B R A I N R E S E A R C H 1 3 1 5 ( 2 0 1 0 ) 1 1 1 – 1 1 8

ava i l ab l e a t www.sc i enced i r ec t . com

www.e l sev i e r . com/ loca te /b ra i n res

Research Report

Acupuncture needling sensation: The neural correlates of deqiusing fMRI

Aziz UR Asghara,b, Gary Greena, Mark F. Lythgoec, George Lewithd, Hugh MacPhersone,⁎aYork Neuroimaging Centre, University of York, Y10 5DG, UKbHull York Medical School and Department of Biological Sciences, University of Hull, Hull, HU6 7RX, UKcRCS Unit of Biophysics, UCL Institute of Child Health, University College London, WC1N 3JH, UKdComplementary Medicine Research Unit, University of Southampton, SO17 1BJ, UKeDepartment of Health Sciences, University of York, YO10 5DD, UK

A R T I C L E I N F O

⁎ Corresponding author. Senior Research FelloHeslington, York YO10 5DD, UK. Fax: +1904 3

E-mail address: hm18@york.ac.uk (H. Mac

0006-8993/$ – see front matter © 2009 Elsevidoi:10.1016/j.brainres.2009.12.019

A B S T R A C T

Article history:Accepted 7 December 2009Available online 16 December 2009

The needling sensation of deqi is considered by most acupuncturists to be an importantcomponent of acupuncture, yet neuroimaging research that investigates this needlesensation has been limited. In this study we have investigated the effect of deqi and acutepain needling sensations upon brain fMRI blood oxygen level-dependent (BOLD) signals.Seventeen right-handed participants who received acupuncture at the right LI-4 (Hegu)acupoint were imaged in a 3T MRI scanner. fMRI datasets were classified, on the basis ofpsychophysical participants' reports of needling scores, into those that were associatedwithpredominantly deqi sensations versus those with predominantly acute pain sensations.Brain areas showing changes in BOLD signal increases (activations) and decreases(deactivations) were identified. Differences were demonstrated in the pattern ofactivations and deactivations between groupings of scans associated with deqi versus painsensations. For the deqi grouping, significant deactivations occurred, whereas significantactivations did not. In contrast, the predominantly acute pain grouping was associated witha mixture of activations and deactivations. For the comparison between the predominatelydeqi sensation grouping and the acute pain sensation grouping (deqi>pain contrast), onlynegative Z value voxels resulted (mainly from deactivations in the deqi grouping andactivations in the pain grouping) in the limbic/sub-cortical structures and the cerebellumregions of interest. Our results show the importance of collecting and accounting for needlesensation data in neuroimaging studies of acupuncture.

© 2009 Elsevier B.V. All rights reserved.

Keywords:Functional magnetic resonanceimagingAcupunctureNeedling sensation

1. Introduction

Acupuncture needling sensation has been considered impor-tant to acupuncture practice since the early classical texts(Kong et al., 2007). A key term that relates to needle sensation is

w, Department of Health21388.Pherson).

er B.V. All rights reserved

deqi, which is assumed by many acupuncturists to beassociated with a therapeutic effect and for this reason isoften sought during needling (MacPherson et al., 2001). Tomonitor needling sensations, Vincent et al. (1989) adapted theMcGill Pain Questionnaire (Melzack, 1975), which was subse-

Sciences, Area 3, Seebohm Rowntree Building, University of York,

.

Table 1 – Psychophysical scores for deqi and pain.

deqi score pain score

deqi score>pain scoreSuperficial needling 5.3±0.3 3.0±0.2Deep needling 6.6±0.3 3.6±0.2

pain score>deqi scoreSuperficial needling 1.1±0.1 3.4±0.3Deep needling 3.6±0.6 5.9±0.7

Data are mean±SEM. For both superficial and deep needling, tenfMRI datasets were classed as deqi>pain and seven datasets wereclassed as pain>deqi.

112 B R A I N R E S E A R C H 1 3 1 5 ( 2 0 1 0 ) 1 1 1 – 1 1 8

quently adapted by Park et al. (2002) to include both pain anddeqi sensations. There have been a number of recent attemptsto clarify more precisely the needle sensations associated withdeqi. MacPherson and Asghar (2006) used expert opinion todelineate and weight the sensations of deqi from those thatmight be more properly associated with acute pain. Based on ahierarchical cluster analysis, a grouping of seven sensationswas found to be associated with the category of deqi (“aching”,“dull”, “heavy”, “numb”, “radiating”, “spreading”and “tingling”),and a grouping of nine sensations (“burning”, “hot”, “hurting”,“pinching”, “pricking”, “sharp”, “shocking”, “stinging” and“tender”) with the category of acute pain (MacPherson andAsghar, 2006). A number of other researchers have sought toestablish a credible rating scale for needling sensations,including the Subjective Acupuncture Sensation Scale (Konget al., 2005), the “deqi composite” (Hui et al., 2007), theMassachusetts General Hospital Acupuncture Sensation Scale(Kong et al., 2007), and the Southampton Needle SensationQuestionnaire (White et al., 2008).

There is considerable debate about whether specific acupunc-ture needling sensations, including deqi, are associated withtherapeutic benefit (Kong et al., 2007). Despite the importancethat practitioners have often placed on the deqi sensation(MacPherson et al., 2001), there has been limited research in thisarea.Anearlyexamplewasamechanisticstudythatdemonstratedcorrelations between acupuncture analgesia and sensations of“numbness”, “fullness” and sometimes “soreness”, sensations thatare often associated with deqi (Chiang et al., 1973). This study alsofound that deqimight be an important component of acupunctureanalgesia because both deqi and acupuncture analgesia could beblocked by intramuscular, but not subcutaneous, procaine injec-tions at acupuncture points (Chiang et al., 1973). Their resultsweresupported by a recent study that also found a similar relationshipbetween acupuncture analgesia, numbness and soreness, but notfor other sensations commonly associated with deqi (Kong et al.,2005). Inasmall clinical trial of acupuncture forosteoarthritis of theknee, the participants' experience ofdeqi sensationwas a predictorfor improved outcome (Takeda andWessel, 1994).

Our aim in the present investigation was to extend ourprevious fMRI analysis, which compared the impact ofneedling at two different depths (MacPherson et al., 2008), bygrouping fMRI datasets into those in which participantsexperienced predominantly deqi sensations and those withpredominantly acute pain sensations with needling at LI-4(Hegu). In the current investigation, we determine if anysignificant differences result from the comparison betweenthe predominately deqi sensation grouping and the acute painsensation grouping (deqi>pain contrast) using fMRI.

2. Results

2.1. Impact of needling sensation (within-group analysis)

For the grouping of scans in which deqi sensations predomi-nated (Table 1), we found no activations (increase in BOLDsignal) above our selected Z threshold level of >4.3 (correctedcluster threshold P=0.05), whether the needling was superfi-cial (Table 2A) or deep (Table 2D). In contrast, for the groupingin which acute pain sensations predominated (Table 1), the

peak activation voxels of the two most significant clusterswere located to the cerebellum (anterior lobe) and insula withsuperficial needling (Table 2B), and in the cerebellum (anteriorlobe) and anterior temporal lobe for deep needling (Table 2E).

In terms of deactivations (decrease in BOLD signal) associ-ated with deqi and acute pain groupings, we report the patternsof responses using a threshold of Z<−4.3 (corrected clusterthreshold P= 0.05) for both superficial and deep needling. Forthe predominately deqi grouping with superficial needling, thedeactivations for the peak voxels of the two most significantclusters were located in the middle temporal gyrus andfusiform gyrus (Table 2G). For the scan grouping with predom-inately deqi sensations with deep needling, the peak voxelsshowing deactivations were located in the fusiform gyrus andthe lingual gyrus (Table 2J). For the scans with predominatelyacute pain sensations with superficial needling, the peak voxeldeactivations were located in the fusiform gyrus and posteriortemporal lobe (Table 2H). For the scans with predominatelyacute pain sensations with deep needling, the deactivationswere located in the posterior temporal lobe (Table 2K).

2.2. Impact of needling sensation (between-group analysis)

For the deqi>pain contrast, with either superficial (Table 2C) ordeep needling (Table 2F), there were no significant voxels withpositive Z values. Significant negative Z values (Z<−2.3,corrected cluster significance threshold of P= 0.05) werefound with the contrast deqi>pain (deactivations in the deqigrouping and activations in the pain grouping); the peak voxelswere located in the limbic lobe (parahippocampal gyrus) withsuperficial needling (Table 2I) and in the cerebellum andthalamus with deep needling (Table 2L).

We used probabilistic masks of brain regions of interest inthe limbic/sub-cortical structures (insula, hippocampus, amyg-dala, thalamus, and the posterior/anterior cingulate gyrus) andthe cerebellum. Using these masks, a count was made of thenumber of significant voxels (Z<−2.3, corrected cluster signif-icance threshold of P= 0.05) in each of the regions of interest forthecontrast,deqi>pain. VoxelswithsignificantnegativeZvaluesfor deqi>pain (mainly from deactivations in the deqi groupingand activations in the pain grouping) were observed bilaterallyin the insula, hippocampus, amygdala, thalamus and cerebel-lum regions of interest for both superficial and deep needling(Table 3 and Fig. 1). In the posterior cingulate gyrus, significantnegative Z value voxels for deqi>pain (deactivations in the deqigrouping and activations in the pain grouping) resulted for both

Table 2 – Localizations of the deqi and pain groupings, and the deqi>pain contrast.

Voxels with maximum effect Cluster

x y z P value(−log10 P)

Z value % BOLD±SD

Numberof voxels

MeanZ value

Superficial needlingA. deqi grouping activations (n=10) No activationsB. pain grouping activations (n=7)Cerebellum (anterior lobe) R 10 −46 −30 17.1 7.2 0.38±0.05 359 5.1Insula (BA13 L −44 −6 −6 8.2 7.3 0.62±0.08 119 5.2C. deqi>pain contrast (unpaired t test,Z>2.3)

No significantpositive Z values

Deep needlingD. deqi grouping activations (n=10) No activationsE. pain grouping activations (n=7)Cerebellum (anterior lobe) R 10 −56 −26 >20 7.9 0.45±0.06 685 5.1Anterior temporal lobe(sub-gyral)/parahippocampal gyrusa

R 36 −12 −14 11.8 9.5 0.47±0.05 212 5.8

F. deqi>pain contrast (unpaired t test,Z>2.3)

No significantpositive Z values

Superficial needlingG. deqi grouping deactivations (n=10)Middle temporal gyrus L −60 −58 2 >20 −8.9 −0.55±0.06 1707 5.2Fusiform gyrus (BA20) L −44 −34 −22 9.4 −7.9 −0.37±0.05 178 4.9H. pain grouping deactivations (n=7)Fusiform gyrus (BA19) L −36 −74 −14 >20 −9.1 −0.62±0.07 1117 5.3Posterior temporal lobe(sub-gyral)/superior temporal gyrusa

L −32 −36 16 4.5 −7.1 −0.28±0.04 48 5.5

I. deqi>pain contrast (unpaired t test,Z<−2.3)

Significant negativeZ values

Limbic lobe, parahippocampal gyrus(BA35)

R 24 −30 −16 17.3 −4.9 (−1.8, 0.3) b −0.73±0.15 4069 2.7

Limbic lobe, parahippocampal gyrus L −22 −18 −30 5.5 −4.3 (−1.3, 2.2) b −0.88±0.14 992 2.7

Deep needlingJ. deqi grouping deactivations (n=10)Fusiform gyrus (BA18) L −22 −92 −16 >20 −9.6 −0.84±0.08 1284 5.7Lingual gyrus R 14 −90 4 14.7 −7.5 −0.45±0.06 512 5.2K pain grouping deactivations (n=7)Posterior temporal lobe(sub-gyral)/fusiform gyrusa

L −42 −44 −6 12.6 −7.4 −0.29±0.04 233 5.0

Posterior temporal lobe(sub-gyral)/middle occipital gyrusa

L −32 −62 8 7.2 −5.7 −0.26±0.05 98 4.8

L. deqi>pain contrast (unpaired t test,Z<−2.3)

Significant negativeZ values

Cerebellum (anterior lobe)c L −12 −58 −32 >20 −5.5 (−2.5, 2.5) b −0.48±0.08 9457 2.9Thalamusc R 18 −12 4 >20 −5.3 (−3.0, 1.8) b −0.31±0.06 9457 2.9

All Z value thresholding was >4.3 for activations, and <−4.3 for deactivations. The most significant clusters (highest P value, maximum of twoclusters reported) with a corrected threshold of P= 0.05 are presented along with the voxel having the peak Z value within each cluster. x, y, andz are MNI152 brain coordinates (mm). L, left; R, right. BA, Brodmann area.a Voxel is part of a cluster which extends to the stated brain area.b Values in parentheses are the Z values for the deqi grouping and the pain grouping, respectively.c As only one large cluster was significant, the two voxels reported for this cluster are given in order of peak Z value.

113B R A I N R E S E A R C H 1 3 1 5 ( 2 0 1 0 ) 1 1 1 – 1 1 8

depths of needling (Table 3 and Fig. 1). No significant voxelsoccurred in the anterior cingulate gyrus for deqi>pain despiteusing a probabilistic mask of ≥5% for this region of interest.

3. Discussion

Our primary finding demonstrated that the fMRI datasetsassociated with predominantly deqi and pain sensations

resulted in different patterns of BOLD signal increases(activations) and decreases (deactivations). For the deqigrouping, significant deactivations were seen without activa-tions. The scans grouped according to acute pain sensationswere associated with a mixture of activations and deactiva-tions. In the present investigation, we additionally consideredthe contrast of predominantly deqi>pain and found voxelswith significant negative Z values (mainly from deactivationsin the deqi grouping and activations in the pain grouping),

Table 3 – deqi>pain contrast: limbic/sub-cortical structures and the cerebellum.

Brain structure Superficial needling Deep needling

Number of voxels x, y, z (mm) deqi >pain maxZ valuea

Number of voxels x, y, z (mm) deqi >pain maxZ valuea

Left insula 75 [27–87%] −42, −10, 0 −3.7 (−0.9, 2.1) 8 [28–72%] −42, −12, −2 −3.1 (1.1, 2.4)Right insula 55 [26–83%] 36, 26, −8 −3.2 (−0.5, 2.0) 89 [30–88%] 44, 0, 0 −3.3 (−0.04, 3.2)Left hippocampus 85 [26–89%] −22, −20, −12 −3.5 (−0.8, 1.2) 119 [25–95%] −34, −22, −10 −3.5 (−1.2, 2.9)Right hippocampus 177 [25–94%] 26, −36, 0 −4.2 (−2.0, 0.1) 176 [26–97%] 24, −22, −12 −4.6 (−2.9, 2.4)Left amygdala 62 [25–87%] −24, −2, −28 −3.4 (−1.9, 1.2) 52 [26–98%] −32, −6, −22 −3.0 (−0.7, 2.4)Right amygdala 133 [26–97%] 26, 4, −22 −3.5 (−1.0, 2.3) 76 [25–100%] 28, −6, −16 −4.0 (−2.0, 2.6)Left thalamus 174 [27–100%] −18, −30, 14 −3.4 (−1.1, 1.7) 129 [27–100%] −2, −8, −4 −4.1 (−2.4, 2.0)Right thalamus 177 [25–100%] 6, −26, 2 −3.7 (0.4, 1.4) 437 [26–100%] 18, −12, 4 −5.3 (−3.0, 1.8)Posterior cingulate gyrus 43 [6–26%] 16, −36, 0 −3.3 (−2.4, 0.2) 166 [7–77%] 24, −48, −2 −4.1 (−3.0, 2.8)Anterior cingulate gyrus No significant voxels No significant voxelsLeft cerebellum 88 [29–100%] −14, −32, −32 −3.6 (−0.5, 0.3) 472 [28–100%] −12, −58, −32 −5.5 (−2.5, 2.5)Left cerebellum 294 [26–100%] 10, −42, −8 −3.6 (−2.0, 0.2) 1394 [28–100%] 8, −36, −18 −4.6 (−2.0, 2.8)

Data were thresholded at Z<−2.3 with a corrected cluster significance level of P= 0.05. No significant voxels occurred with a threshold of Z>2.3.The values in square brackets give the percentage range of the voxels being located within the probabilistic brain region of interest mask used.a Values in parentheses are the Z values for the deqi grouping and the pain grouping, respectively.

114 B R A I N R E S E A R C H 1 3 1 5 ( 2 0 1 0 ) 1 1 1 – 1 1 8

without any voxels having significant positive Z values. Takentogether, our results emphasize the importance of needlesensation, which should be accounted for during neuroima-ging studies of acupuncture. Furthermore, our classification ofgrouping scans according to the needle sensations of predom-inate deqi and pain sensations allows for finer stratification ofneuroimaging results, which may lead to improved interpre-tation of fMRI data. Thus, if deqi and acute pain needlesensations are not considered, our results imply that it ispossible to have a potential source of unaccounted variability.

Interestingly, in the limbic/sub-cortical structures and inthe cerebellum, significant negative Z values resulted forvoxels with the contrast deqi>pain (arising mainly fromdeactivations in the deqi grouping and activations in the paingrouping), without any voxels showing significant positive Zvalues. This result is consistent with our primary finding thatdeqi and pain sensations have different impacts upon theBOLD response to acupuncture at LI-4. Since limbic/sub-cortical structures and the cerebellum are closely involved inthe response to pain (Davis, 2000; Price, 2000; Wager et al.,2004), it is possible that BOLD signal decreases in these brainareas could mediate the putative analgesic effects of acu-puncture needling especially when deqi is present. However, itis not possible to elucidate whether this is a physically orpsychologically mediated mechanism of action. The involve-ment of the limbic/sub-cortical structures in acupuncture-induced BOLD signal decreases has been shown in severalinvestigations. For instance, during acupuncture needling atST-36 with deqi, Hui and colleagues reported a predominanceof signal decreases in various limbic/sub-cortical structuresincluding the insula, hippocampus, amygdala, thalamus,posterior cingulate gyrus and cerebellum, whereas signalincreases were uncommon (Hui et al., 2005). Similarly, theamygdala (acupoints LV2 and ST40) and hippocampus (acu-points LV2, LV3 and ST44) were shown to deactivate inparticipants who mostly experienced deqi without sharp pain(Fang et al., 2009). Wu and colleagues suggested that higherbehavioral scores associated with deqi sensations duringacupuncture at LI-4 and ST-36 were linked to the deactivationof multiple limbic system structures (Wu et al., 1999).

In thepresent study, no significant activationswere reportedfor the deqi grouping alone, whereas other fMRI studies ofacupuncture have reported at least some activations in, forexample, the secondary somatosensory cortex (Fang et al., 2004,2009; Hui et al., 2005). This difference could be explained byvariations in the psychophysical methods used to classify fMRIdatasets intodeqi groupings, use of different acupoints, or itmayhave arisen from differences in the statistical methodology andlevel of thresholding used, which was relatively conservative inour study (cluster thresholding, Z>4.3).

Our observation that acute pain sensations are associatedwith activations is congruent with previous research showingactivations at LI-4 in participants who experienced pain duringacupuncture (Hui et al., 2000). For the acute pain grouping, weobserved a combination of both activations anddeactivations atLI-4. Combinations of BOLDsignal increases anddecreases havealso been reported with needling at ST-36 in participants whowere classified as having a mixture of deqi and pain (Hui et al.,2005). Some neuroimaging studies of acupuncture havereported signal activations rather than deactivations in thethalamus, insula, cingulate cortex and in the cerebellum (Biellaet al., 2001; Fang et al., 2004, 2009; Hsieh et al., 2001; Yan et al.,2005; Yoo et al., 2004). Acupuncture at different acupoints, orinconsistent measurement of psychophysical ratings of acu-puncture needling sensation in these studies, could explain thedifferences in thepatterns of activations anddeactivations seenacross the various neuroimaging investigations of acupuncture.

3.1. Limitations of the present study

The BOLD signals identified in the current investigationresulted from analyzing a time series over a 16 min period,and identifying those voxels that reacted “on” and “off” in amanner that was synchronized with the periods of needlestimulation. Our two periods of 2-min stimulation of theneedle, while modeled on a traditional approach to enhancingthe therapeutic effect (Bovey, 2006), were of a longer durationthan is commonly used in clinical contexts, and perhapsexplain the acute pain sensation experienced by participants.It has also been argued that the subjective sensations of

Fig. 1 – Functional maps showing significant clusters for the deqi>pain contrast with deep needling (n=10 deqi grouping; n= 7pain grouping) for the selected regions of interest. Data were thresholded at Z<−2.3 with a corrected cluster significance levelof P= 0.05. The x, y and z (mm) brain coordinates of the maximal Z values are given in parentheses and indicated by thecrosshairs in each brain region of interest. The values in italics give the probabilistic percentage of the corresponding maximalvoxel being located in the region of interestmask used. Clusters have been overlaid onto aMNI152 standard brain. R, right brainhemisphere; L, left brain hemisphere.

115B R A I N R E S E A R C H 1 3 1 5 ( 2 0 1 0 ) 1 1 1 – 1 1 8

acupuncture are influenced by the type of needling techniques(see Hui et al., 2005). As we used a needle rotation at 2 Hz, ourresults are primarily applicable to the simple “even method”technique (Deadman and Al-Khafaji, 1998).

The physiological effects of acupuncture can continue formany hours, as exemplified by endorphin release (Han, 2004),and clinical benefits have been observed up to 2 years after acourse of treatment (Thomas et al., 2006). However, theputative longer-term impact of acupuncture on neural struc-tures in the brain could not be evaluated with our currentstudy design. Furthermore, our experimental paradigm didnot reflect usual practice, in that it was conducted over asingle session, with only a single needle inserted at oneacupuncture point. Our use of healthy individuals may alsolimit the generalization of our results (Lewith et al., 2005).

3.2. Experimental and clinical implications

Given the value many acupuncturists attach to the elicitationof the deqi sensation during acupuncture (MacPherson et al.,2001), it is interesting that the present results show that a

predominantly deqi sensation is associated with brain deacti-vations rather than activations. Whether or not such deqi-associated brain deactivations could constitute a possiblemechanismwhich underlies the therapeutic effect of acupunc-ture is not ascertainable from our current results, although thismay be a reasonable hypothesis to explore further.

Our results have potential implications for the design ofexperimental studies of acupuncture. For example, ourevidence suggests that investigators should collect data onparticipants experiences of needle sensation, perhaps usingone of the newly validated measures (Kong et al., 2005, 2007;MacPherson and Asghar, 2006; White et al., 2008). In addition,given that our findings indicate that needle sensation affectsspecific neural structures within the brain, researchers shouldaim to minimize the variability of needle sensation in orderthat it does not have the potential to act as a confounder whenanalyzing and interpreting the results. If an accepted stan-dardized and validated behavioral questionnaire could bedeveloped that effectively captured and rated the variousacupuncture needling sensations (especially deqi and painsensations) from study participants, this would allow formore

116 B R A I N R E S E A R C H 1 3 1 5 ( 2 0 1 0 ) 1 1 1 – 1 1 8

meaningful comparisons to bemade across neuroimaging andother studies of acupuncture.

3.3. Conclusions

The acupuncture needle sensations of deqi and acute pain areassociated with different patterns of activations and deactiva-tions in the brain. We found that grouping of fMRI datasetsinto those associated with predominantly deqi sensationsresulted in brain deactivations, but not activations. Scansgrouped according to predominantly acute pain sensationswere associated with a mixture of activations and deactiva-tions. For the deqi>pain contrast, only significant negative Zvalue voxels occurred in the regions of interest in the limbic/sub-cortical structures and the cerebellum. Our resultshighlight the importance of quantifying and accounting forvariations in needle sensations in neuroimaging studies ofacupuncture.

4. Experimental procedures

We recruited 17 healthy right-handed adult participants whowere naive to acupuncture; eight males (mean age 33 years,range 18 to 46) and nine females (mean age 39 years, range 20to 54), with an overall average age of 36 years. The YorkNeuroimaging Centre's Science and Ethics Committeesapproved the study.

4.1. Experimental paradigm

Manual acupuncture needling was performed at the acupunc-ture point Hegu (LI-4) on the right hand by an experiencedacupuncturist (HM). Sterile, disposable and non-magneticstainless steel needles (25 mm long and 0.28 mm diameter)manufactured by Hwato, China, were used. The insertiondepth of the needle into the first dorsal interosseous musclewas randomized to be either superficial (1–2 mm), ascommonly used in Japan (Birch and Felt 1999), or deep(8–12 mm), as commonly used in China (Deadman and Al-Khafaji, 1998).

Fig. 2 – The fMRI block design showing manual LI-4 superficial (participant two fMRI scanswere performed, onewith two consecutwo consecutive deep (D) needling stimulation periods. The ordertypes of needling was randomized.

Scans were conducted over two 16-min periods; one scaninvolved superficial needling and another separate scaninvolved deep needling. Participants were only informed thatthey would be receiving two types of acupuncture during theirtime in the scanner. For both superficial and deep needling thesame block design with two active blocks separated by restperiods was utilized (Fig. 2), a block design used previously(MacPherson et al., 2008). For both superficial or deepinsertion, two repeated 2-min periods of needle “stimulation”were used with an “even” method (Deadman and Al-Khafaji,1998). This involved the needle being rotated clockwise andanti-clockwise continuously at approximately two cycles persecond with an alternating 180° rotation.

4.2. Measuring needle sensation: deqi and acute pain

Immediately after each 16-min scan, participants were askedquestions about their experience of all 25 individual needlesensations from the Park questionnaire (Park et al., 2002).Participants rated the intensity of each sensation on anordinal scale: “none” (zero), “slight” (one), “moderate” (two)or “strong” (three). For each fMRI dataset and for eachsensation we multiplied the participants sensation score(0 to 3) by the expert weighting (0 to 8) for the seven deqisensations (aching, dull, heavy, numb, radiating, spreadingand tingling) and nine acute pain sensations (burning, hot,hurting, pinching, pricking, sharp, shocking, stinging andtender) (MacPherson and Asghar, 2006). We then groupedscans (see Table 1) into those associated with predominantlydeqi sensations (deqi scores greater than pain sensation scores)and those with predominantly acute pain sensations (painscores greater than deqi scores).

4.3. Functional and anatomical MRI acquisition

Functional and structural scanning was performed with a 3Tmagnetic resonance imaging system (GE Signa HD Excite),with an eight channel head coil (GE Signa Excite 3.0T, HighResolution Brain Array, MRI Devices Corp., Gainesville FL) andaxial images were acquired for the whole brain. For functionalimaging, EPI (Echo Planar Imaging) images were acquired

S) and deep (D) needling stimulation periods. For eachtive superficial (S) needling stimulation periods and onewith(whether superficial needling only or deep only) of these two

117B R A I N R E S E A R C H 1 3 1 5 ( 2 0 1 0 ) 1 1 1 – 1 1 8

using a T2⁎weighted gradient echo sequence (TR 4 s, TE 30ms,flip angle 90o, acquisition matrix 128×128, FOV 240 mm, in-plane resolution 2×2 mm, contiguous slice thickness 4.5 mm).The voxel size was 2 mm×2 mm×4.5 mm. High resolutionT1-weighted structural images were acquired prior to thefunctional scans using an IR (Inversion Recovery)-prepared3D-FSPGR (Fast Spoiled Gradient Echo) pulse sequence (TR7.5 s, TE 3 ms, flip angle 20o, acquisition matrix 256×224 inter-polated to 512×512, FOV 260 mm, in plane resolution0.5×0.5mm, slice thickness 2.6mmwith an overlap of 1.3mm).

4.4. fMRI data analysis

FSL software (FMRIB's Software Library, www.fmrib.ox.ac.uk/fsl) based on generalized linear modeling was used to analyzethe fMRI data. The design matrix had a single explanatoryvariable which was “on” during the periods of stimulation andotherwise “off” (see Fig. 2). For each participant, a first levelanalysis was carried out using FEAT (FMRI Expert AnalysisTool) Version 5.63, part of FSL. Following pre-statisticsprocessing (see MacPherson et al., 2008), a time seriesstatistical analysis was carried out using FILM (FMRIB'sImproved Linear Model) with local autocorrelation correction(Woolrich et al., 2001). Registration of EPI images to highresolution T1 images and then subsequently to the MontrealNeurological Institute standard brain (MNI152) was carriedout using FLIRT (Jenkinson et al., 2002; Jenkinson and Smith,2001). All x, y, z (mm) coordinates reported are MNI152 braincoordinates.

The terms activation and deactivation are only used torefer to an increase (positive Z values) or decrease (negative Zvalues), respectively, in the BOLD response in the context ofthe deqi grouping alone or the pain grouping alone. Higher-level within-group analysis was performed using FLAME(FMRIB's Local Analysis of Mixed Effects) (Beckmann et al.,2003; Woolrich et al., 2004). Z (Gaussianized T/F) statisticwithin-group images were thresholded using clusters pre-selected by Z>4.3 (significant activation) or Z<−4.3 (significantdeactivation) and a corrected cluster significance threshold ofP= 0.05 (Worsley et al., 1992). Significant clusters were super-imposed onto the MNI152 standard brain template.

Higher-level between-group contrasts were analyzed usingtwo-sample unpaired t-tests. We compared scan groupingswith predominantly deqi sensations versus scan groupingswith predominantly acute pain sensations (deqi>pain contrast)to identify clusters with a pre-selected threshold of Z>2.3 orZ<−2.3, and corrected cluster significance threshold ofP= 0.05 (Worsley et al., 1992). Clusters of voxels showingsignificant positive (Z>2.3) or negative (Z<−2.3) Z values forthe deqi>pain contrast, were superimposed onto the MNI152standard brain template. Whenever Z values for the deqi>paincontrast are reported for voxels, the corresponding Z valuesfor the deqi grouping alone and the pain grouping alone arealso stated.

For the significant deqi>pain clusters, probabilistic brainarea masks were utilized in a region of interest analysis. Theleft and right insula and cerebellummasks were selected fromthe MNI probabilistic structural atlas, the left and righthippocampus, amygdala, thalamus, and the posterior andanterior cingulate masks selected from the Harvard-Oxford

sub-cortical and cortical probabilistic atlases (FSLview 3.0.2,FSL software 4.1.2). Since fMRI data of individual participantswere transformed onto the MNI152 standard brain template,we opted to utilize relatively less conservative masks of brainstructures; probabilistic threshold ≥ 25% for all masks exceptfor the posterior and anterior cingulate gyri which was ≥5%.The selection of these relatively less conservative probabilisticmasks in the regions of interest analysis allows for moreinclusion of each structure and significant voxels despite thereduction in structure specificity. The number of voxels ineach of the brain regions of interest was counted, and thevoxel with the highest Z value located for each structure.

Acknowledgments

Acknowledgements are due to our 17 participants and thesupport staff at the York Neuroimaging Centre. HughMacPherson is supported by Career Scientist Award fromthe National Institute for Health Research.

R E F E R E N C E S

Beckmann, C.F., Jenkinson, M., Smith, S.M., 2003. Generalmultilevel linear modeling for group analysis in FMRI.NeuroImage 20, 1052–1063.

Biella, G., Sotgiu, M.L., Pellegata, G., Paulesu, E., Castiglioni, I.,Fazio, F., 2001. Acupuncture produces central activations inpain regions. NeuroImage 14, 60–66.

Birch, S., Felt, R., 1999. Understanding Acupuncture. ChurchillLivingstone, Edinburgh.

Bovey, M, 2006. De Qi. J. Chin. Med. 81, 18–29.Chiang, C.Y., Chang, C.T., Chu, H.L., Yang, I.L., 1973. Peripheral

afferent pathway for acupuncture analgesia. Sci. Sin. 16,210–217.

Davis, K.D., 2000. The neural circuitry of pain as explored withfunctional MRI. Neurol. Res. 22, 313–317.

Deadman, P., Al-Khafaji, M., 1998. A manual of acupuncture.J. Chin. Med. Publications.

Fang, J., Jin, Z., Wang, Y., Li, K., Kong, J., Nixon, E.E., Zeng, Y., Ren,Y., Tong, H., Wang, Y., Wang, P., Hui, K.K., 2009. The salientcharacteristics of the central effects of acupuncture needling:limbic–paralimbic–neocortical network modulation. Hum.Brain Mapp. 30, 1196–1206.

Fang, J.L., Krings, T., Weidemann, J., Meister, I.G., Thron, A., 2004.Functional MRI in healthy subjects during acupuncture:different effects of needle rotation in real and false acupoints.Neuroradiology 46, 359–362.

Han, J.S., 2004. Acupuncture and endorphins. Neurosci. Lett. 361(1-3), 258–261.

Hsieh, J.C., Tu, C.H., Chen, F.P., Chen, M.C., Yeh, T.C., Cheng, H.C.,Wu, Y.T., Liu, R.S., Ho, L.T., 2001. Activation of thehypothalamus characterizes the acupuncture stimulation atthe analgesic point in human: a positron emission tomographystudy. Neurosci. Lett. 307, 105–108.

Hui, K.K., Liu, J., Makris, N., Gollub, R.L., Chen, A.J., Moore, C.I.,Kennedy, D.N., Rosen, B.R., Kwong, K.K., 2000. Acupuncturemodulates the limbic system and subcortical gray structures ofthe human brain: evidence from fMRI studies in normalsubjects. Hum. Brain Mapp. 9, 13–25.

Hui, K.K., Liu, J., Marina, O., Napadow, V., Haselgrove, C., Kwong,K.K., Kennedy, D.N., Makris, N., 2005. The integrated responseof the human cerebro-cerebellar and limbic systems to

118 B R A I N R E S E A R C H 1 3 1 5 ( 2 0 1 0 ) 1 1 1 – 1 1 8

acupuncture stimulation at ST 36 as evidenced by fMRI.NeuroImage 27, 479–496.

Hui, K.K., Nixon, E.E., Vangel, M.G., Liu, J., Marina, O., Napadow, V.,Hodge, S.M., Rosen, B.R., Makris, N., Kennedy, D.N., 2007.Characterization of the “deqi” response in acupuncture. BMCComplement. Altern. Med. 7, 33.

Jenkinson, M., Smith, S., 2001. A global optimisation method forrobust affine registration of brain images. Med. Image Anal. 5,143–156.

Jenkinson, M., Bannister, P., Brady, M., Smith, S., 2002. Improvedoptimization for the robust and accurate linear registrationand motion correction of brain images. NeuroImage 17,825–841.

Kong, J., Fufa, D.T., Gerber, A.J., Rosman, I.S., Vangel, M.G., Gracely,R.H., Gollub, R.L., 2005. Psychophysical outcomes from arandomized pilot study of manual, electro, and sham acu-puncture treatment on experimentally induced thermal pain.J. Pain 6, 55–64.

Kong, J., Gollub, R., Huang, T., Polich, G., Napadow, V., Hui, K.,Vangel, M., Rosen, B., Kaptchuk, T.J., 2007. Acupuncture de qi,from qualitative history to quantitative measurement.J. Altern. Complement. Med. 13, 1059–1070.

Lewith, G.T., White, P.J., Pariente, J., 2005. Investigatingacupuncture using brain imaging techniques: the current stateof play. Evid. Based Complement. Altern. Med. 2, 315–319.

MacPherson, H., Asghar, A., 2006. Acupuncture needle sensationsassociated with de qi: a classification based on experts' ratings.J. Altern. Complement. Med. 12, 633–637.

MacPherson, H., Thomas, K., Walters, S., Fitter, M., 2001. Aprospective survey of adverse events and treatment reactionsfollowing 34,000 consultations with professionalacupuncturists. Acupunct. Med. 19, 93–102.

MacPherson, H., Green, G., Nevado, A., Lythgoe, M.F., Lewith, G.,Devlin, R., Haselfoot, R., Asghar, A.U., 2008. Brain imaging ofacupuncture: comparing superficial with deep needling.Neurosci. Lett. 434, 144–149.

Melzack, R., 1975. McGill pain questionnaire: major properties andscoring methods. Pain 1, 275–299.

Park, J., White, A., Stevinson, C., Ernst, E., James, M., 2002.Validating a new non-penetrating sham acupuncture device:two randomised controlled trials. Acupunct. Med. 20, 168–174.

Price, D.D., 2000. Psychological and neural mechanisms of theaffective dimension of pain. Science 288, 1769–1772.

Takeda,W.,Wessel, J., 1994. Acupuncture for the treatment of painof osteoarthritic knees. Arthritis Care Res. 7, 118–122.

Thomas, K.J., MacPherson, H., Thorpe, L., Brazier, J., Fitter, M.,Campbell, M.J., Roman, M., Walters, S.J., Nicholl, J., 2006.Randomised controlled trial of a short course of traditionalacupuncture compared with usual care for persistentnon-specific low back pain. BMJ 333, 623–626.

Vincent, C.A., Richardson, P.H., Black, J.J., Pither, C.E., 1989. Thesignificance of needle placement site in acupuncture.J. Psychosom. Res. 33, 489–496.

Wager, T.D., Rilling, J.K., Smith, E.E., Sokolik, A., Casey, K.L.,Davidson, R.J., Kosslyn, S.M., Rose, R.M., Cohen, J.D., 2004.Placebo-induced changes in FMRI in the anticipation andexperience of pain. Science 303, 1162–1167.

White, P., Bishop, F., Hardy, H., Abdollahian, S., White, A., Park, J.,Kaptchuk, T.J., Lewith, G.T., 2008. Southampton NeedleSensation Questionnaire: development and validation of ameasure to gauge acupuncture needle sensation. J. Altern.Complement. Med. 14, 373–379.

Woolrich, M.W., Ripley, B.D., Brady, M., Smith, S.M., 2001.Temporal autocorrelation in univariate linear modeling ofFMRI data. NeuroImage 14, 1370–1386.

Woolrich, M.W., Behrens, T.E., Smith, S.M., 2004. Constrainedlinear basis sets for HRF modelling using Variational Bayes.NeuroImage 21, 1748–1761.

Worsley, K.J., Evans, A.C., Marrett, S., Neelin, P., 1992.A three-dimensional statistical analysis for CBF activationstudies in human brain. J. Cereb. Blood FlowMetab. 12, 900–918.

Wu, M.T., Hsieh, J.C., Xiong, J., Yang, C.F., Pan, H.B., Chen, Y.C.,Tsai, G., Rosen, B.R., Kwong, K.K., 1999. Central nervouspathway for acupuncture stimulation: localization ofprocessing with functional MR imaging of thebrain—preliminary experience. Radiology 212, 133–141.

Yan, B., Li, K., Xu, J., Wang, W., Li, K., Liu, H., Shan, B., Tang, X.,2005. Acupoint-specific fMRI patterns in human brain.Neurosci. Lett. 383, 236–240.

Yoo, S.S., Teh, E.K., Blinder, R.A., Jolesz, F.A., 2004. Modulation ofcerebellar activities by acupuncture stimulation: evidencefrom fMRI study. NeuroImage 22, 932–940.