Neuroimaging methods infer brain activity from regional cerebral blood flow, glucose metabolism and neurochemical concentrations, and infer changes in brain function from structural measures of water motility and brain volume

The most common functional methods evaluate changes in regional cerebral blood flow (rCBF). Increased neural activity triggers an increase in cerebral blood flow to compensate for the anticipated neural metabolic demand. This increase in rCBF occurs after a variable delay of nearly 3–8 s, is highly localised to the region of increased neural activity and is closely coupled with the magnitude and duration of this activity. This haemodynamic response is evaluated by several methods described below and interpreted as a measure of both basal and evoked neural activity. Some methods use radiolabelled tracers or physical properties of target molecules to assess glucose metabolism or the concentrations of metabolites putatively associated with neural activity, and receptor binding of ligands, such as endogenous opioids. Methods that produce conventional anatomical images provide measures of the relative volume of different brain regions in different clinical populations on a macrostructural scale, while newer methods of water motility can compare structural differences on a finer, microstructural scale .

Neuroimaging methods accomplish these tasks either by detection of radiation from an external or internal source, or by detection of radio waves produced by atomic motion

The well-known method of computerised axial tomography (CAT) provides structural images by computing images resolved from the detection of multiple paths of radiation from an external source to a detector. By contrast, single-photon-emission computed tomography (SPECT) and positron emission tomography (PET) localise the sources of internal radiation resulting from infusion of a radioactive tracer. These latter methods provide a measure of basal rCBF throughout the brain. SPECT has less resolution than PET, with long scans that require subjects to remain motionless for 15–20 min, which can be reduced by multiple head machines. The increased temporal and spatial resolution of PET permits evaluation of evoked changes in rCBF with a time scale of about a minute. Both SPECT and PET can also use radioactive ligands to measure the amount of tracer bound to receptors. This measure is termed ‘binding potential’ and is related to the number of receptors and, inversely, to receptor occupancy by the target endogenous ligand. For example, if mu opioid receptors are fully occupied by endogenous opioids, such as endorphins, there will be little binding by the tracer and a small binding potential. Likewise, an increase in binding potential could represent less occupancy by the endogenous opioid or an increase (up-regulation) in the number of receptors.

Magnetic resonance imaging (MRI) produces high-resolution structural images by placing the patient in a strong magnetic field and resolving radio frequencies that are characteristic of both specific molecules of interest and the field strength of the magnetic field. As the body is comprised of mostly water, MRI imaging focusses on the radio frequencies, known as Lamor frequencies, transmitted by synchronised hydrogen nuclei in water (and fat) molecules, which, in a 1.5-T scanner, corresponds to approximately 64 MHz, near the former analogue TV channel 3. Manipulation of the magnetic field strength selects successive slices to be imaged, and further manipulation of field strength and phase of radio frequency pulses provide two-dimensional images for each slice. A succession of slices provides an image of the entire brain as fast as once per second (depending on resolution and number of slices). The increased resolution of MRI is useful for structural studies of the differences in the volume of brain-processing regions between groups or within individuals over time.

Functional MRI (fMRI) forms images of brain activity by using the magnetic character of blood as an intrinsic tracer. The concentration of oxygen in the blood is inversely related to the magnetic character of haemoglobin, with deoxygenated (blue) blood exhibiting magnetic properties while oxygenated (red) blood is only weakly magnetic. Under normal conditions, the magnetic character of the blood inhibits the MRI signal from adjacent tissue, much like a hairpin or other metal will produce a blacked out ‘hole’ region in an MRI anatomical image. As described above for PET, neural activity leads to a haemodynamic response that overcompensates for the metabolic need. This results in a localised region of increased oxygen in the blood. The decreased magnetic character of this red blood decreases the suppression of the signal from adjacent tissue. The blood-oxygen-level-dependent (BOLD) method of MRI detects this increased signal to form images of inferred neural activity. Like PET, BOLD fMRI can produce images of evoked activity, with improved temporal (seconds) and spatial (millimetre) resolution. Unlike PET, the BOLD method of fMRI cannot produce basal images of activity between groups. This limitation is overcome by injection of tracers similar to PET, although the tracers used for MRI perfusion are not radioactive. A more recently developed fMRI method of arterial spin labelling (ASL) uses no injected tracer. This method magnetises blood (by synchronising water molecules) as it enters the brain and then uses the decay of the magnetisation to form images. It provides basal images of brain activity inferred from basal rCBF. It can also provide measures of evoked activity with less temporal and spatial sensitivity than the BOLD method but with the advantages of less susceptibility to distortion along the edges of the brain near air or fluid, a more linear association with rCBF and better characteristics for repeated measurement over separate sessions. In addition, the signal is more closely localised to the site of neural activity in the capillary beds versus the more distant draining veins in the case of the BOLD contrast method.

Additional methods that provide basal measurements include functional connectivity MRI (fcMRI), which establishes functional connections between brain regions at rest by correlating spontaneous fluctuations in the BOLD signal among these regions. In contrast to the macrostructural changes assessed by the brain volume methods discussed earlier, diffusion MRI uses water motility to assess microstructural changes using two methods: diffusion weighted imaging (DWI) can detect changes related to damage from a stroke. Diffusion tensor imaging (DTI) provides information about the integrity of nerve tracts, referred to as white-matter tracts, or white matter, by the direction of diffusion along the tracts. Finally, while the MRI methods focus on the signal produced by the ubiquitous water molecule, they were developed from older methods that evaluate the spectra of signals from all molecules in a sample. This approach, long used to identify compounds in chemistry, has been applied to humans using magnetic resonance spectroscopy (MRS), which provides another method to evaluate basal and evoked brain function.

Neuroimaging methods can be divided into measures of either basal or evoked brain activity

Rather than present the chronological development of neuroimaging methods for FM, the following will discuss differences between FM patients and controls, first in the absence of any provocation, followed by response to somatosensory and cognitive stimulation.

Structural differences evaluated by voxel-based morphometry

Macroscopic changes in brain structure can be evaluated by voxel-based morphometry (VBM), which uses structural MRI images to assess the volume of specific brain regions. Chronic pain and stress-related disorders, including facial pain, arthritis, dsymenorrhoea, chronic low back pain, tension type headache, chronic fatigue syndrome and post-traumatic stress disorder, are all characterised by regional reductions in brain volume . FM shares many commonalities with these disorders, and an increasing number of studies have examined brain morphology in FM. Many of these focus on the volume of grey matter, which contains neural cell bodies and is associated with neural processing in contrast to white matter, which is composed of neural axons in interconnecting pathways. Kuchinad et al. have shown that individuals with FM exhibit a 3.3 times greater age-associated decrease in grey matter volume than healthy controls, such that each year of FM was equivalent to 9.5 times the normal loss of grey matter with age . Subsequent studies have demonstrated changes in brain volume in FM in a number of regions, with only some consistency in changes in amygdala, cingulate cortex and hippocampus/parahippocampus .

The functional significance of grey matter atrophy in FM is not clear. One hypothesis is that it is related to impaired endogenous pain inhibition and demonstrated deficits in cognitive function . Altered brain morphology associated with cognitive impairment in FM has been observed in brain regions associated with pain modulation . Interpretation of this growing literature is limited by the cross-sectional nature of most studies. Without evidence of causation, it is not known if altered brain morphology increases the probability of a spontaneous or triggered chronic pain condition, or if prolonged pain results in the observed changes in brain volume. The influence of extraneous factors is also a concern. Hsu et al. found differences in FM that disappeared after controlling for affective disorder. At least two studies have addressed the limitations of cross-sectional studies. Tu et al. examined brain volume in women with dysmenorrhoea and control women, and found a number of volumetric changes in brain regions involved in pain transmission, pain modulation, affective regulation and endocrine function when the women were in the pain-free periods of their menstrual cycle. This suggests that these changes could precede symptoms or persist after symptom cessation. A recent study by Gwilym et al. suggests a persistence that may fade with time. Sixteen patients with unilateral hip osteoarthritis were evaluated 9 months before and after hip arthroplasty. Significant changes in brain volume, especially reduced thalamic volume, were observed in patients in comparison to controls preoperatively. Remarkably, these changes normalised postoperatively, suggesting that the observed changes were dynamically associated with pain and were not an accelerated ageing process.

Structural differences evaluated by DTI

VBM characterises gross changes in brain volume that are quantified in terms of cubic millimetres. By contrast, DTI quantifies microstructural organisational changes. This method is based on the movement of water through brain tissue, and can be quantified in terms of flow and direction. Increased water flow is measured as the apparent water diffusion coefficient (ADC). The loss of myelin and axonal membranes during degeneration, which normally restrict diffusion, reduces the degree of diffusion directionality or functional anisotropy (FA) . A combined VBM and DTI study of FM showed regional brain degeneration in FM subjects compared with healthy controls using both methods. Unlike VBM, the DTI results were correlated with FM symptom severity, suggesting improved sensitivity with DTI . In a related study, Sundgren et al. observed reduced FA in the right thalamus of FM patients that was statistically greater in FM individuals with worse clinical pain and an external locus of control. This result and the finding of normal ADC suggest localised neuronal disorganisation rather than ongoing axonal degeneration.

Neurochemical differences evaluated by magnetic resonance spectroscopy (MRS) and PET ligand binding

fMRI uses a standard methodology developed in the 1960s to detect a signal from a specific source, in this case the proton of the hydrogen atom in water and fat. The MRS evaluates the signal from other chosen sources and expresses the magnitude of this signal in relation to a control signal from a specified standard molecule (often creatine). The resultant profile of relative concentration of molecules, such as glutamate, aspartate, glycine and gamma amino butyric acid (GABA), have a number of potential applications, such as detecting the presence of pathological processes before the appearance of more gross structural changes, or identifying the underlying mechanisms of these changes (e.g., excitotoxicity) . MRS has demonstrated differences in concentrations of substances between FM and control subjects, including the ratio of choline/creatine in the dorsolateral prefrontal cortex, and both glutamate and combined glutamate/glutamine within the insula and posterior gyrus . These levels have been found to be associated with the magnitude of spontaneous pain , and with basal experimental pain sensitivity . These levels have also been associated with changes in experimental pain sensitivity and the fMRI BOLD signal in response to a non-pharmacological treatment . These results are consistent with accumulating evidence of the important role of the insula in processing pain magnitude. Further studies are needed to assess whether a substance, such as glutamate, reflects pain activity and, thus, is a surrogate of subjective pain or is also associated with processes involved in initiating or maintaining FM.

Concentrations of neurochemicals may also be evaluated indirectly by SPECT or PET ligand-binding studies, which measure receptor occupancy by a tagged tracer, such as the opioid, carfentanil, which binds to the mu opioid receptor. Harris et al. used administration of [ ¹¹ C]carfentanil to evaluate mu opioid receptor binding in FM. Significantly less binding of the [ ¹¹ C]carfentanil was found in 17 patients compared with 17 control subjects in bilateral nucleus accumbens and left amygdala. A similar reduction in binding approached significance in the right dorsal anterior cingulate cortex ( p < 0.07). Correlational analyses across the individuals found negative associations between binding in the left nucleus accumbens and affective pain ratings, and negative associations between binding in the left amygdala and depression scores. As described above, these effects can represent the joint effect of receptor availability and also the number of available receptors. Thus, the observed lesser binding in the FM patients could be due to: (1) occupancy by endogenous opioids released as a consequence of the ongoing pain and/or (2) receptor down-regulation. Similar results of reduced binding have been observed for other chronic pain syndromes, suggesting that this result is characteristic of chronic pain and is not specific to FM. However, findings of different patterns of results in studies of neuropathic pain and of rheumatoid arthritis suggest that there may be some specificity among pain syndromes.

fcMRI reveals enhanced connectivity in FM

The temporal resolution of fMRI permits evaluation of spontaneous BOLD fluctuations (<0.08 Hz) and the associations of these fluctuations between different brain regions. The analyses compensate for the effects of respiration and heart rate and can be performed using an a priori ‘seed’ or no a priori assumptions. The seeded analysis begins with a reference region or regions that represent either default locations or regions based on a priori hypotheses. The method evaluates the association of activity of brain regions to this reference region and posits functional connections based on the pattern and strength of these associations. In an unpublished pilot study comparing 10 FM patients to 10 control subjects, Welsh et al. found increased connectivity in insula/orbital cortex in the FM patients using the posterior cingulate cortex as a seed. The insular region is active in most brain imaging studies of pain and is involved in affective appreciation of pain. More recent methods need neither seeds nor a priori hypotheses. Procedures such as machine learning , structural equation modelling or multivariate methods have identified networks related to multiple aspects of brain function. In an application of this approach to FM (18 patients and 18 controls), Napidow et al. found greater connectivity in the patients within a resting network and within a network association with attention. Patients had greater connectivity within the bilateral default mode network and the right attentional network. Increased connectivity was also observed between the attentional network and the insula; the magnitude of spontaneous pain was related to greater connectivity between the insula and both the identified default and attentional networks. These networks of brain regions performed with subjects at rest describe what has been termed ‘default mode’ processing, that is, brain activity under normal conditions without any task demands. This processing is described further below in the section on response suppression.

Structural differences evaluated by voxel-based morphometry

Macroscopic changes in brain structure can be evaluated by voxel-based morphometry (VBM), which uses structural MRI images to assess the volume of specific brain regions. Chronic pain and stress-related disorders, including facial pain, arthritis, dsymenorrhoea, chronic low back pain, tension type headache, chronic fatigue syndrome and post-traumatic stress disorder, are all characterised by regional reductions in brain volume . FM shares many commonalities with these disorders, and an increasing number of studies have examined brain morphology in FM. Many of these focus on the volume of grey matter, which contains neural cell bodies and is associated with neural processing in contrast to white matter, which is composed of neural axons in interconnecting pathways. Kuchinad et al. have shown that individuals with FM exhibit a 3.3 times greater age-associated decrease in grey matter volume than healthy controls, such that each year of FM was equivalent to 9.5 times the normal loss of grey matter with age . Subsequent studies have demonstrated changes in brain volume in FM in a number of regions, with only some consistency in changes in amygdala, cingulate cortex and hippocampus/parahippocampus .

The functional significance of grey matter atrophy in FM is not clear. One hypothesis is that it is related to impaired endogenous pain inhibition and demonstrated deficits in cognitive function . Altered brain morphology associated with cognitive impairment in FM has been observed in brain regions associated with pain modulation . Interpretation of this growing literature is limited by the cross-sectional nature of most studies. Without evidence of causation, it is not known if altered brain morphology increases the probability of a spontaneous or triggered chronic pain condition, or if prolonged pain results in the observed changes in brain volume. The influence of extraneous factors is also a concern. Hsu et al. found differences in FM that disappeared after controlling for affective disorder. At least two studies have addressed the limitations of cross-sectional studies. Tu et al. examined brain volume in women with dysmenorrhoea and control women, and found a number of volumetric changes in brain regions involved in pain transmission, pain modulation, affective regulation and endocrine function when the women were in the pain-free periods of their menstrual cycle. This suggests that these changes could precede symptoms or persist after symptom cessation. A recent study by Gwilym et al. suggests a persistence that may fade with time. Sixteen patients with unilateral hip osteoarthritis were evaluated 9 months before and after hip arthroplasty. Significant changes in brain volume, especially reduced thalamic volume, were observed in patients in comparison to controls preoperatively. Remarkably, these changes normalised postoperatively, suggesting that the observed changes were dynamically associated with pain and were not an accelerated ageing process.

Structural differences evaluated by DTI

VBM characterises gross changes in brain volume that are quantified in terms of cubic millimetres. By contrast, DTI quantifies microstructural organisational changes. This method is based on the movement of water through brain tissue, and can be quantified in terms of flow and direction. Increased water flow is measured as the apparent water diffusion coefficient (ADC). The loss of myelin and axonal membranes during degeneration, which normally restrict diffusion, reduces the degree of diffusion directionality or functional anisotropy (FA) . A combined VBM and DTI study of FM showed regional brain degeneration in FM subjects compared with healthy controls using both methods. Unlike VBM, the DTI results were correlated with FM symptom severity, suggesting improved sensitivity with DTI . In a related study, Sundgren et al. observed reduced FA in the right thalamus of FM patients that was statistically greater in FM individuals with worse clinical pain and an external locus of control. This result and the finding of normal ADC suggest localised neuronal disorganisation rather than ongoing axonal degeneration.

Neurochemical differences evaluated by magnetic resonance spectroscopy (MRS) and PET ligand binding

fMRI uses a standard methodology developed in the 1960s to detect a signal from a specific source, in this case the proton of the hydrogen atom in water and fat. The MRS evaluates the signal from other chosen sources and expresses the magnitude of this signal in relation to a control signal from a specified standard molecule (often creatine). The resultant profile of relative concentration of molecules, such as glutamate, aspartate, glycine and gamma amino butyric acid (GABA), have a number of potential applications, such as detecting the presence of pathological processes before the appearance of more gross structural changes, or identifying the underlying mechanisms of these changes (e.g., excitotoxicity) . MRS has demonstrated differences in concentrations of substances between FM and control subjects, including the ratio of choline/creatine in the dorsolateral prefrontal cortex, and both glutamate and combined glutamate/glutamine within the insula and posterior gyrus . These levels have been found to be associated with the magnitude of spontaneous pain , and with basal experimental pain sensitivity . These levels have also been associated with changes in experimental pain sensitivity and the fMRI BOLD signal in response to a non-pharmacological treatment . These results are consistent with accumulating evidence of the important role of the insula in processing pain magnitude. Further studies are needed to assess whether a substance, such as glutamate, reflects pain activity and, thus, is a surrogate of subjective pain or is also associated with processes involved in initiating or maintaining FM.

Concentrations of neurochemicals may also be evaluated indirectly by SPECT or PET ligand-binding studies, which measure receptor occupancy by a tagged tracer, such as the opioid, carfentanil, which binds to the mu opioid receptor. Harris et al. used administration of [ ¹¹ C]carfentanil to evaluate mu opioid receptor binding in FM. Significantly less binding of the [ ¹¹ C]carfentanil was found in 17 patients compared with 17 control subjects in bilateral nucleus accumbens and left amygdala. A similar reduction in binding approached significance in the right dorsal anterior cingulate cortex ( p < 0.07). Correlational analyses across the individuals found negative associations between binding in the left nucleus accumbens and affective pain ratings, and negative associations between binding in the left amygdala and depression scores. As described above, these effects can represent the joint effect of receptor availability and also the number of available receptors. Thus, the observed lesser binding in the FM patients could be due to: (1) occupancy by endogenous opioids released as a consequence of the ongoing pain and/or (2) receptor down-regulation. Similar results of reduced binding have been observed for other chronic pain syndromes, suggesting that this result is characteristic of chronic pain and is not specific to FM. However, findings of different patterns of results in studies of neuropathic pain and of rheumatoid arthritis suggest that there may be some specificity among pain syndromes.

fcMRI reveals enhanced connectivity in FM

The temporal resolution of fMRI permits evaluation of spontaneous BOLD fluctuations (<0.08 Hz) and the associations of these fluctuations between different brain regions. The analyses compensate for the effects of respiration and heart rate and can be performed using an a priori ‘seed’ or no a priori assumptions. The seeded analysis begins with a reference region or regions that represent either default locations or regions based on a priori hypotheses. The method evaluates the association of activity of brain regions to this reference region and posits functional connections based on the pattern and strength of these associations. In an unpublished pilot study comparing 10 FM patients to 10 control subjects, Welsh et al. found increased connectivity in insula/orbital cortex in the FM patients using the posterior cingulate cortex as a seed. The insular region is active in most brain imaging studies of pain and is involved in affective appreciation of pain. More recent methods need neither seeds nor a priori hypotheses. Procedures such as machine learning , structural equation modelling or multivariate methods have identified networks related to multiple aspects of brain function. In an application of this approach to FM (18 patients and 18 controls), Napidow et al. found greater connectivity in the patients within a resting network and within a network association with attention. Patients had greater connectivity within the bilateral default mode network and the right attentional network. Increased connectivity was also observed between the attentional network and the insula; the magnitude of spontaneous pain was related to greater connectivity between the insula and both the identified default and attentional networks. These networks of brain regions performed with subjects at rest describe what has been termed ‘default mode’ processing, that is, brain activity under normal conditions without any task demands. This processing is described further below in the section on response suppression.