Chapter 10. Measuring Pain with Functional Magnetic Resonance Imaging

Pain is recognized as a sensory and emotional experience in humans. Unfortunately, there is no objective test for measuring pain. This has hampered both the clinical management and the scientific understanding of pain. In the clinical setting, physicians daily encounter difficulties in diagnosing chronic pain conditions. The findings of commonly used testing modalities (magnetic resonance imaging [MRI], computed tomography, electromyography) are frequently normal. History and physical examination are highly subjective tests and prone to examiner bias. Complaints of chronic pain patients are frequently labeled “psychogenic” in origin. The majority of chronic pain patients suffer from depression; however, it is difficult to determine if depression is a consequence of chronic pain or vice versa. Indeed, emotional and pain brain networks share similar anatomic structures.

Most of the limited knowledge of central nervous system (CNS) pain processing is derived from animal research using electrophysiologic recordings. The animal data do not provide adequate insight into human aspects of pain processing, such as the affective component of pain. There is need for a better understanding of the following aspects of human pain networks: (1) chronic pain states in which altered CNS processing is taking place; (2) effects of various drugs on brain activation in acute and chronic pain; and (3) perception of pain in altered states of conciousness.

The initial findings that provided an insight into human CNS pain networks were accomplished by functional brain imaging using positron emission tomography (PET) technology, followed by studies using functional magnetic resonance imaging (fMRI). Although still in their infancy, these tools are extremely valuable in bridging the gap between clinical practice and animal research in understanding pain.

Functional brain imaging techniques are based on similar methods of measuring brain neuronal activity. Glucose is the main energy source for human brain metabolism. The coupling between regional cerebral blood flow and local cerebral glucose use has been established and supported by experimental data. Glucose metabolism reflects the brain neuronal synaptic and presynaptic activity needed for maintenance of membrane potentials and restoration of ion gradients. Functional brain imaging, by measuring the regional blood flow, provides indirect data on regional brain neuronal activity, be it local activation or inhibition.

PET is an instrument that allows accurate measurements of small concentrations of radioactivity in living tissue. Positron-emitting isotopes have been incorporated as tracers into a wide range of molecules to provide information about various biologic processes after inhalation or intravenous administration. By using this technique, the whole brain metabolic activity in pain states can be measured. The ability to overlap brain metabolic activity data acquired by PET with structural magnetic images of brain (MRI) enhanced the effective spatial resolution of PET. There are numerous studies using this technology in medicine.

The first brain mapping study focusing on pain used PET. This study demonstrated that painful heat produced activation in multiple brain areas, including cortical and subcortical structures.1 Since then, PET with various experimental paradigms has helped provide a better understanding of the brain circuitry involved in pain processing.2–5 In these studies, noxious stimulation produced activation in multiple brain regions, most commonly the cingulate gyri, somatosensory cortex, thalamus, insula, frontal lobes, and cerebellum. The fact that hypnosis can modulate subjective pain perception in anterior cingulate gyri was also shown using PET technology.6 Recently, fMRI studies have replicated results previously accomplished by PET technology. Taking into consideration its advantages over PET, fMRI has become a promising tool for brain mapping in pain states. However, PET may remain the gold standard for receptor-binding imaging.

Essentially, fMRI is the “ultrafast” MRI. Similar to PET methodology, fMRI is based on the tight relationship among neuronal synaptic activity, metabolism, and blood circulation. Conventional MRI was first performed in the 1940s, and today its principle is used routinely in the clinical setting in the form of the MRI scanner. When a patient is placed in the MRI scanner, hydrogen nuclei from water molecules are oriented along the magnetic fields of the scanner. By administering radiofrequency pulses to these nuclei, their alignment can be disturbed. After the radiofrequency pulse ceases, the nuclei realign themselves by transmitting a radiofrequency signal that can be detected in a receiver coil placed around the patient’s body. Different structures in the human body have different water content, and, therefore, reflect different intensities of radiofrequency signal. The map of these transmitted signals represents a magnetic resonance image. However, obtaining each slice of the magnetic resonance image requires a new radiofrequency pulse to be transmitted and new excitation of water molecules to occur. For that reason, it may take a long time to obtain a magnetic resonance image of the human brain, containing multiple slices.

The slow acquisition of a signal through conventional MRI hampered the development of functional imaging. This problem was overcome by the development of echo planar imaging (EPI). EPI enables the spatial information of an object to be efficiently sampled after a single excitation of the water molecules. In this way, sampling an image can take less than 100 msec as opposed to the several tens of seconds required when using conventional MRI. By using ultrafast MRI with EPI, the brain can be visualized in almost real time. This rapid acquisition of data led to the development of fMRI.7

The first fMRI study using EPI principles was accomplished by injecting a bolus of paramagnetic contrast agent, gadolinium, into the bloodstream and comparing images of blood volumes before and after visual stimulation with flashing light. The results revealed an increase in the primary visual cortex blood volume during stimulation.8

Difficulties with injecting contrast medium have been overcome by using the blood oxygen level–dependent (BOLD) technique in all recent studies. BOLD fMRI relies on the different magnetic properties of oxyhemoglobin and deoxyhemoglobin. In active brain regions, an increase in the regional blood flow causes an increased ratio of oxyhemoglobin to deoxyhemoglobin, which is detected by fMRI as increased activity.9

BOLD fMRI has demonstrated good correlation with PET studies. There are multiple advantages of fMRI over PET, in particular: (1) fMRI does not require injection of radiotracers and, therefore, does not expose patients to radiation; (2) fMRI provides better spatial and temporal resolution than PET; (3) fMRI can be studied in individuals; and (4) every conventional MRI scanner can be converted to fMRI and potentially used for clinical purposes.

The scanners that are often used in hospitals for clinical purposes use 1.5 tesla (T) MRI. It has been demonstrated that magnetic fields higher than 1.5 T result in an improved contrast and elevated sensitivity to the BOLD technology used in fMRI. Recently developed magnetic fields of up to 4 T can provide even better fMRI results.

Many areas in medicine can potentially benefit from fMRI imaging. There are already many reports of fMRI use in brain mapping during auditory, olfactory, visual, somatosensory, and motor tasks. Furthermore, fMRI has been used to study brain activity during complex mental tasks and in people with psychiatric disease and addiction. fMRI could be used for presurgical mapping, to identify certain brain functions to be spared, or in patients with seizure disorders as a way of mapping epileptic foci.

Although there are far more published PET reports on regional brain activity in pain states, recent developments in fMRI provide excellent insight into brain functioning, with even better spatial and temporal resolution than PET.

A study comparing warm (41°C) and noxious heat (46°C) stimuli applied on the dorsum of the hand in human volunteers revealed increased activity in multiple brain regions.10 Positive activation was found in the anterior and posterior cingulate gyri, thalamus, motor cortex, somatosensory cortex (SI and SII), insula, cerebellum, frontal gyri, and supplemental motor area. Negative activation was observed in hypothalamus and amygdala. All areas activated by 41°C were activated by 46°C; however, significantly more activation was found with the 46°C stimulus, except for the thalamus and SI brain regions. With both temperatures, four stimuli were delivered, dispersed by stimuli-free intervals. A significant attenuation of the signal change was observed over the four stimuli at 46°C, but not at 41°C. The signal attenuation concurs with previous studies on time courses of peripheral nociceptor sensitization to noxious heat. The study results showed involvement of somatotopic and limbic brain areas in pain processing. They also revealed an integrative role of various brain areas in pain, as well as the adaptive nature of the pain system to noxious stimuli.

In another study, brain activation was produced by noxious electrical stimuli at two frequencies (5 Hz and 250 Hz) in six human volunteers.11 Both stimuli produced similar subjective levels of pain intensity, but later psychophysical measurements demonstrated a significant difference in pain unpleasantness between the two frequencies. An increased fMRI signal was observed in multiple brain areas, including anterior and posterior cingulate gyri, insulae, and frontal gyri, after stimulation by both frequencies. In these regions, the 5-Hz stimulation produced larger volumes of significant signal change. Other brain areas, such as medial thalamus, brainstem, precentral gyrus, and temporal gyrus, demonstrated significant fMRI signal change only by 5-Hz stimulation. A significant decrease in the fMRI signal was observed in similar brain regions by both 5-Hz and 250-Hz stimuli. These findings suggest that different neural networks are involved in processing sensory-discriminative and affective components of pain, and that the subjective experience of pain unpleasantness is subserved by qualitative and quantitative changes in CNS activation patterns.

Using fMRI , brain activation was measured during the thermal painful task and its overlap compared to cortical areas activated during vibrotactile and motor tasks in ten healthy volunteers.12 Predictor functions were generated from fMRI impulse response function, the time courses of thermal stimuli, and associated pain ratings. The correlation between these predictor functions and cortical activations in the painful thermal task indicated a gradual transition of information processing anteroposteriorly in the parietal cortex. The region best related to the thermal stimulus was the insular cortex at the level of the anterior commissure, and Brodmann’s area 5/7 was the region best related to the pain perception.

Hypersensitivity of the skin to non-noxious stimuli (allodynia) and increased sensitivity to noxious stimuli (hyperalgesia) are commonly found in states of central sensitization and in chronic neuropathic pain states. Capsaicin-induced secondary hyperalgesia in healthy volunteers presents a form of experimentally induced central sensitization. One fMRI study compared brain activation patterns between light touch stimuli and hyperalgesia in healthy volunteers.13 Nonpainful mechanical stimulation of the skin resulted in activation of contralateral somatosensory cortex (SI) and bilateral secondary somatosensory cortex (SII). To the contrary, hyperalgesia induced activation of contralateral prefrontal cortex, along with the middle (Brodmann’s areas 6, 8, and 9) and inferior frontal gyri. Prefrontal activation is thought to result from cognitive evaluation, attention, and planning of motor behavior in response to pain.

A study designed to compare activation of thalamic nuclei, insula, and SII to noxious and tactile stimuli was performed using fMRI.14 Painful thermal stimuli activated regions within the lateral and medial thalamus, predominantly anterior insula, and the contralateral SII in half of the subjects. The innocuous thermal stimuli did not activate SII in any of the subjects but activated the thalamus and posterior insula in 50% of subjects. The innocuous tactile stimulation consistently activated SII bilaterally and the contralateral lateral thalamus. These results indicate different pathways involved in processing somatotopic and emotional aspects of pain. They also demonstrate good spatial resolution of fMRI, enabling researchers to study specific brain nuclei involved in pain processing. Similar fMRI studies15,16 focus on somatosensory cortex and insula.

Spinal cord stimulation is used routinely in the treatment of chronic neuropathic pain, with success rates exceeding 50%. The exact mechanism of pain relief is unclear, but it is thought to result from stimulation of sensory tracts in the dorsal column and inhibition of nociceptive pathways, as described by Melzack and Wall’s gate control theory. Other putative mechanisms for pain relief might be an increase in spinal cord concentration of γ-aminobutyric acid (GABA), an inhibitory neurotransmitter. An fMRI study that investigated brain activation in three patients with chronic lower extremity pain during spinal cord stimulation17 found significant activation of contralateral somatosensory cortex in all patients and of anterior cingulate gyrus in one patient. These findings correlated with patients’ pain relief. Although a case study, this report sheds some light on the mechanisms of action of spinal cord stimulation for the relief of intractable pain.

Visceral pain is transmitted to the brain via slightly different neural networks. The main difference between somatic pain and visceral pain is poor representation of visceral organs in somatosensory brain areas. For this reason, visceral pain is often poorly localized, hampering its diagnosis and treatment. Commonly, visceral pain is referred to the distant site as a result of convergence of somatic and visceral neurons. In another fMRI study, esophageal stimulation was performed by intraesophageal perfusion of 0.1N HCl (1 mL/min), and the results of brain activation were compared with those of saline infusion and balloon distention. Both balloon distention and acid perfusion resulted in significant brain activation, but there was no increase of brain metabolic activity with saline perfusion.18

Leão spreading depression provides a possible explanation for the development of visual migraine headache attacks. These attacks were thought to be associated with decreased cerebral blood flow in affected brain regions. Occipital cortex activation was measured during visual stimulation using fMRI in patients with migraine and in healthy volunteers.19 In 50% of the patients with migraine, the onset of headache or visual change was preceded by suppression of brain activation that propagated into the occipital cortex. Initial suppression was accompanied by signal increase, indicating vasodilation. The vasodilation that follows the initial stage of hypoperfusion may be responsible for the onset of headache in migraine attacks.

Phantom pain following amputation of the extremity is a good example of the CNS role in chronic pain. An fMRI case study of a patient after arm amputation reveals a striking example of acute CNS plasticity following amputation.20 The patient’s phantom limb pain was activated by tactile stimulation of the ipsilateral corner of the mouth 24 hours after amputation. fMRI findings 1 month later revealed contralateral activation of the somatosensory (SII) area in the vicinity where the hand is thought to be represented. These findings were accomplished by tactile stimulation of the ipsilateral corner of the mouth, producing phantom pain in the ipsilateral hand.

Because they process affective and sensory experience, brain pain pathways are complex, changing, and often influenced by psychophysical settings. Only an anticipation of pain produces brain activation in the medial frontal lobe, insular cortex, and cerebellum, as measured by fMRI.21 These areas are distinct but close to the brain regions activated by pain, which indicates that they share similar CNS networks.

Before the era of functional brain imaging, all information about the CNS role in pain processing was derived from animal studies. It was assumed that nociceptive signals made their way to the dorsal horn of the spinal cord, were transmitted to the thalamus, and reached the somatosensory cortex, finally entering the mind through an unclear process.22 Many theories proposed before the advent of functional brain imaging focused on a specific “brain center” responsible for pain. Somatosensory cortex (SI) was thought to have such a role. In the last decade, functional brain imaging studies revealed many brain regions responsible for pain perception (Table 10-1), a scenario known as distributed pain processing.23

Because pain is an emotional experience, brain regions involved in emotions are also involved in pain processing. The best example of these areas is the anterior cingulate gyrus. Cingulotomies have been performed for the treatment of obsessive-compulsive disorder, severe depression, and pain. Indeed, postcingulotomy patients reported sensation of pain but no emotional response attached to it. Medial and lateral thalamic nuclei were thought to have a role in processing affective and somatotopic components of pain, respectively, based on animal experiments. Those data were partially replicated in fMRI studies.14 Several studies have supported the role of insula in affective and motivational aspects of pain.24

Somatosensory cortex has been visualized in some studies but not in others. The SI activation can be highly modulated by factors that alter pain perception. On the other hand, inter- and intrasubject anatomic variations can lead to variable SI activation. Besides having a somatotopic role, the other functions of SI region in pain processing remain to be discovered.

Multiple other brain areas are activated by pain. The pattern of activation varies across the studies and subjects. There are several explanations for these phenomena, but the major reason probably lies in the fact that pain is a complex experience and multiple psychophysical factors modulate perception of pain. Future functional brain imaging studies will gradually explain the complex brain circuitry involved in pain processing.

Functional brain imaging is a new and exciting field that enables clinicians to better understand higher CNS functions in pain states. It is hoped that this methodology will gradually evolve into a useful clinical tool, enabling practitioners to better diagnose and treat painful conditions in everyday practice.