Functional magnetic resonance imaging
| Seiji Ogawa and Yul-Wan Sung (2007), Scholarpedia, 2(10):3105. | doi:10.4249/scholarpedia.3105 | revision #151126 [link to/cite this article] |
Functional magnetic resonance imaging (fMRI) of the brain is a non-invasive way to assess brain function using MRI signal changes associated with functional brain activity. The most widely used method is based on BOLD (Blood Oxygenation Level Dependent) signal change that is due to the hemodynamic and metabolic sequelae of neuronal responses.
One of the most important points for fMRI in investigating human brain function rests on the fact that brain function is spatially segregated, i.e. specific functions are localized at various sites. This functional specialization can be identified with fMRI and mapped at high spatial resolution. BOLD-fMRI has been widely used in various fields of brain science to identify areas in the brain as the neural basis of their corresponding mental behaviors.
Contents |
BOLD based fMRI method
BOLD effects in MR images
Hemoglobin without bound oxygen molecules, deoxyhemoglobin, is paramagnetic because of the high spin state (S = 2) of the heme iron. In contrast, oxygen-bound hemoglobin, oxyhemoglobin, has low spin (S = 0) and is diamagnetic (Pauling & Coryl 1936). The presence of deoxyhemoglobin in red blood cells makes their magnetic susceptibility different from the diamagnetic plasma in blood and, similarly, induces a difference in magnetic susceptibility between the blood and the surrounding tissue. In the large homogenous magnetic fields used in MRI, compartmentalized susceptibility differences induce small magnetic field distortions in the blood as well as in the surrounding extra-vascular area. Water protons in these areas sense these field distortions, which are reflected in the signal decay process, characterized by T2 (spin echo) or T2*(gradient echo) relaxation. When the content of deoxyhemoglobin changes in the blood, the relaxation process of water protons is modified and one can see these changes in MRI. The image intensity that varies with deoxyhemoglobin content has been termed Blood Oxygenation Level Dependent (BOLD) and was suggested for potential use in functional study of the brain by Ogawa et al (1990). It should be noted that Thulborn et al showed, in their in vitro blood experiments, that blood water T2 varies with deoxyhemoglobin content (Thulborn et al 1982).
The era of BOLD-based functional MRI (using the endogenous contrast agent, deoxyhemoglobin) started with three papers that appeared in 1992 (Bandettini et al 1992, Kwong et al 1992, Ogawa et al 1992) (see Raichle 2000 for a historical perspective on fMRI development).
Prior to these reports, another fMRI method for detection of a functional response in the human brain was published in 1991 (Belliveau et al 1991), using an exogenous contrast agent injected into the blood stream. The measurement was only possible during the time window when the agent passed through areas of interest in the brain.
BOLD MR image signal and its relation to physiology
The BOLD effect is related to changes in physiological conditions (Ogawa et al, 1998) and appears as a part of the relaxation rate \(1/T_2^*\ .\)
\[ \frac{1}{T_2^*}=\frac{1}{T_{20}^*}+\frac{1}{T_{2B}^*} \]
\(1/T_{2B}^*\) is the part due to BOLD effect and \(1/T_{20}^*\) is the sum of other terms of the transverse relaxation rate. MRI signal (S) can be expressed by
\[ S=S_0e^{\frac{-TE}{T_2^*}} \]
where \(S_0\) is the signal at the echo time \(TE = 0 \) (spin density or \(T_1\) dependent term). For spin echo signal, \(T_2^*\) is replaced by \(T_2\ .\) The fractional signal change from the resting state to the activated state or induced by changes in the physiology is (\(\Delta \left(1/T_{20}^*\right) = 0\))
\[\tag{1} \frac{\Delta S}{S}=\frac{\Delta S0}{S0}-TE\Delta\left(\frac{1}{T_{2B}^*}\right) \]
The first (TE independent) term is the so-called spin density change and includes contributions from inflow effects ( \(T_1^*\)). When \(TE\) is long, the second term (BOLD effect) dominates. The relation of the MRI parameter, \(T_{2B}^*\ ,\) to the BOLD susceptibility field can be estimated when there is no other microscopic level magnetic field variation interfering BOLD field.
The \(1/T_{2B}^*\)term varies with the content of deoxyhemoglobin in the imaging voxel (Ogawa et al 1998). When the susceptibility field does not change much over the diffusion distance of water molecules (about 10 \(\mu m\) in TE period), the intra-voxel signal average is a simple and static average. This is the case for the signal in extra-vascular space, around a large vessel. The \(1/T_{2B}^*\)component is linearly related to the deoxyhemoglobin content
\[ \frac{1}{T_{2B}^*}=A\left(CBVv\right)C_{Hb}\left(1-Y\right) \] where A is a constant related to susceptibility parameters, \(C_{Hb}\)is the total hemoglobin concentration in the blood, \(Y\) is the hemoglobin oxygenation level and \(CBVv\) is the cerebral blood volume fraction that contains deoxyhemoglobin.
When the susceptibility field varies within the water diffusion distance, the averaging of the intra-voxel signals is dynamic and covers the blood water signal and the signal around capillary. In this instance, the \(1/T_{2B}^*\) is nearly quadratic in the deoxyhemoglobin content in the blood.
\[ \frac{1}{T_{2B}^*}=B\left(CBVv\right)\left\{C_{Hb}\left(1-Y\right)\right\}^2 \]
The change \(\Delta\)(1/\(T_{2B}^*\)) upon neuronal activation for both types of average is \[\tag{2} \Delta\left(\frac{1}{T_{2B}^*}\right)=\frac{1}{T_{2B}^*}\left\{\frac{\Delta CBVv}{CBVv}-a\frac{\Delta Y}{\left(1-Y\right)}\right\} \]
where, \(a=1\) for static average, \(a=2\) for fast dynamic average.
It can be seen from (1) and (2), that MRI signal change is related to the physiological parameters. It is interesting to note that in small anesthetized animals, the CBV change induced by neuronal activation has been reported to come mainly from the arterial side, where the blood does not carry any deoxyhemoglobin. If this holds in the conscious human brain, then \(\Delta CBVv = 0\) in (2).
In high-field MRI, the extra-vascular space contribution is important and at 1.5T the blood water contribution is the major part.
BOLD effect signal changes in (2) can be related to the energy metabolism at activated areas. The regional oxygen balance between the supply and the demand can be described by the Fick’s principle. \(OE=CBF\cdot C_{Hb}\cdot\left(1-Y\right)\) where CBF is the cerebral blood flow and OE is (oxygen extraction). Then the oxygen extraction fraction (OEF) at steady state is \[ OEF=\frac{OE}{CBF\cdot C_{Hb}}=\left(1-Y\right) \]
The change in OEF with variation of the oxygen balance is \[\tag{3} \Delta OEF= -\Delta Y \]
or in another expression \[\tag{4} \frac{\Delta\left(OE\right)}{OE}-\frac{\Delta CBF}{CBF}= \frac{\left( -\Delta Y\right)}{\left(1-Y\right)} \]
Therefore, OEF decreases (more oxygen supply than the consumption) make \(\Delta Y\) positive. This situation corresponds to the case when \(\Delta CBF/CBF\) is larger than the fractional oxygen extraction change and the corresponding BOLD signal increases.
The energy metabolism of the brain is known to be highly oxidative at the resting state, using glucose as the carbon source.
The change in the glucose consumption and CBF upon neural activation has been reported to be \(\Delta CMRglu/CMRglu =\Delta CBF/CBF\ ,\) where \(CMRx\) is the cerebral metabolic rate of \(x\ .\) Then the positive BOLD signal (\(\Delta Y > 0\)) requires that \(\Delta CMRglu/CMRglu > \Delta CMRO2/CMRO2\) and the energy metabolism during neural activation has to be less oxidative.
The neural basis of BOLD signals and neuro-vascular coupling
The neuronal processes causing BOLD signal changes are associated with synaptic inputs at the site of activation, not with the output level of firing of the neuron receiving synaptic inputs (Logothetis et al 2001). This means that fMRI reflects the synaptic activity driving neuronal assemblies, but cannot disclose the information content of the neuronal firing patterns produced by the neurons.
The coupling to synaptic activity is very tight. It is obvious that the metabolic load change is coupled directly to the neuronal activation. Furthermore, the vascular response is also coupled very tightly. One of those coupling mechanisms involves glutamate released for synaptic activation, which induces a change in Ca+2 in neighboring astrocytes. This results in the release of blood vessel dilators (at the contact point of astrocyte to arterioles and through rapidly diffusible substances like Nitric Oxide) and an increase in CBF.
Such tight coupling of BOLD signal to synaptic activity allows us to use the fMRI signal to probe functional responses in the brain, even though the response time of the BOLD signal (or any other signals based on vascular changes) is over several seconds and much slower than the underlying neuronal processes.
Positive and negative BOLD changes
The positive BOLD signal change has been shown to correspond to excitatory activation. An associated increase in the field potential or the appearance of evoked potentials has been observed at the sites where positive BOLD changes are marked. For the positive BOLD ( \(\Delta Y > 0\ ,\) hyper-oxygenation), local neuronal activities at the site of activation are responsible.
In addition to the positive and localized BOLD signal, there are areas that show negative BOLD although their magnitude is in general relatively small. There are several possible explanations for this. A likely candidate is a decrease or suppression of local synaptic activity relative to a control state. In this case, the CBF and CMRO2 both decrease so as to yield \(\Delta Y < 0\ .\) There are some areas that often show negative BOLD to various stimuli such as BA 31 / 7 in the parietal area. The neurobiological mechanisms for these phenomena are still not well understood. There seems to be no case of definite observation where some active inhibition process causes positive BOLD, although active inhibition requires synaptic activity.
With physiological variations in cerebral circulation, it is possible for the BOLD signal to change without local neural activity changes. Such changes are likely to be diffuse; expressed in large areas and not well synchronized with neuronal events.
Advantages and disadvantages of BOLD fMRI
In fMRI applications to functional mapping of the brain, BOLD signal acquisition with gradient echo is most popular because of its relatively high sensitivity (0.5-3% signal change by neural activation) and the simplicity of measurement. BOLD signal acquired by gradient echo measurement, however, has a limited spatial resolution because the signal includes contributions from veins draining the sites of activation. It has been reported that the orientation column structure, in the primary visual area in the cat brain, can be resolved in CBV-based fMRI but cannot be seen with the gradient echo BOLD signal. A further disadvantage of BOLD measurements is the often-observed contamination with large surface vessel signals. Activation-induced signal changes in these can reach very large value of 10-20%, especially when nearby activated areas send their more oxygenated venous blood to a draining surface vessel.
In intra-vascular signals, the BOLD effect is present regardless of the size of the vessel in both gradient and spin echo acquisitions and the signals contribute to the voxel signal (unless they decay out at a long value of echo-time). The intra-vascular signal is the major component at 1.5 T MRI. The BOLD effect with spin echo, in the extra-vascular space around capillaries, is relatively small but becomes important at super-high field MRI. Since the capillary area signal has better functional specificity, it can be advantageous to use such high fields (e.g., above 7T) where BOLD-fMRI at higher spatial resolution can also be performed.
Another main disadvantage of BOLD signals, which is common to all measurements based on vascular changes, is the slow response time (i.e., seconds). If the neural events to be measured are happening slowly, they can be tracked with fMRI since the measurement is in real time. When events occur in short time scales relative to the fMRI response time, the overlap of evoked fMRI signals make it difficult to resolve individual events. With the slow response of fMRI, it is difficult to study fast dynamics of neural processing which proceeds in tens to hundreds of milliseconds.
To measure the latency of an evoked response, one needs help from other non-invasive methods capable of detecting fast dynamics such as MEG and EEG. Some temporal relation between functional systems may be probed by BOLD when there is some inter-system interaction. One needs to look for event related signal changes with a well defined paradigm which contains a few inputs and that can be deconvolved from the BOLD time series (Sung et al, 2007)
Non-BOLD measurements
To overcome the disadvantage of BOLD signal, various non-invasive measurements have been developed. CBF and CBV measurements, both of which are based on the vascular response have better specificity for localizing functional activity in capillaries or functional activity linked blood volume change. Both show the response time to be shorter than BOLD signal by about a second, which corresponds to the transit time of the blood from arterial side to venous side of the capillary bed.
Recently, it has been shown that MRI signals, with an application of strong field gradients as in water diffusion measurement, responds to neuronal activation and the signal change has good spatial specificity (Le Bihan 2006). This response appears to reflect a change in the apparent diffusion constant of water in the tissue. It could be interpreted in terms of cellular swelling caused by evoked neural activity. This phenomenon is interesting but more study is needed to establish the mechanism of this signal change and its characteristics.
It has been an ambition of many people to detect electro-magnetic events caused by neural activity directly using MRI. Such direct detection of current induced local magnetic field or ion movements would provide the means to study neuronal events at high spatial and temporal resolution. Despite active efforts with many experiments using in vitro or model systems, the in vivo detection of electrical events still eludes MR spectroscopists.
Low frequency oscillation of fMRI signal and spontaneous activity at rest
When a long MRI time series data are analyzed in terms of frequency distribution, one can see the oscillation power is largely in the low frequency region, far below respiration rate. There are some peaks at 0.1Hz or at a lower frequency. Such 0.1Hz oscillations used to be attributed to so-called vaso-motion, of the sort seen in in vivo optical measurements. Any vascular modulation could lead to CBF variations. If this is the case, the modulation is not likely due to the local neuronal activity, but some signal to the vascular system from remote areas. However, the presence of connectivity between functionally related sites was shown by correlations between these low frequency oscillations in time series MRI data at resting state (Biswal et al 1995). Furthermore, it has recently been reported that there is a slow modulation of the power of neural oscillations in the gamma range; such modulations can induce low frequency BOLD signal variation (Leopold et al 2003). Slow oscillations of power may reflect long range coordination in a functional network. As reviewed by Fox and Raichle (2007), spontaneous fluctuations of fMRI signals at resting state have been explored to find functional networks among functional sites on the basis of the connectivity. Simultaneous measurements of fMRI and EEG have provided electrophysiological correlates of spontaneous fMRI activity. On the other hand, it is often observed that such a slow fMRI oscillation superimposes on stimulus induced signal activation without any apparent mutual interaction. There is yet more to learn on the physiological bases of slow signal oscillations in fMRI (Fox and Raichle 2007).
Activation maps and functional networks
The use of non-invasive neuro-imaging methods including BOLD-fMRI has been accepted as an approach to elucidate the neural bases of many kinds of sensorimotor and mental processes in neuroscience. For example, fMRI has been used to study functional specialization for visual recognition of objects with what/where information (Ungerleider & Pasternak 2003). Similarly language (Démonet et al 2005) and music perception have been topics of intensive study. Applications of fMRI cover many other high-order functions in the human brain, elucidating the neural attributes of mental processes involved in psychological phenomena such as emotion and feelings. This application has been extended to even wider areas; into the humanities and social sciences, where the neural bases of human moral judgment (Greene and Haidt 2002), of economic decision making, and of the “human mind” or “self” have been addressed.
A major factor in the success in of activation studies with BOLD fMRI should be attributed to the development of dedicated statistical data processing (Friston et al 1995), which has provided easy access to the methodology in spite of the fact that the evoked fMRI signal is fairly weak relative to noise (i.e., inevitable random variations of signal response within and among subjects). The ensuing functional maps represent an ensemble average over subjects selected from the human population. With the help of statistical procedures, specific functional roles of activated sites are often examined by some comparison of fMRI responses between carefully chosen paradigms that are nearly identical but with targeted distinction. In such a comparison, many low-level activation sites as well as common functional sites are eliminated by the criteria of statistically significant differences in evoked responses.
To understand how the brain works we need to know the functional network of sites participating in a functional task. The constellation of observed sites constitutes a major part of these networks. Some of high-order cognitive tasks generate neuronal oscillations. Simultaneous recording of fMRI and EEG data can show the presence of the functional network integrated by neural oscillations. The question there is how and in which order the sites in the network interact each other to transfer information. We need new ways to answer this question even with multimodal imaging with fMRI and EEG.
It is of course necessary to know the functional specificity of each activated site to understand distributed networks in the brain. Many of functional sites are activated with multiple functional inputs, although these inputs may have common features. Without the knowledge of the functional specificity and local functional architecture, the working of the network can only be understood in anecdotal terms. With non-invasive neuroimaging, the functional role or specificity of a site is only established indirectly. This is because we cannot measure the actual input to or output from a site. The only means we have for controlling the site’s activation is through the external or internal stimulus we give to the brain. It is not known which aspects of the original stimulus are delivered to the site or how site-specific processing proceeds. Until we understand the information processing entailed by the local input-output relationships, we can only try to infer this processing by clever manipulations of the stimuli we give to the brain.
References
Bandettini P.A., Wong E.C., Hinks R.S., Tikofsky R.S. and Hyde J.S., (1992) Time course EPI of human brain function during task activation Magn. Res. Med. 25 390-7
Belliveau J.W., Kennedy D.N., McKinstry R.C, Buchbinder B.R., Weisskoff R.M., Cohen M.S., Vevea J.M., Brady T.J. and Rosen B.R., (1991) Functional mapping of the human visual cortex by magnetic resonance imaging. Science. 254 (5032) 716-9.
Biswal B., Yetkin F., Haughton V. and Hyde J., (1995) Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn. Res. Med. 34, 537–541
Démonet J.F., Thierry G. and Cardebat D., (2005) Renewal of the Neurophysiology of Language:Functional Neuroimaging Physiol Rev 85:49-95
Friston K.J., Holmes A., Worsley K.J., Poline J-B., Frith C.D. and Frackowiak R.S.J. (1995) Statistical parametric maps in functional imaging: a general linear approach. . Hum Brain Map 2: 189–210
Fox M.D. and Raichle M.E., (2007) Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging. Nature Review Neuroscience 8 700-711
Greene J. and Haidt J., (2002) How (and where) does moral judgment work? Trends Cogn Sci. 6 517-523
Kwong K.K., Belliveau J.W., Chesler D.A., Goldberg I.E., Weisskoff R.M., Poncelet B.P., Kennedy D.N., Halpemet Halpemet.E.F. and Rosen B.R., (1992) Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. Proc. Natl. Acad. Sci. 89 5675-9
Le Bihan D., Urayama S., Aso T., Hanakawa T., and Fukuyama H., (2006) Direct and fast detection of neuronal activation in the human brain with diffusion MRI. PNAS 103 8263-8268
Leopold, D. A., Murayama, Y. & Logothetis, N. K.(2003) Very slow activity fluctuations in monkey visual cortex:implications for functional brain imaging. Cereb.Cortex 13 423–433 (2
Logothetis N.K., Pauls J., Augath M., Trinath T. and Oeltermann A.(2001) Neurophysiological investigation of the basis of the fMRI signal. Nature 412 150-7
Ogawa S., Menon R.S., Kim S.G. and Ugerbil K., (1998) On the characteristics of functional magnetic resonance imaging of the brain. Annual Review of Biophysics and Biomolecular Structure 27 447-474
Ogawa S., Lee T.M., Kay A.K. and Tank D.W., (1990) Brain Magnetic Resonance Imaging with Contrast Dependent on Blood Oxygenation Proc. Natl. Acad. Sci. (USA), 87, 9868-9872
Ogawa S., Tank D.W., R. Menon R.S., Ellermann J.M., Kim S. G., Merkle H. and Ugurbil K., (1992), Intrinsic Signal Changes Accompanying Sensory Stimulation: Functional Brain Mapping With Magnetic Resonance Imaging. Proc. Natl. Acad. Sci. (USA), 89, 5951-5955
Pauling L. and Coryell C. D., (1936) The magnetic properties and structure of hemoglobin, oxyhemoglobin and carbonmonoxy hemoglobin. Proc. Nayl. Acad. Sci. (USA) 22, 210-216
Raichle M. E., (2000) A brief history of human functional brain mapping. In Brain Mapping: The systems edited by Toga A.W. and Mazziotta J.C., Academic Press, pp33-75
Sung Y.W., Kamba M. and Ogawa S., (2007) An fMRI study of the functional distinction of neuronal circuits at the sites on ventral visual stream co-activated by visual stimuli of different objects. Exp Brain Res. 181 657-63
Thulborn K.R., Warterton C.J., Matthews P.M. and Radda G.K., (1982) Oxygenation dependence of the transverse relaxation time of water protons in whole blood at high field. Biochim Biophys Acta; 714:265-70
Ungerleider L.G. & Pasternak T.,(2003) Ventral and Dorsal Cortical Processing stream. In The Visual Neurosciences edited by. Chalupa & Werner, MIT press, Vol. 1 pp 541-562
Internal references
- Jan A. Sanders (2006) Averaging. Scholarpedia, 1(11):1760.
- Valentino Braitenberg (2007) Brain. Scholarpedia, 2(11):2918.
- James Meiss (2007) Dynamical systems. Scholarpedia, 2(2):1629.
- Paul L. Nunez and Ramesh Srinivasan (2007) Electroencephalogram. Scholarpedia, 2(2):1348.
- William D. Penny and Karl J. Friston (2007) Functional imaging. Scholarpedia, 2(5):1478.
- Mark Aronoff (2007) Language (linguistics). Scholarpedia, 2(5):3175.
- Peter Jonas and Gyorgy Buzsaki (2007) Neural inhibition. Scholarpedia, 2(9):3286.
- Jeff Moehlis, Kresimir Josic, Eric T. Shea-Brown (2006) Periodic orbit. Scholarpedia, 1(7):1358.
- Arkady Pikovsky and Michael Rosenblum (2007) Synchronization. Scholarpedia, 2(12):1459.
See Also
Electroencephalogram, Event Related Brain Dynamics, Magnetoencephalogram, MRI, Functional Imaging, Neurovascular Coupling
