A functional magnetic resonance imaging study of human brain in pain-related areas induced by electrical stimulation with different intensities
Correspondence Address: Source of Support: National Natural Science Foundation of China
(FC No. 30870686) and Youth Innovation Foundation of First Affiliated
Hospital of Xi'an Jiaotong University (FC No. 2009YK7), Conflict of Interest: None DOI: 10.4103/0028-3886.73748
Source of Support: National Natural Science Foundation of China (FC No. 30870686) and Youth Innovation Foundation of First Affiliated Hospital of Xi'an Jiaotong University (FC No. 2009YK7), Conflict of Interest: None
Background: Pain-related studies have mainly been performed through traditional methods, which lack the rigorous analysis of anatomical locations. Functional magnetic resonance imaging (fMRI) is a noninvasive method detecting neural activity, and has the ability to precisely locate related activations in vivo. Moreover, few studies have used painful stimulation of changed intensity to investigate relevant functioning nuclei in the human brain. Aim: This study mainly focused on the pain-related activations induced by electrical stimulation with different intensities using fMRI. Furthermore, the electrophysiological characteristics of different pain-susceptible-neurons were analyzed to construct the pain modulatory network, which was corresponding to painful stimulus of changed intensity. Materials and Methods: Twelve volunteers underwent functional scanning receiving different electrical stimulation. The data were collected and analyzed to generate the corresponding functional activation maps and response time curves related to pain. Results: The common activations were mainly located in several specific regions, including the secondary somatosensory cortex (SII), insula, anterior cingulate cortex (ACC), thalamus, and other cerebral regions. Moreover, innocuous electrical stimulation primarily activated the lateral portions of SII and thalamus, as well as the posterior insula, anterior ACC, whereas noxious electrical stimulation primarily activated the medial portions of SII and thalamus, as well as the anterior insula, the posterior ACC, with larger extensions and greater intensities. Conclusion: Several specified cerebral regions displayed different response patterns during electrical stimulation by means of fMRI, which implied that the corresponding pain-susceptible-neurons might process specific aspects of pain. Elucidation of functions on pain-related regions will help to understand the delicate pain modulation of human brain.
Keywords: Brain, electrical stimulation, fMRI, pain
Pain refers to the subjective impressions on noxious stimulation, which is usually accompanied by somatic movements, autonomic nerve responses, and emotional reactions. The subjective perception alters with variation in intensity of external stimuli. Previous pain-related experiments could not make precise neuroanatomical localization  or quantitative analysis to the stimulation parameters. Functional magnetic resonance imaging (fMRI) helps researchers to understand in vivo detection, localization, and investigation of pain susceptible areas in brain. This curtails the defects of traditional biochemical techniques and establishes a new way for research on central nervous system. Improvement in neuroimaging techniques has facilitated us with detection of pain perceptions, represented by variation in neural substrate. Different regions of the brain probably deal with different aspects of pain, and these regions may be activated by the corresponding levels of stimulation intensities.  On anatomical basis, two separate pain systems, the lateral system and the medial system, could be identified according to the different neurons and afferent fibers types. The lateral system, which is relatively fast and somatotopic, mainly comprises of spinothalamic projections to the contralateral side of lateral thalamic nuclei, SII and SI. The medial system includes polysynaptic and poorly lateralized, nonsomatotopic projections to lentiform nuclei, insular cortex, and anterior cingulate cortex (ACC).  When a fine spatial analysis was performed, the intensity of activation increased along with the stimulus level increased, and the distinctiveness in activation was identified.  However, most prior pain researches regarded thalamic activation as a whole, which lacked the delicate study on pain modulation conducted by subcomponents of the thalamus.  What's more, the activation modes of ACC in several studies were different from our result.  Additionally, several studies used the averaged value over the entire ACC in a region of interest (ROI) in fMRI. ,
In this paper, we used fMRI to investigate the pain-modulation mechanism of brain implicated with reaction to electrical stimulation. The innocuous and noxious tasks were chosen as our study objects to uncover some important features of the cerebral subregions in terms of the human pain-modulation mechanism.
The study subjects for the experiment included 12 healthy right-handed volunteers (23-38 years, 8 males). All the subjects had similar educational background. None of them had a history of neurological and psychiatric diseases. Each volunteer gave written informed consent, which was approved by the ethics committee of First Affiliated Hospital of Xi'an Jiao Tong University, School of Medicine.
The somatic pain threshold for each volunteer was measured by the Medtronic Keypoint-4 Evoked Potential EMG Measuring System (Denmark). A pair of electrodes with conductive pastes was put on the volunteers' dorsal skin of left ankle and tibia separately. Direct square wave current began at 0 mA and was increased by 0.1 mA/s. When the feeling of pinprick was obtained, the stimulus intensity was recorded immediately, designated as the pain threshold. After 10 min, the identical procedure was performed again, and then the mean pain threshold was calculated eventually.
The pain stimulations were generated by the same device, and the stimulation position remained the same as the previous measurement site of the somatic pain threshold. Three current intensities ranging from 0.5 mA below each mean pain threshold to 2.0 mA above it were used as pain tasks. All the volunteers were exposed to all the pain intensities inside the MR scanner for 20 min continuously before scanning. To avoid spatial sensitization and habituation, the selected stimulation sites were slightly changed after each stimulation. To minimize the temporal pain expectancy, three task stimuli were performed on the volunteers in the random order. During scanning, one investigator stayed with the volunteers in the scanning room and each volunteer was free to withdraw halfway if any unpleasantness arose.
Pain task paradigm
An event-related pain generation paradigm was adopted to explore the neural networks related to the pain modulations of the volunteers. Each pain task consisted of 19 repeated trials. In each trial, the task stimulus was kept constant for 3 s. To minimize the effect of temporal pain expectancy, the intervals between stimuli were randomized from 21 to 30 s. The whole procedures, including three stimulation tasks, lasted for 18 min. The paradigm was illustrated in [Figure 1].
All the scanning procedures were performed using 1.5 T MR scanner (GE Signa Horizon, Milwaukee, WI) with standard head coil. The 22 axial T1WI reference images were first acquired with SE sequence (TR/TE 500 ms/60 ms, Matrix 256 Χ 256 pixel, FOV 24 cm Χ 24 cm, slice thickness 3 mm, gap 1 mm), then T2*WI reference images were acquired in the same slice with GRE-EPI sequence (TR/TE 3000 ms/50 ms, Matrix 256 Χ 256 pixel, Flip angle 90°, slice thickness, gap, and FOV were the same to SE sequence). Finally, 160 continuous sagittal slices were obtained using FSPGR sequence (TR/TE 25 ms/3 ms, Matrix 256 Χ 256 pixel, FOV 24 cm Χ 24 cm, slice thickness 1 mm, gap 0 mm) for high resolution, 3D structural images, and spatial realignment.
Imaging processing was carried out using AFNI (version 2.56, Medical College of Wisconsin, USA). The 3D structural images were first realigned to T1WI reference images, spatially normalized to Talairach template. Then, the functional images were registered with motion correction and coregistered with 3D anatomic images. Then, the functional data were spatially smoothed with a 6-mm full-width half-maximum isotropic Gaussian kernel. The fMRI responses were sorted into innocuous and noxious events. A model for estimating cerebral responses was built based on the hemodynamic response functions of the stimulus intensity by using 3D deconvolution program of AFNI. Then, the real observed response data of the brain were compared with those from the estimated model. An F-statistic test was applied to test the fitness between the estimated and the observed responses in each voxel for the pain-generation tasks. Statistical maps were extracted at the threshold, F > 2.868 (P < 0.01, uncorrected), and superimposed on the high resolution anatomic images. The mean time-signal courses in the ROI were analyzed by averaging 19 trials of functional data to generate a 16-s time course of BOLD signal changes with temporal resolution of 1 s. Different time courses belonging to the given ROIs were analyzed to inspect the effects of different stimulation conditions. The volunteers' responses to the different stimulation intensities were characterized by evaluating the BOLD signal intensity variations in each ROI.
The mean pain thresholds varied from 1.15 to 2.20 mA across group [Table 1]. The activated cerebral regions showed the obvious distinction between groups.
Effect of nonpainful stimulation (innocuous task)
A set of cortical and subcortical areas were engaged by nonpainful stimulus (0.5 mA below each pain threshold). The contralateral activations were found in the primary somatosensory cortex (SI), inferior parietal lobule, superior temporal gyri, posterior insula, and lateral thalamus. Bilateral activations were observed in the lateral SII, posterior portion of anterior cingulate cortex (pACC), and middle frontal gyri. In terms of Z scores, the strongest activity was located at SII, which also prevailed in effect size among all the activated regions [Figure 2]a.
Effect of slightly painful stimulation
The activations evoked by slightly painful stimulation (intensity equal to the pain threshold) were spatially similar to the nonpainful stimulus except in the lateral part of ipsilateral thalamus and anterior portion of bilateral anterior cingulate cortex (aACC). In addition, SII still demonstrated the most conspicuous activations in all regions in terms of Z scores.
Effect of markedly painful stimulation (noxious task)
Prominent activations engaging both hemispheres were found in response to the markedly painful stimulation (2.0 mA above each pain threshold) compared with the nonpainful one [Figure 2]b. Meanwhile, the activation maps also expressed higher response magnitudes (higher Z values) and larger spatial extensions (larger effect size) in many cortical and subcortical structures. In contrast to the innocuous task, the noxious task simultaneously activated the bilateral anterior insula, medial thalamus, and putamen. Additional ipsilateral activations were localized in the supplementary motor area, premotor cortex, and contralateral activations in the inferior frontal gyri. Midline activation was observed in most of the ACC, including moderate activations in aACC and prominent activations in pACC. Further investigation also showed bilateral-activated areas, composed of the SII, insula and thalamus, displayed predominant contralateral activations. Simultaneously, Z values and activated spatial extensions in noxious task (i.e., marked painful stimulus) were dramatically enhanced compared to the innocuous one (i.e., nonpainful stimulus).
On the basis of the activation maps of three different stimuli, we selected the lateral portion of contralateral SII for the innocuous task, together with the medial portions of contralateral SII and thalamus, as well as the bilateral pACC for the noxious task, as the ROIs. The averaged time courses of BOLD signal were extracted from each ROI. Intriguingly, the temporal properties of these curves possessed featured characteristic. They all reached the peaks at 5-6 s and returned to the baselines at about 16 s after the stimulations [Figure 3].
Since the electrical nerve stimulation was first used in the functional brain imaging studies in 1995,  a variety of stimulation paradigms were adopted to deepen the understanding how the human brain modulated the pain. However, in this work we chose direct current of which the stimulation parameters could be easily quantified as the stimulus to study the human brain networks. The altered current intensity induced variant modulatory patterns over human central nerve system.
The SII is usually considered as one of the classical cerebral regions modulating pain. The most reliable pain-related activations through the previous functional imaging studies have located in a broad spectrum of regions, including the sylvian fissure, extending from SII to insular lobe.  Electrophysiological experiments in humans have demonstrated that the parasylvian cortex is the first cortical relay station in central process of almost all the nociceptive stimuli.  As a major result, the SII and insular cortex were classified into two subregions, which displayed distinctive BOLD activation patterns in our data. The lateral SII and posterior portion of contralateral insula were activated simultaneously during the innocuous stimulation, however, during the noxious stimulation, remarkable activations were displayed in most parts of the SII and insula, especially in the media SII and anterior insula. Regardless of the slight variations of activations in the SII and insular cortex, both regions could be divided into two distinct cerebral subregions. The segregated activations were due to the differences between the microanatomical structure and physiological characteristics of neurons in most cerebral subregions.  However, the superficial electrical stimulation as a nonspecific method could evoke both Aδ and C fibers at one time,  and the innocuous and noxious elements would be involved in these fibers, respectively. As a result, it tended to speculate that the lateral SII and posterior insula are more related to the innocuous sensory components, while the medial SII and anterior insula are probably involved in the nociception and pain.
Another finding in our study was that the SII and insula displayed contralateral-biased responses to the stimuli, which was in accordance with previous results.  The asymmetric activations in these areas could be reasonably explained by the convergent somatosensory inputs to contralateral SII and insula.  The contralateral SII receives both nociceptive and tactile inputs from periphery via ventroposterior medial thalamic nuclei, furthermore, central inputs from primary somatosensory cortex (SI) of the same hemisphere projects fibers into this region, in comparison the ipsilateral SII mainly receives nociceptive, transcallosal inputs from opposite SII. For this reason, we assumed that these distinct projections might be the basis of the predominant contralateral responses in our study.
Neurophysiological studies already confirmed that the thalamus belongs to the most important somatosensory relay nuclei, through which the sensory information of dorsal horn of spinal cord projects to the cerebral cortex. Three thalamic regions are involved in the pain experience as follows: the post-nuclei for designating stimulus as painful, the ventral posterior nuclei for localizing the painful stimulus, and the intraluminar nuclei for the affective-aversive nature of the stimulus.  Along with increase of stimulation, activations in the lateral subregion of thalamus were moderately enhanced compared to those in media portion, which were changed dramatically. A reasonable explanation was that two distinct types of pain-sensitive neurons may exist in thalamus: the neurons in lateral subregion demonstrated the stimulation-level dependent feature, which exhibited close relation to the sensory discrimination element of pain. In the medial part, the response was not dependent on the stimulation intensity but on sentiment and affection. In contrast to the lateral thalamus, the neurons in the medial thalamus might have had little baseline firing rate related to the stimulation intensity because of no apparent activations during the innocuous stimulation, but these neurons were capable of rapidly modulating the firing rate upward as the raise of simulation intensity during the noxious one. In sum, the spatial distribution of activations in thalamus during pain tasks provides evidence for complex integration of signals.
Asymmetric bilateral activations and contralateral-biased responses in thalamus were observed in most of the subjects, no matter innocuous or noxious stimulation was performed. With the precise analysis, this phenomenon mainly presented in the lateral portion of thalamus and was coincident with the results from other neuroimaging studies.  When it referred to the medial thalamus, the contralateral-biased activations were not obvious. Up to date most researchers believed greater contralateral nociceptive projections to lateral thalamus rather than to medial portion. , Meanwhile, in the medial thalamic region, there might be neurons with large bilateral receptive field, which had great chances to respond similarly regardless of stimulus side. Therefore, it was assumed that neurons in the lateral thalamus processed sensory component in pain modulation, which were concerned with stimulus side. The situation was contrary to the neurons in the medial portion of thalamus, which played a major role on arousal and affective aspects of pain.
Anterior cingulate cortex
The anterior cingulate cortex (ACC) is another essential cerebral region in pain modulation. A few researches showed ACC and adjacent regions in the medial wall had implicated roles in both sensory cognition and affective process of pain.  Anatomic studies have demonstrated the connections between the ACC and several thalamic subnuclei;  What's more, a ligand-PET study further supported the involvement of ACC in pain process, which used opioid agonist as tracers to show high-opioid receptor density in cingulo-frontal region,  and the studies in electrophysiological recordings have indicated the nociceptive neurons in middle portion of ACC (mACC).  As a whole, it is believed that ACC regulates a variety of elements of pain information. However, our data displayed discrete spatial activations within ACC, which could be evoked by the innocuous and noxious electrical stimulation. In order to highlight the functional spatial segregation, we divided ACC into two subregions, namely the anterior and posterior parts of ACC. Our results showed that the pACC was slightly activated in the innocuous task but displayed notable activations during the noxious stimulation, whereas there were no distinct activated variations between the nonpainful and painful stimulation in the anterior ACC. This finding was in accordance with the previous study  and different from other papers, which displayed in the perigenual ACC with increasing activation during self-administered but decreasing activation during externally applied stimulation.  We believed that the different activation patterns in aACC might be due to the diversity of applied stimulation approach. The time course in pACC of the painful task attained its top after 5-6 s, which was similar to the features of the SII and the thalamus despite distinctive shapes. On the basis of the basic characteristics of hemodynamic responses to event-related fMRI and correlative studies,  we believed the activations in these cerebral regions were induced by our stimulation tasks. In the innocuous task, the activations in pACC probably represented the affective component of pain that was beyond the basic sensory processing of pain-evoking activations, and further confirmation was the remission of chronic neuropathological pain after surgical intervention on ACC.  The response of aACC exhibited a plateau on three different stimulation intensities. This "on-off" activation mode was in closed relation to the functions of attention and arousal, which meant activations in aACC had no statically significant correlation with stimulation intensities, but with the presence of the stimuli. Neurophysiological research have discovered that single-cell work in patients undergoing cingulotomy demonstrated similar activity profile in aACC neurons, which located anterior to pain-responsive neurons.  In short, the aACC region displayed pain-intensity independent response mode.
The major drawbacks of this study were the lack of stimuli on left leg and multipoint stimulations, in which the electrodes could be put on upper limbs and trunk. We will pay more attention in further investigations with regard to these points and comprehensively analyze the stimulus-response features in each cerebral area. In general, functional MRI with electrical stimulation tasks is an effective way to analyze pain networks in human brain, rational selections of stimulation sites and parameters will help to draw favorable results on study of pain.
The authors would like to thank all the participants for being too supportive during the preparation of this manuscript.
[Figure 1], [Figure 2], [Figure 3]