Leaders in Pharmaceutical Business Intelligence (LPBI) Group

Functional magnetic resonance imaging

Larry H. Bernstein, MD, FCAP, Curator

LPBI

 

Demystifying BOLD fMRI Data

What does blood oxygen level–dependent functional magnetic resonance imaging actually tell us about brain activity?

By Tim Vernimmen | February 17, 2016   http://www.the-scientist.com/?articles.view/articleNo/45366/title/Demystifying-BOLD-fMRI-Data

BOLD signal in no task (“resting state”) fMRI YOUTUBE, ZEUS CHIRIPA
http://www.the-scientist.com/images/News/February2016/yC7leMG%20-%20Imgur.gif

he relevance and reliability of blood oxygen level-dependent functional magnetic resonance imaging (BOLD fMRI) data have been hotly debated for years, not least because it is still unclear what aspects of brain activity the technique is picking up. “In many ways, this would seem to be an unacceptable method for neuroscience,” said Ed Bullmore from the University of Cambridge, at a Royal Society-organizedgathering of neuroscients late last month. “But if you’re interested in humans, there isn’t much of a choice.” Bullmore and colleagues had convened in Buckinghamshire, U.K., to discuss what, exactly, BOLD fMRI results can tell us.

“What we do know, of course, is what MRI measures,” said Robert Turner, director emeritus of the Max Planck Institute for Human Cognitive and Brain Sciences in Leipzig, Germany. MRI measures the magnetization of hydrogen protons in water molecules excited by pulses of radio waves that lead their spins to temporarily align. “Over the next few tens of milliseconds,” Turner noted, “their orientations fan out again, and the magnetization we measure will quickly decrease.”

But what can this tell us about brain activity?

When hemoglobins—the iron-rich oxygen-carrying proteins in our blood—run out of oxygen, Turner explained, “they become paramagnetic,” disturbing the local magnetic field. This makes the protons spin out of phase more rapidly.” One might think this means BOLD fMRI highlights oxygen consumption by active neurons, but in reality, such activity is rarely measured.

What BOLD does reveal is what usually happens next: fresh blood rushes into the area, flushing out paramagnetic deoxyhemoglobin and replacing it with new, oxygenated hemoglobin. Since this does not interfere with the proton spins, the result is a larger fMRI signal. So BOLD fMRI reflects a combination of changes in blood flow and oxygen consumption within the brain—not neuronal activity itself.

“This means that if BOLD shows you a large blob of activity, that doesn’t necessarily mean that all the neurons in that region are spiking,” said David Attwell of University College London, one of the meeting’s organizers. “So what we really need to know is how neurons are influencing bloodflow.”

To find out, Attwell and his colleagues are studying postmortem slices of rodent brain to better understand the interactions between neurons, blood vessels, and supporting cells such as astrocytes and pericytes. These cells wrap around the vasculature and likely affect its response to local neural activity.

Research on living animals, on the other hand, has suggested that endothelial cells lining the brain’s blood vessels may also play an active role in coordinating such responses, as they are known to do elsewhere in the body. “The wave of vessel dilation resulting in increased bloodflow travels much faster and farther than could be explained by astrocytes and pericytes alone,” said Elizabeth Hillman of Columbia University in New York City, whose lab has developed an optical method to look into rat brains directly. “Moreover, if we disable parts of the endothelium, we can see that wave come to a halt.”

More recently, the Hillman lab unexpectedly uncovered what seems to be a convincing link between neural and vascular activity. “While trying to disprove that resting state activity in the brain could teach us about neural connections we have actually been able to observe seemingly spontaneous neural activity that correlates with bloodflow quite tightly,” Hillman told The Scientist, “which would be hard to show with the very precise single-neuron measurements many neuroscientists prefer, but when you zoom out and look at the larger picture, the synchrony is hard to deny—and believe me, we’ve tried very hard to explain these results away.”

If these unpublished findings stand up to the scrutiny of Hillman’s colleagues,  this would be reassuring news for neuroscientists using BOLD fMRI to study neural activity.

But in some brains, BOLD may not work at all, Hillman cautioned. “In the developing brain of young animals, for example, we find that BOLD activity is very unusual,” she said. “Initially, the bloodflow response doesn’t seem to be attuned to neural activity at all, so fMRI may be as good as blind.”

Diseased brains can also skew results. “Pathology may affect the BOLD signal in the absence of any changes in neurons themselves,” said Bojana Stefanovic of Toronto’s Sunnybrook Research Institute. In patients who suffered a stroke, for example, the amount of water may be reduced where cells have died, and increased by oedema in some of the surrounding tissues. The brain’s bloodflow may also be altered by disruptions to the vasculature, for example, or the formation of scar tissue.

The best way to deal with this depends on the research question, Stefanovic told The Scientist. “There’s this idea that if we can link BOLD to neuronal activity—that would be nirvana,” she said. “Clinicians, however, are looking for measures with a clear link to symptoms. And, fortunately, there is no shortage of disease effects BOLD can sense.”

Cognitive neuroscientist Geraint Rees of University College London sounded a similar note. “If whatever BOLD is measuring reproducibly correlates to the behavior I’m interested in, such as attention or consciousness, I am less worried about the physiological details behind it,” he said. “Which does not mean, of course, I don’t consider them interesting—otherwise, I wouldn’t be here.”

Meanwhile, researchers are developing methods to measure human neural activity more directly, learning more about BOLD fMRI data along the way. “Thanks to over 30 Parkinson’s patients who agreed to play an investment game while undergoing surgery for the placement of a deep-brain stimulation probe, we were able to directly measure the striatal dopamine response we only knew from rodents and human BOLD,” said Read Montague of the Virginia Tech Carilion Research Institute. “Surprisingly, we found that while BOLD responds to expected reward and actual outcome separately, the dopamine response integrates them into one ‘better or worse’ signal.”Montague’s team would next like to explore whether the same is true for people without Parkinson’s disease, which is known to affect dopaminergic neurons.

For now, however, the researchers’ results demonstrate the benefits of applying other techniques in parallel with BOLD fMRI. Not only might this approach reveal insights BOLD cannot, it might also help neuroscientists better understand the results of past fMRI experiments.

Interpreting BOLD: a dialogue between cognitive and cellular neuroscience

Kavli Royal Society Centre, Chicheley Hall, Newport Pagnell, Buckinghamshire, MK16 9JJ

Overview

 

Using an achiasmic human visual system to quantify the relationship between the fMRI BOLD signal and neural response

 

Achiasma in humans causes gross mis-wiring of the retinal-fugal projection, resulting in overlapped cortical representations of left and right visual hemifields. We show that in areas V1-V3 this overlap is due to two co-located but non-interacting populations of neurons, each with a receptive field serving only one hemifield. Importantly, the two populations share the same local vascular control, resulting in a unique organization useful for quantifying the relationship between neural and fMRI BOLD responses without direct measurement of neural activity. Specifically, we can non-invasively double local neural responses by stimulating both neuronal populations with identical stimuli presented symmetrically across the vertical meridian to both visual hemifields, versus one population by stimulating in one hemifield. Measurements from a series of such doubling experiments show that the amplitude of BOLD response is proportional to approximately 0.5 power of the underlying neural response. Reanalyzing published data shows that this inferred relationship is general.

DOI: http://dx.doi.org/10.7554/eLife.09600.001

 

eLife digest

 

Functional magnetic resonance imaging (fMRI) based on the blood oxygenation level dependent (BOLD) signal has provided unprecedented insights into the workings of the human brain. The quantitative relationship between neural signals and the fMRI BOLD response is not precisely known and remains an active area of investigation. Most studies using the BOLD signal to infer brain activity rely on analytical methods (e.g., the general linear model) that assume a linear relationship between the BOLD signal and neural response, despite noticeable deviations from linearity (Boynton et al., 1996).

The BOLD signal is indirectly related to local neural response through mechanisms associated with oxygen metabolism and blood flow (Davis et al., 1998; Hoge et al., 1999; Thompson et al., 2003;Griffeth and Buxton, 2011). The neural response that is associated with information processing is itself multi-faceted. It comprises several interacting components, including subthreshold and suprathreshold electrical activities, the transport, release and reuptake of neurotransmitters, and various maintenance activities. Each of these components has its own metabolic and hemodynamic consequences. The common extracellular measurements of neural response include single- and multi-unit spiking activities and local field potential (LFP). While seminal studies have demonstrated a close relationship between the BOLD signal and these extracellular measurements of neural response (Logothetis et al., 2001; Mukamel et al., 2005), the quantitative nature of this relationship has not been sufficiently characterized. More importantly, since the relationship between these extracellular measurements and the intracellular components of neural activity is complex, the measured relationship between the BOLD signal to any specific extracellular components (e.g., power in the gamma band of LFP) may not reflect the relationship between the BOLD signal and the totality of neural response.

Most applications of fMRI, particularly in human neuroscience, sidestep any need for explicitly estimating neural activity and instead rely on establishing a direct relationship between the BOLD response and the stimulus condition. The general approach is to assume the BOLD responses evoked at different times and in different stimulus conditions sum linearly. Boynton and colleagues (1996) studied how the BOLD signal varied with the contrast and duration of stimulus presentation in the striate cortex and found that the system is approximately linear, in the sense that the BOLD response evoked by a 12 s stimulus was well approximated by summing the responses from two consecutive 6-s stimulations, even though predictions based on stimulations of much shorter durations (e.g., 3 s) failed to accurately predict the long-duration stimulus response. While this and similar studies (Cohen, 1997; Dale and Buckner, 1997; Heckman et al., 2007) have clearly noted the lack of linearity, their general message of an approximately linear system has nevertheless been used to justify the broad application of the general linear model (GLM) in fMRI data analyses. While the neural response is not explicitly involved in this type of analysis, it is always in the background — any nonlinearity observed in the BOLD response, e.g., in surround suppression or adaptation (Grill-Spector and Malach, 2001; Kourtzi and Huberle, 2005; Larsson and Smith, 2012) is often attributed to the underlying nonlinear neural response. The implicit assumption in common practice is that the relationship between the BOLD response and the neural response is essentially linear, a view that is widespread (Logothetis and Wandell, 2004) but under-examined.

An extensive set of biophysical models has been proposed to express either the steady-states (Davis et al., 1998; Griffeth and Buxton, 2011) or the dynamics of the BOLD response (Buxton et al., 1998;Mandeville et al., 1999; Feng et al., 2001; Toronov et al., 2003; Blockley et al., 2009; Kim and Ress, 2016) in terms of more basic physiological components, such as blood flow, blood volume, oxygen saturation, and oxygen extraction fraction in different vascular compartments. These biophysical models are foundational in our understanding of the BOLD signal, yet they do not provide any explicit and quantitative linkage between the neural response and the physiological components that are the inputs to these models. Friston et al. (2000) (see also Stephan et al., 2007), proposed a linkage between the evoked neural response and the blood-flow parameter of the Balloon model by Buxton et al. (1998). While the resulting model is a powerful tool for inferring effective connectivity between brain regions from the BOLD signal, direct empirical support for this specific linkage is limited.

How could we empirically determine the quantitative relationship between the BOLD signal and the neural response, and do so when the constituents of the neural response are not comprehensively defined? A condition known as achiasma or non-decussating retinal-fugal fibre syndrome may provide an excellent model system for this purpose. This congenital condition prevents the normal crossing of optic nerve fibers from the nasal hemi-retina to the brain hemisphere contralateral to the eye (Apkarian et al., 1994; 1995). The result is a full representation of the entire visual field (as opposed to only half the visual field) in each cerebral hemisphere (Williams et al., 1994; Victor et al., 2000; Hoffmann et al., 2012; Davies-Thompson et al., 2013; Kaule et al., 2014). Specifically, the representations of the two visual hemifields are superimposed in the low-level visual areas (V1-V3) ipsilateral to each eye, such that two points in the visual field located symmetrically across the vertical meridian are mapped to the same point on the cortex (Hoffmann et al., 2012). In other words, there are two pRFs for every point on this person’s low-level visual cortex. The two pRFs are symmetrically located across the vertical meridian. Prior to the current study, it was not known if these pRFs were represented by one or two neural populations, or if these neural populations interacted.

In the current study, we found that the two pRFs are each represented by an independent population of neurons. The result is an in-vivo system with two independent populations of spatially intermingled neurons that share the same local control of blood vasculature. Because their population receptive fields (pRFs) do not overlap, an experimenter can independently stimulate each population by presenting a stimulus to its respective receptive field. Such a system is ideal for characterizing the relationship between neural and BOLD responses. Even though we may not know the constituents of the neural response, it will be reasonable to assume that the local neural response evoked by presenting identical stimuli to both pRFs, thereby activating both neuronal populations equally, is twice the neural response evoked by presenting the stimulus to just one of the pRFs. Measuring BOLD responses under these conditions allows us to not only directly test for linearity between the BOLD signal and neural response but also quantify the relationship between them, up to an arbitrary scaling factor. This approach does not require us to know the constituents of neural activity, and it is non-invasive.

To determine the relationship between neural response and the corresponding fMRI BOLD signal, we measured BOLD responses in the cortical areas V1-V3 of our achiasmic subject to luminance-defined stimuli. We presented stimuli of different contrasts to either one or both of the pRFs. From this data set, we used a model-free non-parametric method to infer the quantitative relationship between the BOLD signal (B) and neural response (Z). We found that the resulting B vs. Z function is well approximated by a power function with an exponent close to 0.5. The exponent stayed the same for short and long stimulus durations. We successfully cross-validated this result by comparing the inferred neural responses from this and twelve other fMRI studies to the single-unit responses obtained from non-human primates in similar contrast-response experiments.

 

Article