Publications
APS Bulletin • Volume 7, Number 2, March/April 1997
Research Update
Robert J. Gatchel, PhD, Department Editor
Imaging Pain: Science and Technology
Patricia A. Gelnar, MD; Vania Apkarian, PhD
A 3-day meeting entitled "Imaging Pain: Science and Technology" took place in Syracuse, NY, March 21-23, 1996. The meeting focused primarily on a) the utilization of functional imaging techniques in the study of the pain system and b) understanding the technology of functional imaging and the various data analysis techniques currently in use. The meeting brought together individuals who use various functional imaging techniques to study the pain system, as well as prominent individuals involved in the development and refinement of positron emission technology (PET) and functional magnetic resonance imaging (fMRI) and analysis.
John Gore, PhD (New Haven, CT), started off the symposium Thursday evening with an introduction to fMRI, reviewing deoxyhemoglobin-induced susceptibility contrast effects that indirectly indicate neuronal activity. He discussed the many factors that limit our ability to detect small changes in cortical activity, such as artifacts produced by tissue motion and other physiologic noise. Results from various sensory systems indicate that the areas activated correspond closely to the regions one would expect, given prior neurophysiology. He emphasized that key factors in our ability to detect neuronal activity are the type of stimulus, the stimulus parameters used, and the method of data analysis. For instance, the ability to detect subcortical activity is enhanced by short stimulation periods with many stimulation cycles.
The Friday session began with a suggestion by Gary Duncan, DDS PhD (Montreal, QC, Canada), that perhaps the small changes in cortical activity observed with painful stimuli are not secondary to the stimulus itself, but rather constraints of the statistical approaches used to assess the activity. He proposed that instead of measuring activity in terms of the stimulus parameters (intensity, duration, frequency) presented to the subject, it would be more appropriate to define cortical activity in terms of the individual's perception of the stimulus. He composed traditional condition-based analysis techniques with a regression approach to psychophysical rating and regional cerebral blood flow (rCBF) using PET. The perception-based regression method yielded an increased sensitivity resulting in significant activity with greater than three times the volumes generated by the parameter-based approach. The areas activated were consistent with the regions defined in previous studies, but in this instance no areas ipsilateral to the stimulus were activated. These results led to the conclusion that for response to a noxious stimulus (which reflects both affective and stimulus-based components), cortical activity may indicate an individual's perception of reality rather than reality itself.
Anthony Jones, MD FRCP (Manchester, England), reviewed the cortical and subcortical structures identified thus far in functional imaging studies. He proposed that the individual regions process distinct components of the noxious stimulus. A precisely reproducible and well-controlled stimulus is necessary in order to distinguish the relative contributions of the different components of the task and cortical areas involved in processing those components. Jones suggested that a CO2 laser is the ideal stimulus for differentiating the components of the response to a noxious stimulus. Stimuli with increasing energy levels result in different rCBF response patterns during PET imaging. The thalamus is activated at pain threshold while cortical structures become involved only at higher intensities. Jones noted that these results support the idea of the thalamus being responsible for pain registration, while cingulate/prefrontal and insula responses contribute to affective, autonomic, and motivational aspects of pain.
Kenneth Casey, MD (Ann Arbor, MI), took a theoretical approach to interpreting functional imaging activity. He asserted that we cannot resolve definitive questions about the organization and pathology of the pain system unless we first resolve six fundamental issues:
- The relationship between pain perception and/or stimulus intensity and rCBF requires elucidation.
- The real meaning of a decrease in activity in functional imaging must be determined. Casey believes that such a decrease probably reflects suppression of a remote excitatory projection rather than a local inhibitory process, as local inhibition also requires local synaptic activity. Several participants disagreed with this interpretation; they felt that local inhibitory processes could result in a local decrease in activity as measured in a functional imaging study.
- The cortical activation map may depend upon the individual's past experiences and coping strategies, both of which are difficult or impossible to control.
- Theoretically, the spatial and temporal pattern of pain-induced activity should represent the functional connectivity of the activated regions.
- Cortical regions can respond to multiple tasks, but there must be something about either the spatial or temporal pattern of activity unique to each task.
- We must determine the normal spatial and temporal pattern of activity during pain before we can specify the functional reorganization that occurs within the central nervous system during a pathologic pain state. Having specified the normal pain response, we can then address the degree to which this plasticity is maladaptive, at what point it becomes permanent, and possible intervention or treatment during this reorganization.
Figures 1 and 2 are Casey's PET images showing statistical maps of some of the human brain regions with significant increases in synaptically induced blood flow during the perception of heat pain (50°C applied to the subject's left arm for 5 seconds). The blood flow responses to the perception of warmth (40°C) have been subtracted. After the maps were corrected for anatomical variations, they were superimposed on magnetic resonance images of an anatomically standard brain. The right hemisphere is on the viewer's left.
Figure 1. A highly significant response in the ventral posterior thalamus contralateral to the stimulus.
Activity is also seen in the ipsilateral thalamus and insular cortex bilaterally.
Figure 2. Highly significant activity in the medial dorsal midbrain in the region of the periaqueductal gray matter.
Activity is also seen in the ipsilateral paramedian cerebellum.
Satoshi Minoshima, MD PhD (Ann Arbor, MI), presented a thorough review of the methodology of PET data acquisition, registration, and analysis techniques. The accuracy of intrasubject PET-to-PET registration during a study is quite good, with an accuracy of less than 1 pixel translation and less than 1 (infinity) rotation in all planes. One can perform positron emission technology/functional brain imaging (PET/SPECT)-to-MRI registration by various techniques including matching internal structures, edge contours, and tissue homogeneity. Minoshima reviewed problems of intersubject image registration, the different techniques used to date including stereotaxy, and different methods of anatomic standardization. The method of intersubject anatomic standardization developed at the University of Michigan, while quite impressive, is currently useful only for PET/SPECT data. Minoshima also discussed the effects of different data normalization techniques (proportional, ANCOVA, and stochastic) relating to deactivation and subsequent statistical analyses. In the statistical analyses of activation data, he discussed various confirmatory and exploratory analyses. While PET activation methodology is now well established and allows the investigation of the functioning human brain, several outstanding issues remain, such as intersubject functional variability.
Nikolaus Szeverenyi, PhD (Syracuse, NY), introduced a novel way to deliver stimuli in fMRI studies, which also adds flexibility to data analysis. Szeverenyi described an automated stimulus delivery system that accurately synchronizes data collection and stimulus delivery. A software developed by Szeverenyi monitors the image collection and drives the stimulus. It triggers a precisely timed stimulus and documents which images are obtained during the stimulus and control periods. The test system delivers both low-threshold and noxious stimuli, and provides additional flexibility during data analysis.
Miroslav Backonja, MD (Madison, WI), introduced quantitative electroencephalography (QEEG) and magnetoencephalography (MEG), and discussed the advantages of using multiple techniques to elucidate both the spatial and temporal pattern of cortical responses to a painful stimulus. While PET and fMRI can reveal the spatial pattern of cortical responses, their temporal resolution does not approach the neurophysiologic range. In comparison to PET and fMRI, electrophysiologic methods such as QEEG and MEG offer far greater temporal resolution. QEEG reveals that multiple cortical structures undergo changes in activity in response to a painful stimulus, and that such changes occur in a distinct and predictable temporal sequence (activation followed by inhibition and then a return to baseline activity). This temporal pattern is interesting, but QEEG has limited spatial resolution. MEG can help to better localize the involved regions. MEG reveals activity changes in primary somatosensory (SI) and secondary somatosensory (SII) cortices, the insula, and opercular cortex at different temporal points during the painful stimulus. Backonja believes that combining the existing functional imaging techniques is necessary for elucidating the temporal and spatial pattern of responses to a painful stimulus.
Robert Weisskoff, PhD (Boston, MA), discussed ways to improve the signal-to-noise ratio by either increasing signal or reducing physiological noise. He discussed the deleterious effects on results of motion (both patient motion and movement of the brain due to the cardiorespiratory cycle), and reviewed studies addressing the motion of the brain due to variation in intrathoracic pressure and cardiac cycle. A take-home message for pain researchers is that stimulus-linked changes in cardiovascular status can increase noise and movement artifact.
The Saturday morning session started with Martin Ingvar, MD PhD (Stockholm, Sweden), stressing the importance of anticipatory mechanisms during normal cortical processing of a stimulus. He described two PET experiments designed to reveal activity due to anticipation of a painful stimulus. In the first experiment, the subjects were not forewarned about the painful stimulus. They knew only that sometime during the course of the scan they would receive a painful stimulus. In the second experiment, the subjects were familiar with the stimulus and the experimenter told them that during the course of the scan they would receive the familiar painful stimulus at a tolerable intensity. In neither case did the stimulus occur during the PET scan; instead, it was administered after the completion of the scan. The primary areas of activation in both experiments were located in the right hemisphere (the caudal anterior cingulate cortex [ACC] and the ventromedial prefrontal cortex [VMPFC]), but the type of activity differed in the two experiments. Experiment 1 (unknown stimulus) resulted in increased rCBF in both ACC and VMPFC, while experiment 2 resulted in decreased rCBF in both areas when compared to controls. Ingvar hypothesized that these areas are participating in an anticipatory coping mechanism involving distress and attention, and noted the importance of good experimental design in fMRI experiments for locating the areas contributing to differential response.
Jen-Chen Hsieh, MD PhD (Taipei,Taiwan, R.O.C.), addressed the question of separating cortical processing for the emotional and discriminative (sensorimotor) components of the painful response with PET imaging. Hsieh compared the cortical response to experimentally induced pain in normal subjects with the cortical response in people with existing pathologic pain. He hypothesized that, due to their ongoing inescapable pain, the individuals with pathologic pain would no longer experience the discriminative and motor components of the painful response but would exhibit primarily emotional suffering. Experimentally induced unilateral pain in normal subjects resulted primarily in a response pattern that is contralateral to the stimulus (SI, insula/putamen, ACC, supplementary motor area [SMA], premotor, posterior thalamus). The response pattern in individuals with existing pathologic pain (peripheral mononeuropathy and cluster headaches) was located primarily in the right hemisphere (most notably area 24 and the dorsal medial prefrontal cortex [areas 9 and 32]). Based upon these results, Hsieh contended that the emotional and sensory-discriminative components of the response are separable. The activity in the hemisphere contralateral to the painful stimulus represents a) the perception and localization of the painful stimulus and b) the intention to initiate (or inhibit) an escaping motor behavior. In contrast, the cortical response in the nondominant hemisphere represents the emotional suffering associated with the painful response.
Vania Apkarian, PhD (Syracuse, NY), reviewed the results of investigations using different matrix sizes and coil types (single surface coil, bilateral surface coils, whole head coil) during different tasks. Apkarian revealed that different neurophysiologic systems require unique scanning parameters. For instance, a motor task results in large changes in signal intensity that one can image using a large head coil, while the cortical activation resulting from a painful stimulus is much smaller, requiring the use of a surface coil for optimizing signal-to-noise ratio in the areas of cortical interest. He also discussed the cortical responses to a painful stimulus in pathologic pain patients with reflex sympathetic dystrophy (RSD). Each patient experienced noxious stimulus during the baseline painful state and (in the same imaging session) during a pain-free state following a sympathetic block. This revealed differential activity to the painful stimulus, depending upon the baseline pain. During the pathological pain state there was increased frontal cortical activity as compared to normal subjects and to RSD patients during the pain-free state. After sympathetic block of their baseline pain, noxious stimulation in RSD patients resulted in a cortical activity pattern similar to that seen in normal subjects, primarily increased activity in the primary somatosensory and motor cortices.
Robert Coghill, PhD (Bethesda, MD), pointed out that, while brain regions showing PET activation caused by painful stimulation are identified, quantitative rCBF changes remain unexamined. Coghill presented a PET study that investigated regional and global CBF changes during capsaicin-evoked pain and innocuous vibration. Capsaicin-induced pain resulted in a 22.8% decrease in global CBF from resting levels (p < 0.0001). Coghill asserted that the decrease in CBF was not caused by the cardiorespiratory parameters measured during the study. Ingvar observed that global blood flow can fluctuate by as much as 5%/mm pCO2. Coghill, however, stated that this applies only to increases in pCO2, not to decreases. However, the researchers did not measure pCO2 throughout the study, and as pCO2 levels can change very rapidly, the researchers may have underestimated the influence of decreases in pCO2 on the results.
Michael Iadarola, PhD (Bethesda, MD), presented a PET study designed to identify cortical activity specific to an individual task or condition. Iadarola compared activation maps produced by capsaicin-induced pain, capsaicin-induced allodynia, and low-threshold brushing. The results revealed partial overlap among the three cortical activation maps, but there were some regions in which activation was specific to one condition. The SII and inferior parietal regions were active primarily during the brush stimulus. Capsaicin-induced pain only resulted in activation of the cerebellar vermis and mesencephalon. Capsaicin-induced allodynia primarily activated bilateral prefrontal cortex (area 10). A similar study using fMRI compared the cortical activation maps produced by a low-threshold thermal stimulus, a noxious thermal stimulus, and a motor task involving the thumb. Both noxious and low-threshold thermal stimulation to the thumb produced activation of identical cortical regions (contralateral SI and primary motor cortex), but the area of activated cortex was smaller with the low threshold thermal stimulus than with the noxious stimulus. The motor tasks resulted in activity only in the motor cortex.
Larry Bradley, PhD (Birmingham, AL), used a multidisciplinary approach to study fibromyalgia pain. The study encompassed epidemiologic, functional brain imaging, and clinical data. Psychophysical studies revealed that fibromyalgia patients had low pain thresholds in comparison to normal subjects. Functional brain imaging (SPECT) of healthy controls and patients with fibromyalgia revealed that fibromyalgia patients also exhibit decreased rCBF in the thalamus and caudate nucleus relative to normals. Bradley also discussed preliminary data suggesting an association between the decreases in rCBF in fibro-myalgia patients and levels of substance P in the spinal fluid. Many participants congratulated Bradley on his comprehensive approach to a difficult problem.
Karen Davis, PhD (Toronto, ON, Canada), confirmed the involvement of both SI and cingulate cortex in pain processing using fMRI. SI activity occurred with both noxious and low-threshold stimuli, but noxious stimulation resulted in a greater increase in signal intensity, as well as larger regions of interest. Davis addressed the issue of whether the cingulate activity observed in pain studies indicates processing of the noxious stimuli or simply attention. Comparison among individuals of cingulate cortex activity during pain with cingulate cortex activation due to attentional tasks revealed that the responses are spatially distinct. Areas of activation in the cingulate cortex as related to a painful stimulus appear in the contralateral posterior area 24, while an attentional task activates areas bilateral and anterior and/or superior to the areas activated during pain.
This is only a brief description of the work presented at the meeting. During the meeting it became obvious that, while the cortical activation maps presented from many centers appear at first to be quite variable, there are in fact many similarities. Numerous cortical and subcortical structures are activated or deactivated by experimental painful stimuli. Investigators identified the caudal anterior cingulate gyrus and the primary and secondary somatosensory and motor regions with the greatest consistency. Perhaps even more important than the data presentations was the overall mood of the meeting, which allowed for a truly open and honest exchange of ideas. Investigators openly addressed problems they had encountered and offered useful suggestions to others.
While the presentations were informative and interesting, participants felt the discussion sessions were the highlight of the meeting. The final discussion session addressed approaches to reaching intelligent conclusions about the functional anatomy and physiology of pain, given that the studies from different centers are quite variable. Apkarian suggested that the inconsistent findings might reflect differing sensitivities of PET and MRI machines. Ingvar agreed that different PET machines have varying sensitivities. As it is impossible to standardize the sensitivity of each imaging system, Casey suggested that investigators try to standardize experimental stimuli, scanning conditions, and parameters. Finally, the participants discussed clinical application of these functional imaging studies of the pain system. The group believes that the field is still in its infancy. Until we are able to determine the functional connectivity (both temporal and spatial) of the cortical activation maps produced in normal subjects by these imaging techniques, our ability to determine pathology in the system is extremely limited. We may need to combine different imaging techniques in order to characterize the full extent of the functional connections between brain regions.
Patricia A. Gelnar is a resident in neurosurgery and a doctoral candidate in neuroscience at the State University of New York (SUNY) Health Science Center in Syracuse, NY. Vania Apkarian is director of the neurosurgery research labs and the computational neuroscience program at SUNY Health Science Center in Syracuse, NY.