Functional imaging techniques applied for the study of pain are positron emission tomography (PET), functional magnetic resonance imaging (fMRI), near-infrared spectroscopy (NIRS), multi-channel electroencephalography (EEG) and magnetoencephalography (MEG), and intracranial recordings. PET measures cerebral blood flow, glucose metabolism, or neurotransmitter kinetics. A very small amount of a labeled compound (called the radiotracer) is intravenously injected into the patient or volunteer. During its uptake and decay in the brain, the radionuclide emits a positron, which, after traveling a short distance, “annihilates” with an electron from the surrounding environment. In case of the most common use of O¹⁵-water injection, counting and spatial reconstruction of these occurrences within the brain anatomy allow visualization of the regional cerebral blood flow response (rCBF) as an indicator of neuronal activity. Radio-labeled fluorodeoxyglucose (FDG) is applied to measure regional energy consumption as a function of metabolic rate. An interesting refinement of PET technology represents the use of neurotransmitters as tracers to investigate binding mechanisms and kinetics, for example, in the opioidergic system.

One method of fMRI images blood oxygenation, a technique called blood oxygen level-dependent (BOLD), which exploits the phenomenon that oxygenated and deoxygenated hemoglobin possess different magnetic properties resulting in different relaxation behavior following radiofrequency pulses inside the magnet. An additional fMRI technique called arterial spin labeling (ASL) uses a radio pulse to magnetically “label”
hydrogen atoms ascending through the vasculature and then determines alterations in paramagnetic characteristics as the labeled blood perfuses the brain. The rCBF measures using O¹⁵-water PET, the BOLD technique, and the ASL technique rely on neurovascular coupling mechanisms that are not yet fully understood, but which overcompensate local oxygen consumption, thus causing a flow of oxygenated blood into neuronally active brain areas in excess of that used.⁴

NIRS is an optical approach that also analyses changes in hemoglobin oxygenation levels by using light in the near infrared range (650 to 950 nm). Shortly, NIRS measures the attenuation of light over the cortex and can be applied noninvasively through the skull. Another advantage is the portability of NIRS systems and the possibility to measure in different environments. However, the spatial and temporal resolution in noninvasive measurements is low for NIRS.⁵

EEG and MEG are noninvasive neurophysiologic techniques that measure the respective electrical potentials and magnetic fields generated by neuronal activity of the brain and propagated to the surface of the skull where they are picked up with EEG electrodes or, in the case of its magnetic counterpart, received by supra conducting quantum interference device (SQUID) sensors located outside the skull. Compared with PET and fMRI, EEG and MEG are direct indicators of neuronal activity and yield a higher temporal resolution of investigated brain function. The spatial distributions of EEG potentials and MEG fields at characteristic time points following noxious stimulation are analyzed using an inverse mathematical modeling and spatial filtering to estimate the anatomic origin of the recorded pain-related activity. The spatial acuity of MEG is higher than that of EEG because the latter measures the extracellular volume currents that are distorted by the differentially conducting tissues such as gray and white matter, cerebrospinal fluid, durae, and bone. In contrast, MEG measures the magnetic field perpendicular to the intracellular currents undistorted by the surrounding tissue.

Invasive recordings, which are assessed during neurosurgical interventions in patients ongoing epileptic surgery, or deep brain stimulation (DBS) procedures (most common in Parkinson disease), are extremely helpful to directly measure supraspinal pain signals. Different invasive recording techniques, namely, depth electrodes, subdural grids, subdural strips, and stereoencephalography, are used today.⁶ Advantage of all these invasive recordings is that the neuronal activity is picked up directly from the neurons and not being distorted by the cerebrospinal fluid, durae, and bone as in EEG or MEG. Disadvantage lies in the invasive nature of the procedure and the only limited area that can be investigated using invasive recordings.

The periaqueductal gray (PAG) is a midbrain territory located medially adjacent to the RF surrounding the cerebral aqueduct in a horse shoe-shaped manner by sparing ventral regions subserving distinct ocular-motor functions unrelated to pain.¹⁷ The PAG plays a critical role in the expression of a variety of emotion-related behaviors including pain.¹⁸ It represents a key structure in relaying descending pathways from the limbic system (prefrontal cortex, amygdala, and hypothalamus) to midbrain (e.g., inferior and superior colliculi), pons (e.g., locus coeruleus, lateral parabrachial nucleus, dorsolateral
pontine tegmentum), the raphe nuclei of the rostroventral medulla (RVM) and the deeper layers of the spinal cord (Rexed layers V, VII, VIII). Midbrain, pontine, medullar, and spinal connections are all reciprocal, such that the PAG is connecting afferents from these origins to medial thalamic nuclei and hypothalamus.¹⁹ Early systematic studies identified PAG and RVM as brainstem sites that elicit powerful surgical levels of analgesia through focal brain stimulation,²⁰ subsequently more elaborated and referred to as stimulus-induced analgesia.^21,22,23 A milestone contribution to the understanding of the interaction between PAG and RVM and its role in opioid analgesia was delivered by Fields and colleagues’²⁴ working group who identified two classes of pain modulatory cells in the RVM exerting inhibitory and facilitatory actions through respective off- and on-cells. Off-cells are activated by local infusion of µ-opioid agonists, and their activity inhibits nociceptive transmission. In contrast, on-cells facilitate nociceptive transmission, are inhibited by local µ-opioids, and are activated by naloxone and morphine abstinence. Approximately 15% of RVM neurons are serotonergic and are neither on- nor off-cells and do not respond to opioids.²⁵ Some respond to baroreceptor input integrating cardiovascular and nociceptive function.²⁶ Descending serotonergic fibers from the RVM project to dorsal horn neurons via the dorsolateral funiculus and mediate the analgesic effect of opioid receptor activation. Also noradrenergic structures in the brainstem, namely, the locus coeruleus, are regarded to be pain inhibitory through activation of α₂-adrenoceptors on central terminals of primary afferent nociceptors (presynaptic inhibition), by direct α₂-adrenergic action on spinal pain-relay neurons (postsynaptic inhibition), and by α₂-adrenergic activation of inhibitory interneurons.²⁷ However, the central as well as the peripheral efficacy of pain inhibition can vary depending on neuroplastic changes following inflammation and injury.²⁷ Taylor and Westlund²⁸ recently reviewed evidences that argue for a pain facilitatory action of locus coeruleus neurons under conditions of neuropathic pain, especially at later stages after the traumatic lesion.

The biologic significance of endogenous pain control is generally seen in the context of behavioral conflicts in which the subject needs to disengage from pain in order to fight or escape at the presence of body injury. Analogous human life situations are sporting competition or combat, during which a subject may fail to be aware of even severe tissue damage, which becomes painful when the victim releases engagement in these activities. Thus, forebrain input to the PAG mediates contextual information, from the prefrontal cortex, the amygdala, the anterior cingulate cortex (ACC), and the hypothalamus about momentary behavioral goals, past experience, and bodily needs. Evidence furthermore indicates that injury and inflammation causing increased sensitivity toward painful stimuli (primary hyperalgesia) triggers the RVM as key structure of a feedback pain inhibiting circuitry.²⁹ There is solid evidence by fMRI that PAG and RVM are also mediating the placebo response in humans.^30,31

Using the expression of the immediate early gene, c-fos, as a marker of neuronal activation, an interesting regional distinction for deep versus cutaneous pain had been demonstrated within the midbrain PAG. Noxious stimulation of a range of deep somatic and visceral structures evoked a selective increase in Fos expression in the ventrolateral PAG column (vlPAG), whereas, noxious cutaneous stimulation evoked Fos expression predominantly in the lateral PAG column (lPAG).³² vlPAG and lPAG areas are suggested to represent different modes of behavioral adaptation characterizing inescapable and escapable types of pain, respectively. Earlier studies showed that both deep pain as well as microinjection of excitatory amino acids (EAA) into the vlPAG of freely moving animals evoked a response of quiescence, decreased vigilance, decreased reactivity, hypotension, and bradycardia. In contrast, cutaneous pain as well as activation of the lPAG-evoked fight-and-flight behavior, increased vigilance, hyperreactivity, hypertension, and tachycardia. Lumb et al.^33,34 presented evidence that differential representation of escapable and inescapable pain in the PAG extends to distinct representations of “first” and “second” pain, as indicated by the columnar distribution of neurones activated by inputs from respective Aδ and C nociceptors. Furthermore, the functional organization of projections from circumscribed regions of the hypothalamus to the different columns of the PAG indicates that the behavioral significance of the pain signal is also represented in brain regions other than the PAG. Such specificity may coordinate antinociception with adequate behavioral and autonomic responses to prevent damage, in case of an imminent threat, or promote healing when an injury is already manifest.

Finally, the PAG is a promising target in invasive DBS to treat chronic pain, first applied by Hosobuchi et al.³⁵ and Richardson and Akil³⁶ in humans. However, its use is limited due to loss of efficacy over time and intolerable side effects. It remains to be cleared whether more precise electrode positioning and simultaneous stimulation of distinct locations³⁷ or a better understanding of pain mechanisms that are sensitive or insensitive to DBS will help to achieve better outcome. In chronic neuropathic and central pain syndromes, DBS of the PAG appears less effective than nociceptive pain.³⁸

The mesolimbic system comprises dopaminergic brainstem sites, that is, the VTA and the SN projecting to the striatum. Neurons originating in the SN pars compacta innervate D1-and D2-receptors of the dorsal striatum (putamen and caudate nucleus) that are functionally integrated into the motor loop of the basal ganglia (see the following text). The nucleus accumbens of the ventral striatum is regarded the key structure of a mesolimbic circuitry receiving input from SN and VTA mediating motivational salience and valence of painful stimuli to drive pain avoidance and pain endurance depending on the situational context.³⁹ The mesolimbic circuitry is furthermore regarded important for the encoding of the rewarding effect of pain relief.⁴⁰ Chronic fibromyalgia and low back pain have been observed to be associated with reduced responses of the mesolimbic circuitry to salient stimuli^41,42 that could account for the depressive comorbidity in these patients.³⁹

The cell bodies of spinothalamic tract (STT) fibers are located in the most superficial layers, lamina I, the outer region of lamina II, and deeper laminae V to VI⁴⁹ according to the Rexed scheme. STT axons cross via the anterior commissure to the anterolateral portion of the contralateral hemisphere and have their main thalamic targets in lateral nuclei, namely, ventral posterolateral (VPL; from the body) and posteromedial (VPM; from the face) nuclei and the ventral posterior inferior (VPI) nucleus. These fibers contribute to thermal and pain sensation. The lateral thalamic nuclei have small receptive fields and mostly gradual stimulus response functions over nonnoxious and noxious intensities, representing the “wide-dynamic-range,” or to a lesser extent, over noxious range only, representing the “nociceptive-specific” type of cells. These features render lateral thalamic targets of spinal nociceptive afferents ideally suited for the encoding of spatial localization and intensity of painful stimuli, similar to the properties of touch. The sensory-discriminative determinant of pain is thus governed by a spinal afferent pathway that mainly reaches lateral thalamic nuclei, from where neuronal activity is projected into the contralateral primary (SI) and bilateral secondary (SII) somatosensory cortices and mid and posterior sections of the insula (see the following text). Invasive DBS of the VPL demonstrated improvement of chronic neuropathic and central pain.⁵⁰

Although direct connections of lamina I STT cells exist with medial thalamic nuclei, namely, the central lateral nucleus and intralaminar complex,^51,52,53 the major source of nociceptive input to the medial thalamus is likely indirect through the brainstem that relays spinoreticular, spinomesencephalic, and spinoparabrachial input from both superficial and deeper dorsal horn. Medial thalamic nuclei project densely into key structures of the limbic system, such as the ACC, the amygdala, the hippocampus, the anterior insula, and prefrontal cortex, which represent the perceived intrusion and threat by pain, referred to as affective-motivational and cognitive-evaluative determinants of pain.⁴⁸

The contention that slowly and rapidly conducting nociceptors project differentially to respective medial and lateral thalamic nuclei had been recently challenged. Using intracerebral recordings of laser-evoked potentials in patients receiving neurosurgery due to refractory partial epilepsy, Bastuji et al.⁵⁴ identified three major thalamic regions, namely, the centrolateral (CL), VPL, and anterior pulvinar responding equally and simultaneously to phasic nociceptive laser input. The authors conclude that at least part of the nociceptive input driven by rapidly conducting Aδ fibers reaches both lateral and medial thalamic regions arguing for rapid medial thalamic projection to cingulate structures to subserve rapid orienting reactions to, and withdrawal from, the noxious stimulus. This suggests that both medial and lateral thalamic sites contribute to the exteroceptive function of escapable pain.

In contrast, Craig⁵⁵ hypothesized that pain is a purely interoceptive perception such as hunger, thirst, or itch. It originates in specific lamina I neurons which impinge on specific thalamic nuclei, such as the posterior part of the ventral medial nucleus (VMpo) and the ventral caudal part of the mediodorsal nucleus (MDvc). These distinct thalamic nuclei relay afferent input to the dorsal posterior insula and caudal ACC, respectively, and form separate pathways regarded as important elements of a hierarchical system subserving homeostasis, linking thermal sensation and pain contributing to the sense of the physiologic condition of the body (interoception) with subjective feelings and emotion.