Now in its seventeenth year, fMRI has become the tool of choice for the cognitive neuroscience community. It is essential for those interested understanding the functional correlates of behavior and disease in populations and, in a growing degree, individuals. Functional MRI has grown largely because of its noninvasiveness, relative ease of implementation, high spatial and temporal resolution, and importantly, signal fidelity. The fMRI signal is robust and for the most part, highly reproducible and consistent. Another development which further enabled the rapid spread of fMRI was the introduction in 1996 of echo planar imaging (EPI) as a standard capability on clinical scanners sold by most vendors.
Advancements in fMRI have been defined by higher resolution both in time and space, better understanding of the signal, more robust or sensitive processing methods, new applications, and new neuroscience or clinical findings with fMRI. What fMRI can do in terms of probing brain organization is growing rapidly and in some unanticipated directions. New insights into the limitations of fMRI - or rather, what fMRI cannot do, have been just as important as advances in what fMRI can do. A deepening collective understanding is shared that while fMRI can very effectively answer questions about what brain areas - at a specific spatial scale - are active in association with specific tasks, it cannot as effectively be used to address questions of mechanisms of cognition or provide insights into deep principals of how the brain is organized. It can perhaps be able to tell us what decisions you are going to make or what object you are looking at or even how you are feeling. it can also provide information regarding where in the brain processing related to (but not necessarily essential to) a specific processing task occurs. It might be able to provide sensitive probes, complementary to other clinical techniques to an array of diseases, and perhaps provide useful information for predicting future illnesses including psychological disorders or dementia.
Over the past year or so, several exciting advancements have come out - thus continuing to push the horizon of what fMRI can and might do. In addition, the emergence of a "standard practice" in fMRI for performing and reporting studies has come about(Poldrack et al., 2008
). Specific cutting edge findings include insights into the mechanisms of blood oxygenation level dependent (BOLD) contrast, new hypotheses about hemodynamic contrast, new fMRI contrast mechanisms, higher resolution and increased sensitivity, and new paradigms and methods for analyzing and generating data. All of these have led to surprising new applications and findings. Below, these advancements are discussed in detail.
a. Functional MRI Interpretation
Over the years, fMRI has been defined by a certain ambiguity about the origins of the functional signal. It is based in hemodynamics and therefore is only an indirect measure of brain activation. Even though it is extremely robust and consistent, many assumptions have to be made if more in depth interpretation of the signal is desired. The field has benefitted from efforts to better understand not only the neuronal correlates of fMRI but the sources of fMRI signal variability. In the past year, these efforts have continued - as indicated by a large number of new fMRI contrast mechanism papers as well as some excellent review articles.
To begin, a recently published opinion piece by Greg Miller in Science (G. Miller, 2008
) and a review article by Nikos Logothetis in Nature(Logothetis, 2008
) do an excellent job in laying out some limits and misuses of fMRI while also discussing its potential. At the heart of a long-recognized problem with some fMRI studies is the issue of data interpretation. Data interpretation can be confounded on several levels. First, the fMRI signal is based on the complex interaction of neuronal activity, neuronal metabolism, blood flow and blood volume on a spatial scale that lumps together hundreds of thousands of neurons in each MRI voxel. These factors may vary not only across subject populations, individuals, and regions in the brain but also across voxels and even within voxels. This hemodynamic variability poses a severe limit on how fMRI can be used. On an individual level, this limit prevents an investigator from suggesting that one part of the brain shows "more" activity simply because the fMRI signal is greater. Among other things, this signal intensity is highly weighted by the blood volume in each voxel(Bandettini & Wong, 1997
). Logothetis goes further in his review to suggest that the limits in our knowledge of precisely what neuronal activity (excitation, inhibition, sub threshold activity, top down or bottom up modulation) is manifest by the hemodynamic response limits the depth to which we can draw inferences even from parametrically modulated signal within a region. While most researchers have known these caveats all along, this particular one is a potentially significant confound in the interpretation of the many parametric studies that have been performed over the years.
Because of the variability of the timing of hemodynamics, we also have well understood limits (on the order of seconds) on the inferences of the relative timing of brain activity between regions. Without a clear understanding of the non-neuronal sources of this spread in latency, inferences about causality based on relative timing (below 2 second differences) are inherently weak and likely to be so weighted by the underlying vasculature-influenced timing that they are incorrect. A growing area in fMRI method development is that of hemodynamic calibration - to measure and subsequently account for the hemodynamic variability over space(Ances et al., 2008
; Chiarelli, Bulte, Wise, Gallichan, & Jezzard, 2007
; Hoge et al., 1999a
; Kida, Rothman, & Hyder, 2007
; Thomason, Foland, & Glover, 2007
). Nevertheless, the problem of parsing excitatory and inhibitory brain activity, for example, remains unable to be "calibrated" with fMRI.
Another perhaps more common problematic level of interpretation in fMRI is not that of inferring the level, degree, or timing of BOLD signal change but that of drawing unsubstantiated inferences about mental states or conditions from assessment of brain activation maps. Functional MRI is not immune to the problem of researchers creating the "just so" story that can explain any complicated pattern of activation shown in the data. The best fMRI studies are those with testable hypotheses, tightly controlled paradigms, appropriate calibrations, and careful inferences. It is typically correct to say "this area is associated with processing visual stimuli." It is much more risky (but not necessarily impossible) to say "Because there is more activation in this area, this subject prefers this political candidate more."
Regarding fMRI interpretation, a surprising amount of progress has been made in past year. Several papers are worth mentioning. First, an exciting study by Schummers et al(Schummers, Yu, & Sur, 2008
) describes in detail the spatial and temporal behavior of astrocyte activation relative to neuronal activation, and convincingly demonstrates that it is astrocytes that signal blood flow to increase with activation. Interestingly, this work shows that the spatial tuning of visual cortex astrocytes to orientation column stimulation is sharper than that of neurons. In addition, the activation timing of astrocytes is delayed about 4 seconds from that of neurons. To show astrocyte influence on hemodynamics, astrocyte activation was selectively modulated using isoflurane, which has less of an effect on neurons. When astrocyte activity was suppressed, the measured hemodynamic response to stimulation was also shown to be suppressed. These findings certainly shed light on the mechanism of fMRI contrast mechanisms and will cause us to refine our models of neurovascular coupling.
As is well known, considerable controversy continues as to the physiologic origin of several specific aspects of BOLD contrast, including BOLD nonlinearity, pre-undershoot, post-undershoot, and fluctuations. Many papers published this year have provided provocative data perhaps shedding light on several of these unknowns. In particular, Devor et al. (Devor et al., 2007
) provides a unique angle on decreased BOLD signal. Devor et al. shows, in a rat model, that when given somatosensory stimulation, a concentric pattern of neuronal depolarization (and corresponding increase in oxygenation) surrounded by hyperpolarization (and corresponding decrease in oxygenation) was created. The most intriguing part of this study was that, in the areas of hyperpolarization (and deoxygenation), which was up to 3 mm away from the central depolarization, there was a clear vasconstriction. This appears to imply that inhibitory synaptic activity, rather than a decrease in firing rate, drives vasoconstriction. These data appear to contradict the current understanding, supported by a study by Shmuel et al. (Shmuel et al., 2002
) that shows a clear decrease in spiking that spatially corresponds to a decrease in BOLD signal. If this turns out to be a general finding, this may not only explain negative signal changes (inhibitory synaptic activity in distinct locations in space) but perhaps also the post stimulus undershoot (inhibitory synaptic activity at distinct times following activation). This group also published an intriguing hemodynamic model called the vascular anatomical network model (VAN) (Boas, Jones, Devor, Huppert, & Dale, 2008
) that goes beyond the Balloon model(Buxton, Uludag, Dubowitz, & Liu, 2004
) in that it is based on physically measurable parameters and models effects both in time and over space. It has been able to accurately predict, based on low level input parameters, blood pressure in vascular compartments, and dynamics of blood pressure and hemoglobin saturation.
Regardless of these pieces of interesting data and insightful modeling, controversy still reigns. One clear cut example, touched on above, is that of the post stimulus undershoot. As is well understood, BOLD signal increases with an increase in blood oxygenation or decrease in blood volume. Three popular hypotheses for the post undershoot exist: The first is that there is a perseveration of increased blood volume, after blood flow and oxygenation have returned to baseline. The second is that there is a perseveration of oxidative metabolic rate after blood flow and volume have returned to baseline, thus reducing blood oxygenation. The third is that there is a post stimulus decrease in flow, causing blood oxygenation to decrease since the time for oxygen to be delivered to active tissue increases, thus enhancing delivery. The work by Devor et al (Devor et al., 2007
) , mentioned above, may suggest the latter explanation. Other groups have experimental evidence for a blood volume perseveration (Silva, Koretsky, & Duyn, 2007
), while yet another set of researchers have experimental evidence for oxidative metabolic rate perseveration (H. Z. Lu, Golay, Pekar, & van Zijl, 2004
) or just against blood volume perseveration (Frahm et al., 2008
). There exist different assumptions, experimental conditions, and spatial scales with all of these studies. As with most fMRI contrast mechanism work, the explanation for an effect, when discovered, is always more complicated than we expect it to be. To most people outside of the subspecialty of fMRI contrast mechanism studies, this seems like a somewhat esoteric discussion since the post stimulus undershoot is typically not even used as a source of contrast. Understanding it fully is nevertheless extremely important as this understanding may lend insight into other aspects of BOLD contrast and potentially lead to an enhanced capability of using BOLD contrast to understand neuronal activity.
Turning to negative signal changes, a common network showing negative signal changes in response to a wide range of cognitive tasks has been termed the default mode network(McKiernan, D'Angelo, Kucera-Thompson, Kaufman, & Binder, 2002
; Raichle et al., 2001
). An ongoing debate continues with regard to precisely what these negative signal changes imply, and what the functional role of this network is. One recent study combined spectroscopy and fMRI to further probe these areas(Northoff et al., 2007
). Specifically, relative concentrations of GABA (an inhibitory neurotransmitter) were measured in the anterior cingulate cortex (ACC) - one node of this negative signal change network - during a task which caused a decrease in signal there. The findings demonstrated that a clear negative relationship exists between GABA concentrations and the magnitude of BOLD signal decrease in the ACC. Both the previous study and this seem to be aspects of a similar evolving story - that inhibitory activity apparently leads to negative signal changes in the area showing the inhibitory activity. Another study combined EEG and fMRI data to probe the default mode(Scheeringa et al., 2008
). When EEG power at theta frequencies (2 to 9 Hz) was used a regressor for resting state fMRI signal, only negative correlations were found which spatially corresponded to the default mode network. The conclusion of this study was that increased frontal theta activity can be seen as an indicator of default mode network activity. There is more on theta activity in the MEG/EEG section of this review.
Functional MRI contrast continues to be compared with that of other modalities, with occasionally curious and surprising results. Spatial location, timing, and magnitude have been studied as they relate to EEG, MEG, optical imaging, and electrophsyiologic measures. Generally, these have been confirmatory studies, showing a general agreement in space, magnitude, and timing between modalities. A recent eye-opening report is not quite so confirmatory. In their study, Muthukumaraswamy and Singh compared fMRI and MEG Gamma frequency (40 to 60 Hz) responses to a visual stimulus which was varied in spatial frequency and temporal frequency(Muthukumaraswamy & Singh, 2008
). While the spatial overlap between brain activation measured with the two modalities was substantial, the parametric dependence of the signals varied between the two. This is shown in . For fMRI, nearly identical temporal frequency dependence curves were generated for both spatial frequencies. With MEG, high spatial frequency showed much higher power across temporal frequency than low spatial frequency. In addition, both temporal frequency dependence curves were generally flatter for MEG than for fMRI. These results strongly suggest that the two measures are tuned to different aspects of neuronal activity.
Figure 2 Obtained from (Muthukumaraswamy & Singh, 2008). This shows the temporal frequency tuning curves for peak responses in fMRI and MEG Gamma frequency (40×60Hz) data in primary visual cortex, clearly indicating a difference between fMRI and (more ...)
Taking the investigation of temporal frequency dependence further with high resolution BOLD contrast, Sun et al. demonstrate in a recent study (Sun et al., 2007
) a dramatic columnar organization in V1 of visual temporal frequency preference. This map is shown in . So the story of frequency dependence is complicated by this fine grained structure of temporal frequency responsiveness.
Figure 3 Obtained from (Sun et al., 2007). This is NOT an ocular dominance column map. It is an fMRI-derived temporal frequency domain map, showing the fine structure of areas in primary visual cortex that are either selective to high temporal frequency visual (more ...)
b. New Functional MRI Contrasts
As the imaging community comes to terms with the limits of fMRI, many are busy trying to develop an MRI-based functional contrast that may somehow be a more direct, sensitive, quantitative measure of neuronal activity. Indeed, many other functional contrasts have been put forward as possible. These include the non-invasive measures of activation induced changes in perfusion (Detre, Leigh, Williams, & Koretsky, 1992
), blood volume(H. Lu, Golay, Pekar, & Van Zijl, 2003
), diffusion(Le Bihan, Urayama, Aso, Hanakawa, & Fukuyama, 2006
; A.W. Song et al., 2003
(Davis, Kwong, Weisskoff, & Rosen, 1998
; Hoge et al., 1999b
), temperature (Yablonskiy, Ackerman, & Raichle, 2000
), and presumably magnetic field changes in the vicinity of active neurons(Bandettini, Petridou, & Bodurka, 2005
). These results have always been met with extreme interest by the imaging community but none have proven superior to BOLD - because the techniques are more technically challenging or, more often, the effect size produced using the specific contrast sensitivity is considerably less than BOLD contrast. Perfusion imaging has perhaps come the closest to BOLD in terms of utility and functional contrast. While it generally has less brain coverage, lower temporal resolution, and lower functional contrast to noise than BOLD, it does have the advantages of higher specificity, baseline information, and less baseline drift. The latter advantage is significant when probing very slow brain activation timing(Aguirre, Detre, Zarahn, & Alsop, 2002
Progress in an area may not always mean that a technique grows in utility or validity but rather that papers are published that increase our understanding of the results. In the past year, some progress has been made with regard to two types of new MRI-based brain activation contrast: The measurement of decreases in diffusion coefficient with brain activation and direct measurement of neuronal activity as indicated by MR phase or magnitude changes.
Depending on the degree of sensitization to diffusion, brain activation causes either an increase or decrease in measured diffusion coefficient. If a very small diffusion weighting is used - sensitizing the signal to extremely rapid random motion of spins - specifically that of capillary flow, then, hypothetically, if brain activation causes an increase in flow, this diffusion coefficient will increase and will be manifest as a signal decrease(A.W. Song, Woldorff, Gangstead, Mangun, & McCarthy, 2002
). If the diffusion weighting is much larger, the sensitization is weighted towards changes in diffusion in very slow moving spins. Hypothetically, with brain activation, a small amount of cell swelling occurs, increasing the proportion of presumably slower diffusing intracellular spins. Therefore, with brain activation and when using a large diffusion weighting, the measured diffusion coefficient will decrease, resulting in an increase in this highly diffusion weighted signal(Le Bihan et al., 2006
). It is hypothesized that this contrast is a more direct measure of neuronal activity. In recent years, this technique has received a considerable amount of publicity but has been difficult to replicate. Last year, two papers were published which contradicted the hypothesis that the origins of this diffusion coefficient change were in tissue(Jin & Kim, 2008
; K. L. Miller et al., 2007
). These papers suggested that the origin of these activation-related diffusion coefficient decreases was in the vasculature. Currently, the story remains to be fully resolved.
A second perhaps more popular potential functional contrast is that which allows detection of neuronal currents directly. Following a few results in phantoms (Bodurka & Bandettini, 2002
; Konn, Gowland, & Bowtell, 2003
), preliminary (and non-replicated) results in humans (Xiong, Fox, & Gao, 2003
), and some studies with cell cultures (Petridou et al., 2006
) and in intact systems (Park, Lee, & Park, 2004
) - all suggesting that direct detection of neuronal currents in humans might be possible, researchers in the field are taking a second, more in depth look at what hypothetically they are looking for and what the prospects realistically are for detection. The hypothesis for this effect is as follows: Neuronal activity produces a transient current, presumably in dendrite clusters, setting up a transient, small in extent, and very subtle magnetic fields superimposed on the existing magnetic field. The precise timing, extent, and magnitude of these fields is a source of debate and study. These small field inhomogeneities contribute to NMR phase shifts or phase dispersion, depending on the geometry. If the fields are larger and of a single orientation, a predominant phase shift would occur. If they are smaller and more random in orientation, phase dispersion occurs. Optimal contrast weighting for these so far has been long TE (TE=T2* of tissue) gradient echo, which unfortunately, is also the optimal contrast weighting for BOLD contrast. One of the reasons why clear detection is so difficult is because any hint of BOLD contrast has to be clearly ruled out.
In the past year, several outstanding papers have contributed to this effort to image and understand neuronal current effects in MRI. These include some extremely detailed models of magnetic fields produced by neuronal activity (Cassara, Hagberg, Bianciardi, Migliore, & Maraviglia, 2008
; Paley, Chow, Whitby, & Cook, 2008
), very carefully executed but unfortunately, negative findings (Tang, Avison, Gatenby, & Gore, 2008
), some highly innovative pulse sequences that may be more sensitive to periodic neuronal currents (Buracas, Liu, Buxton, Frank, & Wong, 2008
), and some novel ideas regarding the potential to detect neuronal currents using extremely low magnetic field strength scanners (Cassara & Maraviglia, 2008
; Kraus Jr, Volegov, Matlachov, & Espy, 2008
). At extremely low fields (i.e. fields such that the Larmor frequency is on the same order of magnitude as predominant brain frequencies as measured with MEG), the hypothesis is that brain activity at specific frequencies (i.e. 10 Hz to 60 Hz) would enhance spin relaxation for spins precessing at a corresponding frequency. This is an intriguing hypothesis that is certainly worth exploring.
A unique paper, tangentially related to fMRI contrast mechanisms, published this year is worth mentioning. Moore and Cao have put forth what they call the "Hemo-Neural" hypothesis stating that activation induced hyperemia (increase in blood flow) has a neuromodulatory function (Moore & Cao, in press
). In other words, an increase in neuronal activity increases blood flow, which, in turn further increases the gain of local cortical circuits, perhaps influencing detection and discrimination of sensory stimuli. They propose the following mechanism for this effect: The delivery of blood-borne messengers, mechanical modulation, thermal modulation, or hemodynamic modulation of astrocytes exerts an influence on neuronal activity. Evidence for some of these mechanisms is provided in their study. In general, this highly novel paper is certainly worth taking a look at, as, in the context of creating models of neuronal activity, if hemodynamic changes up regulate neuronal excitability significantly then their effects need to be taken into account.
c. MRI Technology
Just about every significant step in the advancement in fMRI applications has been made as a direct result of an advancement in MRI technology. MRI technology includes: magnetic field, rf coil conFiguration, and acquisition and/or image reconstruction strategies. Advancements in MRI technology include increases in field strength, increases in image signal to noise, increases in resolution, and increases in imaging speed. In the past year, several dramatic improvements in MRI technology have been published. These are discussed below.
A look at the field strength used for scanning humans over the years reveals a surprising linear trend, as shown in . A new MRI Center, NeuroSpin, has proposed developing a human 11.7 T scanner by 2012 - thus maintaining the linear trend. In addition, at the time of writing this review article, there are over 28 human 7T scanners in operation. A summary of the countries where the 7T scanners reside is shown in . The push for higher field strength is driven not only by a direct proportionality of sensitivity to field strength, but by two other related factors: With this increased sensitivity comes the ability to collect high SNR functional and anatomical images faster and at higher resolution - allowing investigation of more subtle signal changes or anatomic features or allowing scanning of volunteers or patients in a faster time. Also, the images at 7T are not simply higher signal to noise or higher resolution, but qualitatively different in terms of anatomical contrast. Qualitatively new anatomic contrast is apparent in 7T anatomical images. An example of the qualitative difference in contrast at 7T is a study by Duyn et al (Duyn et al., 2007
), which demonstrates that MRI phase is a highly sensitive contrast at 7T, revealing a 10 x improvement in gray matter-white matter delineation over magnitude contrast at lower field strengths, as well as cortical layers, and perhaps even white matter tracts. The authors speculate that the phase shifts are mostly due to differences in susceptibility effects, but other hypothesis have been put forward (Zhong, Leupold, von Elverfeldt, & Speck, 2008
) suggesting that macromolecules play a role. Examples of detailed, high contrast images at 7T are shown in .
Plot of the highest current MRI field strength used for human imaging as it has increased over the years, showing a surprising linear trend.
Pie chart showing the current distribution (as of the summer of 2008), by country, of 7 Tesla human scanners. (Data is courtesy of Hellmut Merkle, NINDS)
Figure 6 Obtained from (Duyn et al., 2007). This is an illustration of the image quality of magnitude and phase images obtained at 7T. The GRE data, left and center, had a resolution of 240 × 240 um with a scan duration of 6.5 min, whereas the MP-RAGE (more ...)
Performance of fMRI at high field is technically extremely challenging. Challenges include magnetic field inhomogeneity, RF power deposition in some sequences, RF excitation inhomogeneity, acoustic noise, increased physiologic noise, and patient discomfort, among others. Nevertheless the benefits are beginning to manifest themselves in fMRI. A recent publication by Yacoub et al (Yacoub, Harel, & Ugurbil, 2008
) has shown that activation of human orientation columns (about a third of the size of ocular dominance columns) can be imaged. These results are shown in . To achieve this, scanning was performed at 7T on a Siemens/Varian console using a four shot spin-echo sequence, resulting in an in-plane resolution of 0.5 × 0.5 mm2
. A novel finding from this study is that there is a bias toward processing orientations around the vertical direction. An important point also to be taken from this study is that at 7T, the intravascular contribution is reduced such that spin-echo sequences do not suffer from large vessel intravascular effects - as occurs at lower field strengths. This enables spin-echo sequences to have a smaller functional point spread function than gradient echo sequences(Shmuel, Yacoub, Chaimow, Logothetis, & Ugurbil, 2007
; Yacoub, Shmuel, Logothetis, & Uğurbil, 2007
), thus being critical for the imaging of orientation columns.
Figure 7 Obtained from (Yacoub et al., 2008). This shows ocular dominance columns for two subjects in panels a and b. Red and blue represent voxels that showed preference to right and left eye stimulation, respectively. Maps shown in panels c and d, show maps (more ...)
Other technological innovations have been in the direction of faster image acquisition. Currently, almost all functional MRI data involves the collection of a stack of 2D images which takes about 2 seconds. With higher bandwidth receivers, parallel RF coil arrays, stronger gradients, and novel image reconstruction strategies, the goal of obtaining a relatively high resolution volume of data in a single excitation or with a single acquisition is drawing closer. In the past year, two papers have described novel echo volume imaging (EVI) techniques. The first paper, by Lindquest et al (Lindquist, Zhang, Glover, & Shepp, 2008
), describes a method by which a central cubic portion of 3D raw data space is sampled in 100 ms, leading to a volume of low resolution (25 × 25 × 17 matrix size to 46 × 46 × 17 matrix size) but rapidly obtained data. This approach was then used to study transient dynamics of the hemodynamic response. First the pre-undershoot, presumably associated with a rapid increase in oxidative metabolic rate before an increase in flow, was observed. Secondly, in a task that involved motor cortex activation cued by a visual stimulus, a difference in onset time of the pre-undershoot was observed. The second paper, by Rabrait et al (Rabrait et al., 2008
), demonstrates how parallel acquisition, outer volume suppression, and SENSE reconstruction can be combined to obtain a 20 × 20 × 24 matrix size with a 120 × 120 × 144 mm3
FOV to create isotropic 6 mm voxel sizes. It is important to note that, in typical fMRI studies with voxel sizes of about 3 mm3
, spatial smoothing, normalization, and multi-subject averaging reduces the effective resolution to about this. In fact it is more optimal to acquire images initially at this lower resolution then to acquire at a higher resolution (and lower SNR) and then spatially smooth. It appears that in the near future, EVI will be a new standard in image acquisition for those researchers performing multi-subject averaging or interesting in studying subtle dynamics of the response across the brain.
Lastly, Lin et al (Lin et al., 2008
) and Hennig et al (Hennig, Zhong, & Speck, 2007
) have taken parallel acquisition to a new level in minimizing the need for time-consuming spatial encoding gradients. Lin et al, using a 32 channel parallel array with the coil configuration resembling EEG or MEG sensor geometries, apply similar source localization algorithms to reconstruct these images. They have named this approach magnetic resonance Inverse Imaging (InI). Hennig et al, have reduced the need for spatial encoding gradients even further with their one-voxel-one-coil (OVOC) imaging method. While the obvious advantage of these methods is the extreme rapidity with which useful volumes are obtained, another advantage is the presumably silent acquisition. This avenue of technology development appears to be just beginning. It's easy to imagine improvement in coil configurations optimized for specific brain areas or more efficient sampling schemes. Higher bandwidth acquisition rates will also lead to further improvements in resolution. Perhaps in the next several years, relatively silent, whole brain fMRI will be the standard.
d. fMRI Paradigms and Processing
The powerful approach of analyzing fMRI that involves extraction of subtle voxel-wise activation patterns rather than mapping blobs of activation has, in the past few years, seen a tremendous surge of interest due to some of the dramatic results that it has produced (Carlson, Schrater, & He, 2003
; Cox & Savoy, 2003
; Davatzikos et al., 2005
; Eger, Ashburner, Haynes, Dolan, & Rees, 2008
; Hanson, Matsuka, & Haxby, 2004
; Haxby et al., 2001
; Haynes & Rees, 2005a
; Haynes et al., 2007
; Kamitani & Tong, 2005
; Kriegeskorte & Bandettini, 2007a
; Kriegeskorte, Formisano, Sorger, & Goebel, 2007
; Kriegeskorte, Goebel, & Bandettini, 2006
; LaConte, Strother, Cherkassky, Anderson, & Hu, 2005
; Mitchell et al., 2004
; Mourao-Miranda, Bokde, Born, Hampel, & Stetter, 2005
; O'Toole, Jiang, Abdi, & Haxby, 2005
; Pessoa & Padmala, 2005
; Shinkareva et al., 2008
; Strothert et al., 2002
) (Friston et al., 2008
) This new class of techniques is generally known as multivariate pattern recognition, classification, or decoding: (i.e. aiming to identify a perceptual representation, activation, or cognitive state on the basis of multivoxel regional fMRI signals.) When a perceptual representation can be decoded from the activity pattern, the brain region studied contains information about the stimulus. In general, such “information-based” analysis requires multivariate techniques, but not necessarily decoding. In many instances "decoding" has been performed using univariate techniques.
Multivariate techniques in neuroimaging were introduced over a decade ago (Worsley, Poline, Friston, & Evans, 1997
), but the current interest in the information-based approach is derived from a the idea that brain activation patterns reveal information carried by a neuronal population code at the scale of imaging voxels. This idea motivates multivariate analysis in single subjects without smoothing of the data. An information-based analysis determines whether there is a statistical dependency (i.e. mutual information) between the experimental conditions and the regional spatiotemporal activity patterns. Information undetected by activation-based mapping is often present in neuroimaging data. If the information resides in the fine-scale pattern of the activity, the spatial average may be similar between conditions, so no effect may be found by conventional methods with the data spatially averaged for ROI analysis or smoothed for statistical mapping. Information-based analysis can be applied to predefined ROIs. Alternatively, a continuous information-based mapping can be performed with a multivariate searchlight in order to discover regions carrying a particular type of information (Kriegeskorte et al., 2006
It should be emphasized that this approach to neuroimaging, while, on the surface, appears to approach the brain as a black box from which to extract information about what the subject is doing, can be used as a highly sensitive probe to understand which aspects of a response are most informative, and therefore most relevant to behavior. In addition, as researchers begin to explore and understand precisely why certain classification algorithms work better in the context of extracting information from messy biological systems, perhaps this information will also be useful in lending insight into brain function that could not be probed any other way. Ultimately, the technique may, at the moment, appear to be a flashy way of doing tricks with fMRI, but in fact, this approach has tremendous potential to pose and answer extremely subtle questions about how the brain is organized.
In the past year, two outstanding studies have further pushed the limits of decoding mental content from brain activity. In previous decoding studies, the experiments have been divided between a training set in which the spatial pattern of brain activity was associated with an object (or object category), orientation, or position; and an experiment set in which the same stimuli used in the training set were used. The training set and test set was the same or highly similar stimuli. The two studies described below extend fMRI decoding to allow the accurate identification of brain activation associated completely novel stimuli or tasks.
An exciting paper by Kay et. al (Kay, Naselaris, Prenger, & Gallant, 2008
) has demonstrated the ability for fMRI patterns to be used to identify completely novel images that the subject was viewing. A schematic diagram of the principle steps of this study is shown in . This study consisted of a training set in which subjects viewed 1750 natural images. From this training set a quantitative receptive field model for each voxel was created using Gabor wavelets, characterizing tuning dimensions in space, orientation, and spatial frequency. Subjects then viewed 120 completely novel images. The corresponding brain activity patterns were compared with the receptive field models obtained from the training set. Accuracy ranged from 72% to 92%. In addition, the decrease of accuracy was extremely small as the size of the test set increased, providing hope that with the appropriate training set size and tuning dimensions, activation for just about any object that exists may be characterized with a high level of accuracy.
Figure 8 Obtained from (Kay et al., 2008). This is a schematic diagram showing the steps in the fMRI decoding experiment. In the first stage, fMRI data were recorded as subjects viewed a large collection of natural images. From these data, receptive field models (more ...)
In the second study, Mitchell et al (Mitchell et al., 2008
) describe a method by which whole-brain patterns associated with statistical relationships between words representing abstract semantic concepts are determined and used to predict mental states not present in the training set. The essence of this method is shown in . These results establish a direct predictive relationship between the statistics of word occurrences in text and the activation patterns associated with processing word meanings. A new insight into how the brain processes concrete nouns is also put forth in this study. Essentially, the meaningful fMRI patterns for these nouns are distributed across the brain (i.e. also in prefrontal regions) rather than only existing in typical sensory-motor regions.
Figure 9 Obtained from (Mitchell et al., 2008). This figure shows the process by which predicted fMRI maps were created for specific stimulus words. The top three maps in A. are learned spatial coefficients for 3 of the 25 semantic features related to “celery” (more ...)
Both of these studies represent an important paradigm shift in fMRI decoding - that of using a large training set focused on more elemental aspects of a brain state (i.e. from visual stimulus or word set) to predict the brain activation pattern associated with novel stimuli or brain states associated with processing novel semantic content. Not only do these studies pave the way for extensive applications of fMRI for "brain reading" but also lend a unique insight into how the brain is processing information by being able to identify the "most informative" voxels and regions. Similar insights regarding the significance of tuning functions and whole brain activation patterns would not likely have been uncovered using standard univariate mapping techniques. These studies represent the very beginning of an extremely powerful new approach to fMRI in reading brains and in understanding brain organization.
A related study by Williams et. al (Williams, Dang, & Kanwisher, 2007
) addresses a previously overlooked aspect of pattern information - that which regards which aspects of a specific pattern in the brain are actually used in the processing of a visual stimulus or word and which are nonessential epiphenomenon. In their study, subjects were scanned while observing novel objects that belonged in one of three categories. They were asked to decide which category each viewed object fell into. The corresponding fMRI activation patterns in retinotopic and lateral occipital cortex (LOC
) were shown to contain category information associated with objects, but only in the LOC were the patterns stronger for correct than for incorrect trials. Put another way, in trials in which the subjects did not correctly categorize the stimulus, the correct stimulus information was present in retinotopic cortex but not in LOC. This work represents an alternative direction in fMRI pattern effect imaging - that of determining pattern representations and then comparing with corresponding behavior to further classify cortical areas as they influence behavior.
Lastly, every so often a paper captures the attention of those in the field simply because of the cleverness and elegance of a paradigm. A recent study by Hassan et. al (Hasson, Yang, Vallines, Heeger, & Rubin, 2008
) does just that. The study sets out to identify brain regions responsive to sensory information accumulated over different temporal windows. The paradigm simply involves viewing of a movie played either forward, backward, or in a scrambled manner. Low level processing regions such as primary visual (V1) cortex and motion sensitive area (MT+) showed minimal dependence on the direction in which the movie was played, while higher regions such as the superior temporal sulcus (STS), precuneus, posterior lateral sulcus (LS), temporal parietal junction (TPJ), and frontal eye field (FEF) were highly affected by how the movie was played. In addition LS, TPJ, and FEF activation depended on information accumulated over time periods of about 36 sec while STS and precuneus activation depended on information accumulated over about 12 sec. Using a simple, natural paradigm involving movie viewing, temporal processing scales were elegantly and effectively probed. The summary of this study is shown in .
Figure 10 Obtained from (Hasson et al., 2008). This shows a summary of their clever experiment testing which cortical areas are sensitive and insensitive to time reversals in movie sequences. Plots are from area MT+. A. Representative frames from the silent films (more ...)
e. Endogenous Oscillations in Functional MRI
One area of rapid development in fMRI is that of using endogenous or "resting state" fluctuations or oscillations to explore resting state networks. The basic observation is that when subjects are not performing any task in the scanner, the time series signal consists of fluctuations that are not simply thermal noise of the scanner. Components of these fluctuations include cardiac pulsations of blood and CSF (inflow effects), breathing oscillations (changes in Bo field from chest expansion), bulk motion, scanner instability, vasomotion, BOLD and flow fluctuations that occur with slow changes in end tidal CO2 variations, and BOLD and flow fluctuations that occur with spontaneous neuronal activity. The last component is of course what's most interesting to neuroscientists. The predominant frequency component of these "neuronally related" fluctuations is at or below 0.1 Hz.
Research related to this phenomenon has been along determining three avenues: 1. How to most efficiently and robustly extract and map these correlated resting state fluctuations. 2. How to best interpret these fluctuations, as well as to best separate neuronal from physiologic fluctuations, and 3. How to best apply the phenomenon of resting state fluctuations towards neuroscience questions and towards understanding patient populations. Hundreds of papers have covered these topics in only the past 5 years. In the past year, the trend of an explosive interest in endogenous oscillations continues. A recent special issue of the journal Human Brain Mapping (Volume 29, Issue 7, July 2008) was completely devoted to the publication of studies about endogenous oscillations. Several excellent review articles have also been written in the past year.
Regarding each of the three areas of endogenous oscillations research, the following are some notable advancements in the past year.
1. Endogenous oscillation processing
The current problem in resting state processing is as follows: Given at about 60,000 voxels, and 60,000 × 300 = 18 million time points in a single 5 minute volumetric fMRI scan, the task of determining the correlation of every voxel time series with every other voxel in the brain is computationally prohibitive. Two alternative solutions have been popular. The first is to chose "eed voxel" regions (Biswal, Yetkin, Haughton, & Hyde, 1995
) (M. D. Greicius, Krasnow, Reiss, & Menon, 2003
) and the second is to use relatively unsupervised model-free approaches such as independent component analysis (ICA) (Damoiseaux et al., 2006
). Each of the methods has a limitation. With the seed voxel approach, the results are highly dependent on the choice of ROI or voxel from which the "seed" reference time series is chosen from. Within this chosen time series, much of the energy in the signal driving the correlation may also be artifactual. In addition, many correlated areas are of course, missed, depending on how many "seeds" are chosen. With the ICA approach, it's difficult to sort individual components and to subsequently further interpret what each component means (i.e. What frequency components are correlated? What about phase offsets over space?). As ICA methodology improves, it appears that the problem of robustly extracting a stable set of networks for comparison and clinical use is being solved. Papers that have discussed how to process resting state data last year include the following: (Calhoun, Kiehl, & Pearlson, 2008
; Cohen et al., 2008
; Duff et al., 2008
; Esposito et al., 2008
; Salvador et al., 2008
; van den Heuvel, Mandl, & Hulshoff Pol, 2008
; Wink, Bullmore, Barnes, Bernard, & Suckling, 2008
; Zhu et al., 2008
2. Endogenous oscillation interpretation and separation
A central issue is whether or not the extracted networks actually mean anything neuronal, and if they do contain neuronally-relevant fluctuations, how are these separated robustly from data that contain non-neuronal fluctuations? In the past year, several excellent multi-modal studies have been carried out to demonstrate the neuronal basis of fMRI-based endogenous oscillations (de de Munck et al., 2008
; Hamandi et al., 2008
; Helmut Laufs, 2008
; Shmuel & Leopold, 2008
). In addition, several studies have recently addressed the issue of how physiologic processes such as breathing and cardiac function can influence fluctuations (Birn, Murphy, & Bandettini, 2008
; Birn, Smith, Jones, & Bandettini, 2008
; Shmueli et al., 2007
). Other studies have shown functionally correlated endogenous oscillations in the generally "cleaner signal" perfusion data (Chuang et al., 2008
). While identification of potentially confounding artifacts is relatively straightforward, a clear assessment in each subject of what regions show artifactual correlation and which do not is much more difficult. It is problematic to cleanly "regress out" respiration or other variations. Most would agree that there is sufficient evidence that spontaneous and correlated neuronal activity certainly does contribute to signal fluctuations. Work is ongoing, and will be ongoing for quite a few years, to minimize any false positives in the identification of these networks.
One other point that is not discussed as much as other aspects of endogenous oscillations is in regard to what the "purposes" or precise mechanisms of these neuronally mediated oscillations in flow and BOLD are? Are they only an epiphenomenon of regions being physically connected or functionally related, or do they serve a specific critical function to how we interact with the world? Four interesting studies have come out in the past year which speak to a functional role - or at least an effect on behavior - of endogenous oscillations.
The first study, carried out by Boly et. al (Boly et al., 2007
), starts with the observation that baseline signal magnitude in the brain spontaneously varies. They then demonstrate that a temporal variance in perception of identical stimuli is positively correlated with pre-trial activity in medial thalamus and lateral frontoparietal network (thought to be associated with vigilance and external monitoring) and negatively correlated with posterior cingulate/precuneus and temporoparietal cortices, which are hypothesized to be related to introspection and incidentally, the primary areas associated with what has been identified as the "default network." These results are summarized in .
Figure 11 Obtained from (Boly et al., 2008). Maps of baseline activity that predicts conscious perception of subsequent somatosensory stimuli. A. Increased activity of the medial thalamus (Th), dorsolateral prefrontal cortex (DLPF), intraparietal sulcus/posterior (more ...)
The second study, by Fox et. al (Fox, Snyder, Vincent, & Raichle, 2007
) demonstrates not only that endogenous oscillations persist during tasks and that they contribute to a significant fraction of the trial to trial variance of the BOLD response in supplementary motor cortex with a simple finger tapping task, but, amazingly they also affect motor performance. The performance of the task itself (force of the button press) is affected by when the task is performed relative to the phase of the endogenous oscillation. It appears that when the oscillations are at the peak of their cycle at the beginning of activation-induced BOLD changes, the force of the finger tap force is lightest, and vice versa…when the trough is at the beginning of the BOLD-signal change, the finger tap force is greatest. About 74% of the behavioral (finger tapping force) variance is attributed to ongoing endogenous oscillations in supplementary motor cortex. These results are shown in .
Figure 12 Obtained from (Fox et al., 2007). This is a demonstration that coherent spontaneous activity in the motor cortex influences the force at which subjects press a button. A. Average left SMC BOLD time courses for the hard (blue) and soft (red) button presses. (more ...)
The third study, by Mason et al, (Mason et al., 2007
) suggests that the default network is most active during mind wandering. In a clever study, they demonstrate a positive correlation between the percent signal change in specific nodes of the default network and the frequency of stimulus independent thoughts (or mind wandering), suggesting a conscious introspection role of this network. They conclude with a few speculations: Stimulus independent thoughts may maintain the brain in an optimal state of arousal, thus facilitating performance on mundane tasks. Stimulus independent thoughts maintain a sense of coherence with one's past and future or may be a epiphenomenon of the brain's ability to manage multiple tasks.
Fourth, a comparison study of the resting state networks (default, oculomotor, somatomotor, and visual) in anaesthetized monkeys was carried out by Vincent et al (Vincent et al., 2007
). Not only did they show that the same "default" network was activated in humans and monkey but provided anatomical connectivity and evoked response pattern support for the oculomotor network. This work strongly argues that the "default" network is not dependent on the level of consciousness, as suggested in the study by Mason, mentioned above, nor is it a uniquely human system.
All four of these studies provide intriguing evidence that endogenous oscillations influence how we perceive the world and interact with it. What to make of this information with regard to understanding the functional significance of endogenous oscillations remains to be fully worked out.
3. Applications of neuronal fluctuation assessment
A surprisingly large number of applications of studies involving endogenous oscillations have been reported in a very short time. In the past year, these have included children and infant studies (Liu, Flax, Guise, Sukul, & Benasich, 2008
; Thomason et al., 2008
), populations with multiple sclerosis (Lowe et al., 2008
), intelligence (M. Song et al., 2008
), acupuncture (Dhond, Yeh, Park, Kettner, & Napadow, 2008
), sleep (Horovitz et al., 2008
), sedation (M. Greicius et al., 2008
), seizure activity (Lui et al., 2008
), Alzheimer's (Y. Liu, K. Wang et al., 2008
), schizophrenia (Y. Liu, M. Liang et al., 2008
; Zhou et al., 2008
), chronic pain (Baliki, Geha, Apkarian, & Chialvo, 2008
), ADHD (Castellanos et al., 2008
), depression (M. D. Greicius et al., 2007
), and consciousness (Boly et al., 2008
). For future clinical studies, this is the paradigm that clinicians are looking to with hope: fMRI signal changes that do not require a probe task but that are extremely sensitive to patient population and conscious state.