and referred to as FAIR by Kim et al. (8). Variations of these techniques have since been introduced (10, 11, 33). In the case of the pulsed techniques, pairwise subtraction of sequential images with and without the application of the RF tag outside the plane gives a perfusion related signal. Variation of the delay time between the inversion or tag outside the imaging plane and the acquisition of the image gives perfusion maps highlighting blood at different stages of its delivery into the imaging slice. Because there is necessarily a gap between the proximal tagging region and the imaging slice, there is a delay in the time for tagged blood to reach the arterial tree, this delay time can be highly variable, ranging from about 200 ms to about 1 sec for a gap of 1 cm. At 400 ms, typically only blood in larger arteries has reached the slice and the pulsed arterial spin labeling signal is dominated by focal signals in these vessels, while at 1000 ms, tagged blood has typically begun to distribute into the capillary beds of the tissue in the slice. Images acquired at late inversion times can be considered qualitative maps of perfusion. For the application of mapping human brain activation, (i.e. to only observe activation-induced changes in blood perfusion), a more commonly used flow sensitive method is performed by application of the inversion pulse always in the same plane. In this case, the intensity of all images obtained will be weighted by modulation of longitudinal magnetization by flowing blood and also by other MR parameters that normally contribute to image intensity and contrast (proton density, T1, T2). Therefore, this technique allows only for observation of changes in flow that occur over time with brain activation. This technique was first implemented by Kwong et al. (13) to observe activation - induced flow changes in the human brain. In this seminal paper, activation - induced signal changes associated with local changes in blood oxygenation were also observed. 1.3 Blood Oxygenation In 1990, pioneering work of Ogawa et al. (20, 21, 22) and Turner et al. (24) demonstrated that MR signal in the vicinity of vessels and in perfused brain tissue decreased with a decrease in blood oxygenation. This type of physiological contrast was coined blood oxygenation level dependent (BOLD) contrast by Ogawa et al. (22). The use of BOLD contrast for the observation of brain activation was first demonstrated in August of 1991, at the 10 th Annual Society of Magnetic Resonance in Medicine meeting (34). The first papers demonstrating the technique, published in July 1992, reported human brain activation in the primary visual cortex (13, 23) and motor cortex (13, 17). Two (13, 17) of the first three reports of this technique involved the use of single shot EPI at 1.5 Tesla. The other (23) involved multishot fast low angle shot (FLASH) imaging at 4 Tesla. Generally, a small local signal increase in activated cortical regions was observed using gradient echo pulse sequences - which are maximally sensitive to changes in the homogeneity of the main magnetic field. The working model constructed to explain these observations with gradient-echo imaging was that an increase in neuronal activity causes local vasodilatation which, in turn, causes an increase in blood flow. This results in an excess of oxygenated hemoglobin beyond the metabolic need, thus reducing the proportion of paramagnetic deoxyhemoglobin in the vasculature. This hemodynamic phenomenon was previously suggested using non-mri techniques (35, 36, 37). A reduction in deoxyhemoglobin in the vasculature causes a reduction in magnetic susceptibility differences in the vicinity of veinuoles, veins and red blood cells within veins, thereby causing an increase in spin coherence (increase in T2 and T2*), and therefore an increase in signal in T2* and/or T2 - weighted sequences. This effect scales with field strength.

9 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, Presently, the most widely used fmri technique for the non-invasive mapping of human brain activity is gradient-echo imaging using BOLD contrast. The reasons for this are that a) gradient-echo T2* - sensitive techniques have demonstrated higher activation-induced signal change contrast, by about a factor of two to four, than T2 - weighed, flow-sensitive, or blood volume-sensitive techniques, and b) BOLD contrast can be obtained using more widely available high speed multi-shot non-epi techniques. c) While T2* - weighted techniques are sensitive to blood oxygenation changes in vascular structures that include large vessels that may be spatially removed from the focus of activation, for most applications the sacrifice in functional contrast to noise ratio in techniques more sensitive to microvascular structures does not outweigh the necessity for a for the highest possible contrast to noise in functional images. 1.4 Embedded Contrast Recently, several methods for obtaining both perfusion and oxygenation information simultaneously have been implemented (38, 39). These offer the unique possibility to more directly study coupling between neuronal activity and blood flow changes. 1.5 General Pulse Sequence Modulations While it is generally accepted that MR signal changes are transduced through neuronally - induced hemodynamic changes, the following two relationships are not clear: a) the relationship between the magnitude, timing and spatial extent of neuronal activation and the magnitude, timing, and spatial extent of the hemodynamic changes, and b) the relationship between the degree, timing and spatial extent of induced hemodynamic changes and the degree, timing and spatial extent of the MR signal changes. The difficulty in activation - induced signal change contrast mechanism research is that the actual location, magnitude, and timing of neuronal activation is imprecisely known. In general, contrast mechanism studies have involved well controlled modulation of many potentially significant parameters. By parameter modulation and subsequent comparison with an ever-growing model which includes MR physics, cerebral physiology, and neurology, a large amount of convergent information about the relative contributions to MR signal changes has been and continues to be obtained. Strategies for a more complete understanding of fmri contrast mechanisms have involved: a) observation of the dependence of the magnitude, timing, and spatial extent of activation - induced signal changes upon MR parameters such as TE (23, 40, 41, 42, 43, 44, 45, 46), TI (7), flip angle (47), Bo (23, 25, 43, 44, 48), slice thickness and resolution (45, 49, 50, 51), outer volume saturation (52), and diffusion weighting (53, 54, 55, 56) b) observation of the resting signal, and activation - induced signal changes, during different degrees of hemodynamic stress (20, 21, 22, 24, 57, 58, 59, 60, 61, 62) or during different degrees of neuronal activity (visual flicker rate, finger tapping rate, syllable presentation rate) (13, 39, 63, 64, 65). c) comparison of signal locations with macroscopic vessel maps (44, 49, 50, 66, 67) or neuronal activation maps (68, 69, 70) and d) modeling of activation - induced MR signal changes based upon knowledge of MR physics and human cerebral physiology (30, 71, 72, 73, 74, 75, 76, 77, 78). References 1. B. R. Rosen, J. W. Belliveau, D. Chien, Magn. Reson. Quart 5, (1989). 2. B. R. Rosen, J. W. Belliveau, J. M. Vevea, T. J. Brady, Magn. Reson. Med. 14, (1990). 3. J. W. Belliveau, et al., Magn. Reson. Med. 14, (1990).

10 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, C. T. W. Moonen, P. C. M. vanzijl, J. A. Frank, D. LeBihan, E. D. Becker, Science 250, (1990). 5. D. S. Williams, J. A. Detre, J. S. Leigh, A. S. Koretsky, Proc. Natl. Acad. Sci. USA 89, (1992). 6. J. A. Detre, J. S. Leigh, D. S. WIlliams, A. P. Koretsky, Magn. Reson. Med. 23, (1992). 7. R. Edelman, P. Wielopolski, F. Schmitt, Radiology 192, (1994). 8. S.-G. Kim, Magn. Reson. Med. 34, (1995). 9. K. K. Kwong, et al., Magn. Reson. Med. 34, (1995). 10. E. C. Wong, R. B. Buxton, L. R. Frank, Quantitative imaging of perfusion using a single subtraction (QUIPSS), 2 nd Int. Conf. of Func. Mapping of the Human Brain, Boston (1996). 11. E. C. Wong, R. B. Buxton, L. R. Frank, NMR in Biomedicine (in press). 12. J. W. Belliveau, et al., Science 254, (1991). 13. K. K. Kwong, et al., Proc. Natl. Acad. Sci. USA. 89, (1992). 14. R. R. Edelman, B. Sievert, P. Wielopolski, J. Pearlman, S. Warach, JMRI 4(P), [Abstr.], 68 (1994). 15. K. K. Kwong, Magn. Reson. Quart. 11, 1-20 (1995). 16. M. K. Stehling, F. Schmitt, R. Ladebeck, JMRI 3, (1993). 17. P. A. Bandettini, E. C. Wong, R. S. Hinks, R. S. Tikofsky, J. S. Hyde, Magn. Reson. Med. 25, (1992). 18. A. M. Blamire, et al., Proc. Natl. Acad. Sci. USA 89, (1992). 19. J. Frahm, H. Bruhn, K.-D. Merboldt, W. Hanicke, D. Math, JMRI 2, (1992). 20. S. Ogawa, T.-M. Lee, A. S. Nayak, P. Glynn, Magn. Reson. Med. 14, (1990). 21. S. Ogawa, T.-M. Lee, Magn. Reson. Med 16, 9-18 (1990). 22. S. Ogawa, T. M. Lee, A. R. Kay, D. W. Tank, Proc. Natl. Acad. Sci. USA 87, (1990). 23. S. Ogawa, et al., Proc. Natl. Acad. Sci. USA. 89, (1992). 24. R. Turner, D. LeBihan, C. T. W. Moonen, D. Despres, J. Frank, Magn. Reson. Med. 27, (1991). 25. R. Turner, et al., Magn. Reson. Med. 29, (1993). 26. P. Jezzard, et al., NMR in Biomedicine 7, (1994). 27. P. Mansfield, J. Phys. C10, L55-L58 (1977). 28. M. K. Stehling, R.Turner, P. Mansfield, Science 254, (1991). 29. M. S. Cohen, R. M. Weisskoff, Magnetic Resonance Imaging 9, 1-37 (1991). 30. R. M. Weisskoff, C. S. Zuo, J. L. Boxerman, B. R. Rosen, Magn. Reson. Med. 31, (1994). 31. J. A. Detre, et al., in Diffusion and Perfusion: Magnetic Resonance Imaging D. LeBihan, Ed. (Raven Press, New York, 1995) pp K. K. Kwong, D. A. Chesler, R. M. Weisskoff, B. R. Rosen, Perfusion MR imaging, Proc., SMR, 2nd Annual Meeting, San Francisco (1994). 33. E. C. Wong, R. B. Buxton, L. R. Frank, Quantitative perfusion imaging using EPISTAR and FAIR, Proc., ISMRM 4 th Annual Meeting, New York (1996). 34. T. J. Brady, Future prospects for MR imaging, Proc., SMRM, 10th Annual Meeting, San Francisco (1991). 35. A. Grinvald, R. D. Frostig, R. M. Siegel, E. Bratfeld, Proc. Natl. Acad. Sci. USA 88, (1991). 36. R. D. Frostig, E. E. Lieke, D. Y. Ts o, A. Grinvald, Proc. Natl. Acad. Sci. USA 87, (1990). 37. P. T. Fox, M. E. Raichle, Proc. Natl. Acad. Sci. USA 83, (1986). 38. E. C. Wong, P. A. Bandettini, Two embedded techniques for simultaneous acquisition of flow and BOLD signals in functional MRI, Proc., ISMRM 4 th Annual Meeting, New York (1996). 39. P. A. Bandettini, E. C. Wong, Analysis of embedded-contrast fmri: interleaved perfusion, BOLD, and velocity nulling, Proc., ISMRM 5th Annual Meeting, Vancouver (1997). 40. P. A. Bandettini, E. C. Wong, R. S. Hinks, L. Estkowski, J. S. Hyde, Quantification of changes in relaxation rates R2* and R2 in activated brain tissue, Proc., SMRM, 11th Annual Meeting, Berlin (1992). 41. P. A. Bandettini, E. C. Wong, A. Jesmanowicz, R. S. Hinks, J. S. Hyde, Simultaneous mapping of activation - induced R2* andr2 in the human brain using a combined gradient-echo and spin-echo EPI pulse sequence, Proc., SMRM, 12th Annual Meeting, New York (1993). 42. P. A. Bandettini, E. C. Wong, A. Jesmanowicz, R. S. Hinks, J. S. Hyde, NMR in Biomedicine 7, (1994).

11 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, P. A. Bandettini, et al., MRI of human brain activation at 0.5 T, 1.5 T, and 3 T: comparisons ofr2* and functional contrast to noise ratio, Proc., SMR, 2nd Annual Meeting, San Francisco (1994). 44. R. S. Menon, S. Ogawa, D. W. Tank, K. Ugurbil, Magn. Reson. Med. 30, (1993). 45. J. R. Baker, et al., The effect of slice thickness and echo time on the detection of signal changes during echo-planar functional neruroimaging, Proc., SMRM, 11th Annual Meeting, Berlin (1992). 46. J. Frahm, K. D. Merboldt, W. Hanicke, A. Kleinschmidt, H. Steinmetz, High-resolution functional MRI of focal subcortical acivity in the human brian. Long - Echo Time FLASH of the Lateral Geniculate Nucleus During Visual Stimulation, Proc., SMRM, 12th Annual Meeting, New York (1993). 47. J. Frahm, K.-D. Merboldt, W. Hanicke, A. Kleinschmidt, H. Boecker, NMR in Biomedicine 7, (1994). 48. A. Jesmanowicz, P. A. Bandettini, E. C. Wong, G. Tan, J. S. Hyde, Spin-echo and gradient-echo EPI of human brain function at 3 Tesla, Proc., SMRM, 12th Annual Meeting, New York (1993). 49. E. M. Haacke, et al., NMR in Biomedicine 7, (1994). 50. S. Lai, et al., Magn. Reson. Med 30, (1993). 51. J. Frahm, K.-D. Merboldt, W. Hanicke, Magn. Reson. Med. 29, (1993). 52. J. H. Duyn, C. T. W. Moonen, G. H. vanyperen, R. W. d. Boer, P. R. Luyten, NMR in Biomedicine 7, (1994). 53. J. L. Boxerman, et al., Magn. Reson. Med. 34, 4-10 (1995). 54. A. W. Song, E. C. Wong, P. A. Bandettini, J. S. Hyde, The effect of diffusion weighting on task-induced functional MRI, Proc., SMR, 2nd Annual Meeting, San Francisco (1994). 55. J. L. Boxerman, R. M. Weisskoff, K. K. Kwong, T. L. Davis, B. R. Rosen, The intravascular contribution to fmri signal change. Modeling and diffusion-weighted in vivo studies, Proc., SMR, 2nd Annual Meeting, San Francisco (1994). 56. R. S. Menon, et al., Comparison of spin-echo EPI, asymmetric spin-echo EPI and conventional EPI appled to functional neuroimaging. The effect of flow crushing gradients on the BOLD signal, Proc., SMR, 2nd Annual Meeting, San Francisco (1994). 57. H. Bruhn, et al., J.Cereb Blood Flow Metab. 14, (1994). 58. S. Ogawa, T. M. Lee, B. Barrere, Magn. Reson. Med. 29, (1993). 59. A. J. decrespigny, M. F. Wendland, N. Derugin, E. Kozniewska, M. E. Moseley, Magn. Reson. Med. 27, (1992). 60. A. J. decrespigny, M. F. Wendland, N. Derugin, Z. S. Vexler, M. E. Moseley, JMRI 3, (1993). 61. P. A. Bandettini, et al., Hypercapnia and hypoxia in the human brain: effects on resting and activation-induced MRI signal, Proc., SMR, 2nd Annual Meeting, San Francisco (1994). 62. F. Prielmeier, Y. Nagatomo, J. Frahm, Magn. Reson. Med. 31, (1994). 63. J. R. Binder, et al., Cognitive Brain Research 2, (1994). 64. P. A. Bandettini, et al., The functional dynamics of blood oxygen level dependent contrast in the motor cortex, Proc., SMRM, 12th Annual Meeting, New York (1993). 65. S. M. Rao, et al., J. Cereb. Blood Flow and Metab. 16, (1996). 66. A. T. Lee, G. H. Glover, C. H. Meyer, Magn. Reson. Med. 33, (1995). 67. A. T. Lee, C. H. Meyer, G. H. Glover, JMRI 3(P), [Abstr.], (1993). 68. J. A. Sanders, J. D. Lewine, J. S. George, A. Caprihan, W. S. Orrison, Correlation of FMRI with MEG., Proc., SMRM, 12th Annual Meeting, New York (1993). 69. A. Connelly, et al., Radiology 188, (1993). 70. C. R. Jack, et al., Radiology 190, (1994). 71. R. P. Kennan, J. Zhong, J. C. Gore, Magn. Reson. Med. 31, 9-21 (1994). 72. C. R. Fisel, et al., Magn. Reson. Med. 17, (1991). 73. D. A. Yablonsky, E. M. Haacke, Magn. Reson. Med 32, (1994). 74. S. Ogawa, et al., Biophysical J 64, (1993). 75. J. L. Boxerman, R. M. Weisskoff, B. E. Hoppel, B. R. Rosen, MR contrast due to microscopically heterogeneous magnetic susceptibility: cylindrical geometry, Proc., SMRM, 12th Annual Meeting, New York (1993). 76. E. C. Wong, P. A. Bandettini, A deterministic method for computer modelling of diffusion effects in MRI with application to BOLD contrast imaging, Proc., SMRM, 12th Annual Meeting, New York (1993). 77. J. L. Boxerman, L. M. Hamberg, B. R. Rosen, R. M. Weisskoff, Magn. Reson. Med. 34, (1995). 78. P. A. Bandettini, E. C. Wong, International Journal of Imaging Systems and Technlogy 6, (1995).

12 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, Fast Imaging Techniques with Applications to fmri - Non-EPI (Claudia Österle) 2.1 Basics of Image Generation in MR An introduction to the most important features in MRI: image-space and k-space; spin echoes (SE) and gradient echoes (GE); longitudinal and transversal relaxation; pulse program schemes and k-space trajectories, etc. For a very thorough introduction, see for example [1]. 2.2 Non - EPI - Techniques FLASH FLASH (Fast Low Angle Shot) [2] is a fast GE-technique and the basis for many non-epit2-sensitive techniques. Imaging parameters shall be discussed with reference to functional MRI experiments. A severe disadvantage of FLASH for use in fmri is its long acquisition time leading to low temporal resolution and limited volume coverage. The next two sections show different ways to improve these limitations Improving the Temporal Resolution: Reduced k-space Techniques Imaging time can be significantly reduced if only a part of k-space is scanned in one experiment. To understand methods that exploit this fact, a basic comprehension of Fourier Transform is necessary. Acquiring only the center of k-space will lead to fast but blurred image updates. Keyhole Imaging [3] is a well established tool in contrast enhanced perfusion imaging and uses a high resolution reference image to improve the intrinsic low resolution of such fast image updates. However, difference images of a time series of Keyhole images still have low spatial resolution. The IKE-algorithm (Improvement of Keyhole Effect images) [4] enhances the spatial resolution of such difference images using information from the reference data set and low resolution difference images. RIGR (Reduced-encoding Imaging by Generalized-series Reconstruction) [5] offers an alternative reconstruction method for images acquired with the keyhole strategy, based on the generalized series model Improving the Volume Coverage: MUSIC and PRESTO MUSIC (Multislice Interleaved Excitation Cycles) [6] uses the longtethat is necessary for the requiredt2contrast for an interleaved excitation of several additional slices. Depending on the gradient strength, up to 7-16 slices can be examined in the same total experiment time as a single slice FLASH experiment. In PRESTO (PRinciples of Echo-Shifting with a Train of Observation) [7] the same slice is excited several times during the long echo time, which means thatteis longer thantr. The echo shifting leads to a shorter acquisition time and thus can be used for improved volume coverage especially in 3D-sequences RARE RARE imaging (Rapid Acquisition with Relaxation Enhancement) [8] was initially developed as a very fastt2-sensitive multi-se-technique. Additional contrast preparations enable the use of RARE

13 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, imaging for multiple applications.t2-sensitivity is realised by a delay between the excitation and acquisition periods [9]. As a single shot SE-technique, RARE offers a short acquisition time and the good imaging properties of a SE-generated image Functional Spectroscopy In functional spectroscopy the time evolution of the water signal from a small voxel in the activated brain region is measured. Comparison of signals in the resting and stimulated state allows the examination of the signal response in a time interval of ms after onset of the stimulus and lead to the discovery of a fast negative response [10]. Compared to imaging techniques functional spectroscopy has the advantages of improved SNR and high temporal resolution. References [1] D. D. Stark, W. G. Bradley, Magnetic Resonance Imaging. 2nd Ed. Mosby-Year Book, St. Louis, 1992 [2] A. Haase, J. Frahm, W. Hänicke, K.D. Merboldt, FLASH imaging. Rapid NMR imaging using low flip angle pulses. J. Magn. Res. 67, (1986). [3] J.J. van Vaals, M.E. Brummer, W.T. Dixon, H.H. Tuithof, H. Engels, R.C. Nelson, B.M. Gerety, J.L. Chezmar, J.A. den Boer, Keyhole method for accelerating imaging of contrast agent uptake. J. Magn. Res. Imaging 3, , (1993). [4] C. Oesterle, J. Hennig, Improvement of spatial resolution of keyhole effect images. in Proc., ISMRM, 4th Scientific Meeting, New York, 1996, p [5] A.G. Webb, Z.-P. Liang, R.L. Magin, P.C. Lauterbur, Applications of reduced-encoding MR imaging with generalizedseries reconstruction (RIGR). J. Magn. Reson. Imaging 3, , (1993). [6] T. Loenneker, F. Hennel, J. Hennig, Multislice interleaved excitation cycles (MUSIC): an efficient gradient-echo technique for functional imaging, Mag. Reson. Med. 35, , (1996). [7] G. Liu, G. Sobering, J. Duyn, C.T. Moonen, A functional MRI technique combining principles of echo shifting witt a train of observations (PRESTO), Mag. Reson. Med. 30, , (1993). [8] J. Hennig, A. Nauerth, H. Friedburg, RARE-Imaging: A fast method for clinical MR. Magn. Res. Med. 27, (1986). [9] J. Hennig, F. Hennel, C. Oesterle, O. Speck, C. Janz, J.F. Nedelec, Fast and robust measurements of brain activation using modified RARE sequences with variable contrast. in Proc., SMR, 2nd Scientific Meeting, San Francisco, 1994, p [10] T. Ernst, J. Hennig, Observation of a fast response in functional MR. Magn. Res. Med. 32, (1994).

14 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, Fast Imaging Techniques with Applications to fmri - EPI (Claudia Österle) 3.1 EPI Sequence Echo planar imaging (EPI) [1] was first introduced in During the last years it has become more and more important to a wide spread group of MR scientists due to the increasing availability of MR scanners with the required hardware. In contrast to multishot conventional sequences, EPI scans the entire k-space in one acquisition period after a single excitation. It offers a very short acquisition time and intrinsict2contrast and is therefore well suited for fmri experiments. The most important disadvantages of EPI however are its strong sensitivity to chemical shift and susceptibility induced artefacts that can lead to severe distortions in the images. 3.2 Examples of BOLD Sensitive EPI Studies Work in progress on our 1.5T Siemens Magnetom Vision includes a study about the reproducibility of BOLD effect results measured with a simple fist closure paradigm, examinations of the visual cortex with regard to motion perception [2], and fmri in patients with tumors affecting the sensorimotor area [3]. 3.3 Comparison EPI - FLASH for fmri A recent study compared fmri results obtained from EPI and FLASH acquired in one experimental session [4]. This allows a direct evaluation of the advantages and disadvantages of both methods. References [1] P. Mansfield, A.A. Maudsley, Planar spin imaging by NMR. J. Magn. Res. 27, (1977). [2] F.M. Krämer, K. Singh, A.T. Smith, M.W. Greenlee, J. Hennig, Echo-planar imaging of visual cortex during second order motion perception. in Proc., ISMRM, 5th Scientific Meeting, Vancouver, 1997, p [3] A. Schreiber, S. Ziyeh, U. Hubbe, J. Spreer, M. Büchert, R. Scheremet, Functional MRI in patients with intracerebral tumors affecting the sensorimotor area. in Proc., ISMRM, 5th Scientific Meeting, Vancouver, 1997, p [4] M. Büchert, S. Ziyeh, J. Hennig, O. Heid, E. Müller, Comparison of functional MRI with EPI and FLASH on a 1.5T MR-System. in Proc., ISMRM, 4th Scientific Meeting, New York, 1996, p

15 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, Spatial and Temporal Resolution (Peter Bandettini) The maximum temporal and spatial resolution of fmri can only be fully realized by the combination of: a high contrast to noise ratio, hemodynamic specificity, and well controlled and carefully executed experiments. 4.1 Temporal Resolution Specifically, temporal resolution can be referred to: a) the minimal stimulus duration necessary to cause a signal change, b) the minimal inter stimulus interval necessary to elicit activation from a single cortical region, c) the minimal difference in time that can be discriminated between activation of separate regions, d) the accuracy to which the activation onset is determined, e) the minimal time necessary for the fmri signal to significantly deviate from baseline, and f) The minimal time necessary to form a functional image. As of yet, a) the minimal stimulus duration necessary to cause a signal change is less than one second (1, 2, 3). b) the minimal inter stimulus interval has been determined to be about 3 sec on and 3 sec off (2, 3) or even faster when a small inter - stimulus cluster interval is allowed (4), c) the minimal difference in time that can be discriminated between activation of separate regions is on the order of a second (5, 6) or less (500 ms) when variations in the hemodynamic response over space are taken into account (7), d) the standard deviation of the measured activation onset is on the order of 200 ms, e) the minimal time necessary for the fmri signal to significantly deviate from baseline is about 1 to 2 seconds, and f) the minimal time necessary form a functional image is on the order of 10 seconds. For many types of investigations it may be desirable to use experimental paradigms similar to those used in event related potential recordings (ERP) or magneto - encephalography (MEG) (8), in which multiple runs of transient stimuli are averaged together. Impulse response cognitive fmri studies have recently been demonstrated (6). As a side note, because of the brief collection time of EPI relative to typical TR values (e.g. 50 ms relative to about 1 sec), the between-image waiting time allows for performance of EEG in the scanner during the imaging session without electrical interference from MR pulse sequences (9). 4.2 Spatial Resolution The upper limit on functional spatial resolution, similar to the limit on temporal resolution, is likely determined not by MRI resolution limits but by the hemodynamics through which neuronal activation is transduced. Evidence from in vivo high resolution optical imaging of the activation of ocular dominance columns (10, 11, 12) suggests that neuronal control of blood oxygenation occurs on a spatial scale of less than 0.5 mm. MR evidence suggests that the blood oxygenation increases that occur on brain activation are be more extensive than the actual activated regions (13, 14, 15, 16). In other words, it is possible that, while the local oxygenation may be regulated on a submillimeter scale, the subsequent changes in oxygenation may occur on a larger scale due to a spill-over effect. In general, to achieve the goal of high spatial resolution fmri a high functional contrast to noise and reduced signal contribution from draining veins is necessary. Greater hemodynamic specificity, accomplished by proper pulse sequence choice (selective to capillary effects), innovative activation protocol design (phase-tagging) (17, 18), may allow for greater functional spatial resolution. If the contribution to activation-induced signal changes from larger collecting veins or arteries can be easily identified and/or eliminated, then, not only will the confidence in brain activation localization increase, but also the upper limits of spatial resolution will be determined by scanner resolution and functional

16 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, contrast to noise rather than variations in vessel architecture. Currently, voxel volumes as low as 1.2 microliters have been obtained by functional FLASH techniques at 4T (19), and experiments specifically devoted to probing the upper limits of functional spatial resolution, using spiral scan techniques, have shown that fmri can reveal activity localized to patches of cortex having a size of about 1.35 mm (17). These studies and others using similar methods (17, 18, 20, 21, 22), have observed a close tracking of MR signal change along the calcarine fissure as the location of visual stimuli was varied. The voxel dimensions typically used in single - shot EPI studies are in the range of 3 to 4 mm, in plane, and having 4 to 10 mm slice thicknesses. These dimensions are determined by practical limitations such as readout window length, sampling bandwidth, limits of db/dt, SNR, and data storage capacity. Other ways to bypass the practical scanner limits in spatial resolution include partial k-space acquisition (23) and multi - shot mosaic or interleaved EPI (23, 24, 25). References 1. R. L. Savoy, et al., Exploring the temporal boundaries of fmri: measuring responses to very brief visual stimuli, Book of Abstracts, Soc. for Neuroscience 24 th Annual Meeting, Miami (1994). 2. P. A. Bandettini, et al., The functional dynamics of blood oxygen level dependent contrast in the motor cortex, Proc., SMRM, 12th Annual Meeting, New York (1993). 3. P. A. Bandettini, et al., in Diffusion and Perfusion: Magnetic Resonance Imaging D. LeBihan, Ed. (Raven Press, New York, 1995) pp A. Dale, R. Buckner,. (1997). 5. J. R. Binder, et al., Analysis of phase differences in periodic functional MRI activation data, Proc., SMRM, 12th Annual Meeting, New York (1993). 6. R. L. Buckner, et al., Porc. Nat l. Acad. Sci. 93, (1996). 7. R. L. Savoy, et al., Pushing the temporal resolution of fmri: studies of very brief visual stimuli, onset variablity and asynchrony, and stimulus-correlated changes in noise, Proc., SMR 3rd Annual Meeting, Nice (1995). 8. W. W. Orrison, J. D. Lewine, J. A. Sanders, M. F. Hartshorne, Functional Brain Imaging (Mosby-Year Book, Inc., St. Louis, 1995). 9. J. R. Ives, S. Warach, F. Schmitt, R. R. Edelman, D. L. Schomer, EEG and Clinical Neurophys 87, (1993). 10. A. Grinvald, R. D. Frostig, R. M. Siegel, E. Bratfeld, Proc. Natl. Acad. Sci. USA 88, (1991). 11. R. D. Frostig, in Cerebral Cortex, vol. 10 A. Peters, K. S. Rockland, Eds. (Plenum Press, New York, 1994) pp R. D. Frostig, E. E. Lieke, D. Y. Ts o, A. Grinvald, Proc. Natl. Acad. Sci. USA 87, (1990). 13. A. T. Lee, G. H. Glover, C. H. Meyer, Magn. Reson. Med. 33, (1995). 14. J. Frahm, K.-D. Merboldt, W. Hanicke, A. Kleinschmidt, H. Boecker, NMR in Biomedicine 7, (1994). 15. E. M. Haacke, et al., NMR in Biomedicine 7, (1994). 16. S. Lai, et al., Magn. Reson. Med 30, (1993). 17. S. A. Engel, et al., Nature 369, 370 [erratum], 525, 106 [erratum] (1994). 18. M. I. Sereno, et al., Science 268, (1995). 19. K. Ugurbil, et al., Magn. Reson. Quart. 9, (1993). 20. W. Schneider, D. C. Noll, J. D. Cohen, Nature 365, (1993). 21. E. A. DeYoe, P. Bandettini, J. Neitz, D. Miller, P. Winans, J. Neuroscience Methods 54, (1994). 22. E. A. DeYoe, et al., Proc. Natl. Acad. Sci. 93, (1996). 23. M. S. Cohen, R. M. Weisskoff, Magnetic Resonance Imaging 9, 1-37 (1991). 24. K. Butts, S. J. Riederer, R. L. Ehman, R. M. Thompson, C. R. Jack, Magn. Reson. Med. 31, (1994). 25. G. C. McKinnon, Magn. Reson. Med. 30, (1993).

17 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, Whole Brain fmri Data Analysis (Robert Cox) 5.1 The Problem The detection, mapping, and quantification of neural activation with FMRI relies on extracting information from noisy measurements. In any such problem, it is necessary to understand both the signal that is to be detected, and the noise that interferes with the detection process. In the first part of this talk, I will discuss some characteristics of the FMRI signal and noise, and will explain why both parts of the problem are active areas of research. I will then turn to some practical issues in whole brain FMRI data analysis. 5.2 The Signal: Pattern Matching When one knows the form of the signal, then detection of activation is essentially a pattern matching problem: which parts of the brain show the expected pattern, and which do not? The most widely used of such approaches is the correlation method (of which thet-test is a special case). 5.3 The Signal: Pattern Hunting When there is no expected form to the received signal, then detection of activation involves looking for unknown temporal patterns. There are two methods widely used for this: principal components analysis, and fuzzy clustering. 5.4 The Signal: Multiple Comparisons and Spatial Models If each voxel is considered separately, so that any possible spatial activation pattern is allowed, then a large (10000) number of activation decisions must be made this is the multiple comparison problem. To ensure that few of these decisions produce false activations, the threshold for detection must be made very stringent, making the detection of weak activations unlikely. One way to overcome this is to restrict the allowable spatial activation patterns; for example, to detect only in clusters above a certain volume. 5.5 The Noise: Subject Motion Gross motion of subjects is a major problem in FMRI. If neighboring voxels differ in MR intensity by 10%, and the subject moves by 10% of a voxel dimension, then a 1% signal change will occur. This is at the same level as the activation signal changes, and can either produce false positives (if the motion correlates to the the stimulus) or false negatives (if the motion is noise-like ). Head restraints and image registration techniques can mitigate this problem to some extent, but the ideal solution has yet to be found. 5.6 The Noise: Physiological Fluctuations ForB01:5Tesla, most of the noise in an EPI time series is due to respiratory and cardiac effects. These effects can be measured separately, or using MR techniques, and can be filtered out of the data time series. 5.7 The Noise: Scanner Fluctuations Echo planar imaging pushes MR scanners to the edge of the hardware technology. Small instabilities that would not visibly affect structural images will manifest themselves in functional time series. The

18 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, timing between system components and the stability of the many power supplies are two important problem areas. 5.8 The Noise: Variance and Covariance The noise variance in different subjects, in the same subject on different occasions, and in the same subject in different brain regions, can be quite different. This means that the power of any detection method will vary. This puts some constraints on how FMRI activation results should properly be reported. Additionally, FMRI data are correlated in time (due to physiological fluctuations) and also in space (depending also on the imaging method used). This means that most detection methods, which assume white noise (uncorrelated in space and time) are overly optimistic. This is a difficult problem to overcome, since there usually isn t enough data to estimate the covariance parameters accurately. 5.9 Practical Points It is important to have several stimulus cycles present in each imaging run the more the better. A practical minimum is three. Fewer cycles will make detection more uncertain. Always examine the image time series for motion. Nearly cubical voxels are the best for brain mapping. Know your scanner! Do some stability tests on water phantoms. Collect some purely resting state data on human subjects. Time series of phase images are very useful tools. When just starting functional MRI, begin with some simple experiments: finger tapping (almost everyone has the needed equipment) and flashing lights. If you have problems with these, then you can adjust your scanner and protocols until they work well. Readings [1] Bandettini PA, Jesmanowicz A, Wong EC, and Hyde JS. Processing strategies for time-course data sets in functional MRI of the human brain. Magn. Reson. Med. 30: (1993). (Where the linear analysis of FMRI time series began.) [2] Cox RW. Notes on Motion and Functional MRI, and references therein. Available at [3] Ford I. Commentary and opinion: III. Some nonontological and functionally unconnected views on current issues in the analysis of PET datasets. J. Cereb. Blood Flow & Metab. 15: (1995). (Although this paper is about PET analysis, many of the cautions also apply to FMRI datasets.) [4] Forman SD, Cohen JD, Fitzgerald M, Eddy WF, Mintun MA, and Noll DC. Improved assessment of significant activation in functional magnetic resonance imaging (fmri): use of a cluster-size threshold. Magn. Reson. Med. 33: (1995). (Addresses to some extent both the issue of multiple comparisons and spatial correlations in the data.) [5] Cox, RW. AFNI: Software for analysis and visualization of functional magnetic resonance neuroimages. Comput. Biomed. Res. 29: (1996). (Description of AFNI, as of nearly 2 years ago.)

19 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, Towards Real-Time fmri Sessions (Robert Cox) 6.1 The Goal Development of human brain functional MRI into a flexible interactive tool for advanced neuroscience and clinical applications to make FMRI scanning sessions more brain scientist friendly. 6.2 The Rationale Scanning sessions will be able to proceed much more quickly when the investigator has immediate feedback on the quality of the functional results being obtained. For any clinical applications of FMRI, it will be essential to know that the desired results have been obtained before the patient leaves the scanner. For experimentalists, since their scanner time is highly limited, knowing the quality of their results will let them use their time allotment optimally. New classes of FMRI protocols will be possible when the investigator has up to the minute knowledge of results, and can use that knowledge to choose which stimulus to present to the patient or subject. This will allow customization of FMRI to the subject Exploration of new FMRI protocols will be much easier when the investigator can compare the effects of changing the stimulus to immediately previous results in the same subject. This will make it possible to design new experiments and form new hypotheses, both much more quickly than is currently possible. 6.3 Software Issues Real-time FMRI sessions should have all the analysis tools available that are practicable for the time constraints. This means that the image acquisition and initial analysis software must be integrated into a more general analysis and visualization tool. We are developing such a comprehensive system, in which real-time FMRI is simply another component. This system AFNI will be demonstrated at the Symposium (if all goes well). 6.4 The Methods To achieve the goal of interactive FMRI, functional displays must be available as an imaging run proceeds, not just at its end. Computational efficiency requires the use of recursive algorithms for the detection and quantification of functional activation. 6.5 The Progress Single slice recursive real-time FMRI was first accomplished at the Medical College of Wisconsin in March Distractions (such as development of AFNI, and lack of computer resources) caused little work to be done on this until recently. Our first real-time whole brain scan, integrated with AFNI, was achieved on 2 May The Plans Generalize the existing recursive algorithm to allow real-time fitting to more complex waveforms. Create methods for real-time data quality analysis and correction, including image registration and response variance analysis.

20 First Norwegian Symposium on fmri of the Brain - Bergen, May 15-16, Develop AFNI into a neuroscience console for interactive monitoring of FMRI scanning sessions. Readings [1] Cox RW, Jesmanowicz A, and Hyde JS. Real-time functional magnetic resonance imaging. Magn. Reson. Med. 33: (1995). (Details about the recursive algorithm to compute the correlation coefficient.)