Available online at www.sciencedirect.

com

Neuroscience Letters 434 (2008) 144–149

Brain imaging of acupuncture: Comparing

superficial with deep needling
Hugh MacPherson a,∗ , Gary Green b , Angel Nevado b , Mark F. Lythgoe c ,
George Lewith d , Ross Devlin b , Robyn Haselfoot b , Aziz U.R. Asghar b,e
aDepartment of Health Sciences, University of York, YO10 5DD, United Kingdom
b York Neuroimaging Centre, University of York, Y10 5DG, United Kingdom
c RCS Unit of Biophysics, UCL Institute of Child Health, University College London, WC1N 3JH, United Kingdom
d Complementary Medicine Research Unit, University of Southampton, SO17 1BJ, United Kingdom
e Hull York Medical School and Department of Biological Sciences, University of Hull, Hull HU6 7RX , United Kingdom

Received 5 November 2007; received in revised form 16 January 2008; accepted 22 January 2008

Abstract
The difference between superficial and deep needling at acupuncture points has yet to be mapped with functional magnetic resonance imaging
(fMRI). Using a 3 T MRI, echo planar imaging data were acquired for 17 right-handed healthy volunteer participants. Two fMRI scans of acupuncture
needling were taken in random order in a block design, one for superficial and one for deep needling on the right hand at the acupuncture point
LI-4 (Hegu), with the participant blind to the order. For both scans needle stimulation was used. Brain image analysis tools were used to explore
within-group and between-group differences in the blood oxygen level dependent (BOLD) responses. The study demonstrated marked similarities
in BOLD signal responses between superficial and deep needling, with no significant differences in either activations (increases in BOLD signal) or
deactivations (decreases in BOLD signal) above the voxel Z score of 2.3 with corrected cluster significance of P = 0.05. For both types of needling,
deactivations predominated over activations. These fMRI data suggest that acupuncture needle stimulation at two different depths of needling,
superficial and deep, do not elicit significantly different BOLD responses. This data is consistent with the equivalent therapeutic outcomes that are
claimed by proponents of Japanese and Chinese styles of acupuncture that utilise superficial and deep needling, respectively.
© 2008 Elsevier Ireland Ltd. All rights reserved.

Keywords: Acupuncture; Neuroimaging; fMRI

Acupuncture is a treatment modality that is growing in pop-
ularity [5,23] and is practised in many different ways, with
one of the fundamental differences in approach being the depth
that acupuncture needles are inserted. Practitioners trained in
Japanese styles of acupuncture commonly insert their nee-
dles superficially just below the skin to a depth of perhaps
1–2 mm [3]. In contrast, practitioners trained in Chinese styles
of acupuncture, including what has become known as Tradi-
tional Chinese Medicine (TCM), are of the view that therapeutic
responses are associated with needling depths that are often
called “deep”, usually 1–2 cm depending on the local mus-
culature [4]. Practitioners of both acupuncture styles claim
therapeutic benefits while insisting that the depth of needling
∗ Corresponding author at: Department of Health Sciences, Area 3, Seebohm
Rowntree Building, University of York, Heslington, York YO10 5DD, United
Kingdom. Tel.: +44 1904 321394; fax: +44 1904 321388.
E-mail address: hm18@york.ac.uk (H. MacPherson).
is integral to the putative therapeutic gain. A primary interest in
this area therefore is whether these different depths of needling
elicit similar or different physiological responses.
The primary aim of this investigation therefore was to utilize
fMRI to measure the increases (activations) and decreases
(deactivations) in the blood oxygen level dependent (BOLD)
response with acupuncture needling at LI-4, comparing the
impact of needling at two different depths: (1) superficial
needling as commonly used in Japanese styles of acupuncture
(3), and (2) deep needling to a depth commonly used in
Chinese-based styles of acupuncture (4).
We recruited 17 healthy right-handed adult volunteer partic-
ipants who were naı̈ve to acupuncture: eight males (mean age
33 years, range 18–46, S.D. = 11.9) and nine females (mean age
39 years, range 20–54, S.D. = 10.2), with an average age of 36
years. Participants were provided with an information leaflet
describing the study and a safety questionnaire that screened

0304-3940/$ – see front matter © 2008 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.neulet.2008.01.058
 H. MacPherson et al. / Neuroscience Letters 434 (2008) 144–149
for contraindications for MRI scanning. When the York Neu-
roimaging Centre received a completed safety questionnaire,
participants were screened for eligibility and informed consent
was obtained. Approval for the study was obtained from the York
Neuroimaging Centre’s Science and Ethics Committees, and all
ethical requirements were met.
For all participants, and for all scans, acupuncture needling
was performed at the acupuncture point Hegu (LI-4) on the
right hand by an experienced acupuncturist who is a member
of the British Acupuncture Council (HM). Needles were ster-
ile, disposable, non-magnetic stainless steel, 25 mm long, and
0.28 mm diameter, manufactured by Hwato, China. The inser-
tion depth of the needle into the interosseus dorsalis muscle was
randomised to be either superficial (1–2 mm) or deep (8–12 mm).
The depth for deep needling was adjusted by the acupuncturist to
take into account the variability of the participant’s interosseus
dorsalis muscle, as would be judged appropriate in normal prac-
tice. The randomisation was computer generated with allocation
concealed in envelopes that were opened at the onset of the scans.
fMRI scans were conducted over two 16 min periods; one
scan involved superficial needling only and another separate
scan involved deep needling only. Participants were informed
that they would be receiving two types of acupuncture during
their time in the scanner. To minimise participants’ expectations
about which depth of needling they were about to receive, they
were not informed of the order of the two scans. They were
requested to lie still during scans with their eyes closed.
Each fMRI scan involved two identical active blocks of nee-
dle stimulation separated by a rest period (Fig. 1). Although
fMRI studies typically contain more stimulation blocks, previ-
ous published research has shown that this experimental design
paradigm using two long stimulation blocks is robust in detecting
changes in the BOLD response (both activations and deacti-
vations) to acupuncture needling at LI-4 [11]. For the active
blocks, whether superficial or deep insertion, the needle was
“stimulated” with a continuous alternating 180 degree rotation
of the needle clockwise and anti-clockwise at approximately
two cycles per second. In acupuncture theory, this manipulation
technique is also known as needle stimulation using the “bal-
anced” or “even” method [4]. For the baseline, the needle was
retained with no stimulation.
After each scan, participants reported experiential data on
needle sensation using the Park questionnaire [18], leading to
Fig. 1. Block design showing the two needle stimulation periods which were
provided consecutively within each scan, with both stimulation periods involving
identical needling. For each participant two fMRI scans were performed using
this block design, one with two consecutive superficial needling stimulation
periods and one with two consecutive deep needling stimulation periods. The
order of these two types of needling intervention was randomised.
145
ratings of both pain and deqi [16]. Immediately after each 16 min
scan, participants were asked questions about their experience
of 25 individual needle sensations from the Park questionnaire
[18]. Participants rated the intensity of each sensation on an
ordinal scale: “none” (zero), “slight” (one), “moderate” (two) or
“strong” (three). In a previous study, these 25 sensations from
the Park scale were separated using a hierarchical cluster analy-
sis into two categories: deqi being associated with aching, dull,
heavy, numb, radiating, spreading and tingling sensations and
acute pain being associated with burning, hot, hurting, pinch-
ing, pricking, sharp, shocking, stinging and tender sensations
[16]. Deqi and acute pain sensation scores for each scan were
determined from the participant’s 25 sensation scores (range 0–3
each sensation, divided by number of sensations reported). The
impact of prior expectation/anticipation was equalised across
superficial and deep needling as the participant was blind to the
intervention.
Scanning was performed with a 3 T magnetic resonance
imaging system (GE Signa HD Excite) with an eight chan-
nel head coil (GE Signa Excite 3.0T, High Resolution Brain
Array, MRI Devices Corp., Gainesville FL). Axial images were
acquired for functional and structural scans which covered the
whole brain. Localiser scans were used to align axial images
along the anterior commissure-posterior commissure line, cov-
ering both hemispheres.
For functional imaging, Echo Planar Imaging (EPI) images
were acquired using a T2* weighted gradient echo sequence
(TR 4 s, TE 30 ms, flip angle 90◦ , acquisition matrix 128 × 128,
FOV 240 mm, in-plane resolution 2 mm × 2 mm, contiguous
slice thickness 4.5 mm).
High resolution T1-weighted structural images were acquired
prior to the functional scans using an IR (Inversion Recovery)-
prepared 3D-FSPGR (Fast Spoiled Gradient Echo) pulse
sequence (TR 7.5 s, TE 3 ms, flip angle 20◦ , acquisition matrix
256 × 224 interpolated to 512 × 512, FOV 260 mm, in plane
resolution 0.5 mm × 0.5 mm, slice thickness 2.6 mm with an
overlap of 1.3 mm).
The primary outcomes evaluated were activations and
deactivations as represented by increases and decreases in
BOLD response. FSL software (FMRIB’s Software Library,
www.fmrib.ox.ac.uk/fsl) based on generalised linear modelling
was used. The design matrix consisted of a single explanatory
variable which was “on” during the periods of stimulation and
otherwise “off” (see Fig. 1). Positive and negative parameter esti-
mates, referred to as activations and deactivations, respectively,
were calculated for each of the two conditions (superficial and
deep needling) in each participant.
For each participant, a first level analysis of individual scans
was carried out using FEAT (FMRI Expert Analysis Tool)
Version 5.4, part of FSL. The following pre-statistics process-
ing was applied; slice-timing correction using Fourier-space
time-series phase-shifting; motion correction using MCFLIRT
[13]; non-brain removal using BET [21]; spatial smoothing
using a Gaussian kernel of FWHM 5mm; mean-based inten-
sity normalisation of all volumes by the same factor; highpass
temporal filtering (Gaussian-weighted LSF straight line fitting,
with sigma = 480 s). Time-series statistical analysis was car-
146 H. MacPherson et al. / Neuroscience Letters 434 (2008) 144–149
ried out using FILM (FMRIB’s Improved Linear Model) with
local autocorrelation correction [27]. Both positive (activations)
and negative (deactivations) interactions were modelled. Reg-
istration of EPI images to high resolution T1 images and then
subsequently to the Montreal Neurological Institute standard
brain (MNI152) was carried out using FLIRT [13,14].
Motion correction was applied to the fMRI scans and
for each participant the mean (across voxels) voxel absolute
(each time point with respect to the reference image) and
relative (each time point with respect to the previous time
point) was calculated by the FSL software (Motion Correc-
tion Report). Displacements across 34 separate scans from
17 participants were found to be small: mean ± S.D. of the
absolute displacement = 0.48 ± 0.28 mm, and relative displace-
ment = 0.09 ± 0.04 mm.
Within-group analysis of scans were performed separately
for each of the two depths of needling using FLAME (FMRIB’s
Local Analysis of Mixed Effects) [1,26]. Z (Gaussianised T/F)
statistic within-group images were identified using clusters
determined by individual voxels above the threshold of Z > 4.3
and a (corrected) voxel cluster significance threshold of P = 0.05
[28].
Between-group differences were analysed by subtracting the
scans for superficial needling from those of deep needling in
a two sample paired t-test using FSL software. We identified
clusters with individual voxels having Z > 2.3, and (corrected)
voxel cluster significance threshold of P = 0.05 [28] For both
within-group and between-group analyses, significant clusters
(activations or deactivations) were registered and superimposed
onto the MNI152 standard brain template. Clusters sizes of
≤20 voxels were not considered further.
The primary comparison in our study was between nee-
dle stimulation at superficial compared to deep needling. This
between-group analysis, in which superficial needling responses
were subtracted from deep needling responses (paired t-tests),
showed no significant increases (activations) or decreases (deac-
tivations) in BOLD response above the threshold of Z = 2.3
with a corrected cluster significance threshold of P = 0.05
(Table 1A and 1D, Fig. 1).
The participant-reported sensory data showed levels of deqi
and pain to be of the same order for both superficial and
deep needling. The mean deqi score was 0.50 (S.D. = 0.48)
per sensation for superficial needling and slightly higher at
0.71 (S.D. = 0.49) for deep needling, while the acute pain
score was 0.44 (S.D. = 0.33) for superficial needling and again
slightly higher at 0.62 (S.D. = 0.57) for deep needling. There
were no significant differences between superficial and deep
scans for deqi scores (P = 0.08) or for acute pain scores
(P = 0.17).
The participant-reported sensory data showed levels of deqi
and pain to be of the same order for both superficial and
deep needling. The mean ± standard error of the deqi score
was 0.50 ± 0.12 per sensation for superficial needling and
at 0.71 ± 0.12 for deep needling, while the acute pain score
was 0.44 ± 0.08 for superficial needling and 0.62 ± 0.14 for
deep needling. There differences between superficial and deep
needling did not reach statistical significance (paired t test) for
deqi scores (P = 0.08) or for acute pain scores (P = 0.17).
From the within-group analyses of deep needling scans
(considered as a single group), we observed a pattern of acti-
vations above the Z = 4.3 threshold (with a corrected cluster
significance of P = 0.05) predominantly in the culmen of the
cerebellum (Table 1B). Deactivations were observed in the
occipital lobe (BA19), the frontal lobe and the medial frontal
gyrus (BA10) (Table 1E). From the within-group analyses of
superficial needling scans (considered as a single group), the
activations were similar to those of deep needling, predomi-
nantly in the culmen of the cerebellum (Table 1C). Superficial
needling deactivations were primarily located in the precuneus
(BA39), the precentral gyrus (BA4), and the inferior tempo-
ral gyrus (Table 1F). For both depths of needling, deactivations
were predominantly in the left hemisphere (contralateral to
the needling side). Despite these apparent within-group differ-
ences in patterns of deactivation, the statistical analysis based
on between-group subtraction of superficial from deep needling
scans (see above) showed no significant differences (i.e. no clus-
ters with Z > 2.3 and a corrected cluster significance of P = 0.05).
Based on the sizes of significant clusters, overall the deac-
tivations were substantially larger than activations for both
superficial and deep needling. For example with deep needling,
the largest area of deactivation involved a cluster with 1496 vox-
els, compared to a cluster of only 62 voxels for the largest area
of activation. For superficial needling, the largest deactivation
cluster contained 4209 voxels compared to 141 for activation.
A region of interest analysis in the secondary somatosensory
area demonstrated bilateral activation with both types of
needling (superficial needling:- right 60, −22, 16, Z = 3.17 and
left −68, −24, 20, Z = 3.26; deep needling:- right coordinates x =
56, y = −26, z = 22, Z = 5.36 and left −68, −22,16, Z = 3.36, un-
corrected P = 0.05) thus confirming the activations in this area
as reported previously in fMRI studies of acupuncture [11,31].
The main finding in this study is that, when subtracting the
BOLD responses between needle stimulation at superficial and
deep needling, no significant differences are seen in activa-
tions (increase in BOLD response) or deactivations (decrease
in BOLD response) at the threshold of Z > 2.3 and corrected
cluster significance level of P = 0.05. The data on needle sen-
sation, whether associated with deqi or acute pain, indicated
that sensation was not a confounder. The percentage BOLD
response is of a similar order to that of other studies [11,12].
Previous research has primarily explored the impact of deep
needling, and we know of no fMRI study directly comparing
needle stimulation at superficial versus deep needling. Previ-
ous fMRI studies of deep needling have reported results similar
to our data for deep needling, with deactivations predominat-
ing, whether the same point at LI-4 has been used or at other
acupuncture point locations [11,12,15,17,29,30]. Similar data on
activations as a result of deep needling have also been reported
[31,12,15,17,29,30]. Taken together, our study replicates the
core findings of other studies on deep needling, which are useful
in facilitating interpretations of the comparison between super-
ficial and deep needling that we report here.
Table 1
Brain areas showing activation and deactivation with deep and superficial needling
Voxels with maximum effect Cluster

x y z P value (−log10 P) Z value % BOLD ± S.D. Number of voxels Mean Z value

H. MacPherson et al. / Neuroscience Letters 434 (2008) 144–149

Activation
A. Deep > Superficial (paired t test) (this comparison is No significant differences in activations between deep and superficial groups
equivalent to superficial > deep with respect to
deactivation)
B. Deep needling (N = 17)
Cerebellum (culmen) I 4 −48 −24 3.8 5.6 0.43 ± 0.07 62 4.9
C. Superficial needling (N = 17)
Cerebellum (culmen) I 10 −44 −28 7.5 6.8 0.36 ± 0.05 141 5.2
Cerebellum (culmen) I −30 −34 −26 3.2 6.6 0.49 ± 0.07 36 5.0
Deactivation
D. Deep > Superficial (paired t test) (this comparison is No significant differences in deactivations between deep and superficial groups
equivalent to superficial > deep with respect to activation)
E. Deep needling (N = 17)
Occipital lobe (BA19) I −44 −80 2 >20 7.6 −1.02 ± 0.13 1469 5.2
Frontal lobe I −28 0 44 5.6 6.2 −0.27 ± 0.03 113 5.0
Medial frontal gyrus (BA10) I −12 34 −10 4.1 6.1 −0.23 ± 0.04 69 4.9
F. Superficial needling (N = 17)
Precuneus (BA39) I −40 −68 34 >20 7.5 −0.52 ± 0.07 4209 5.2
Precentral gyrus (BA4) I −56 −14 30 5.8 6.7 −0.36 ± 0.05 93 5.3
Inferior temporal gyrus I −56 −56 −16 4.9 6.7 −0.48 ± 0.07 71 4.9

All Z value thresholding is >4.3 except for paired differences which are thresholded at Z > 2.3. The most significant clusters (maximum of three shown) with a (corrected) cluster significance threshold of P = 0.05
are presented along with the voxel having the highest Z score within each cluster, x, y, z are MNI152 brain coordinates (mm). L: left; R: right. BA: Brodmann area.

147
148 H. MacPherson et al. / Neuroscience Letters 434 (2008) 144–149

Fig. 2. Functional maps for cluster analysis showing significant clusters for activations and deactivations on (A) deep needling and (B) superficial needling of LI-4.
Functional data were thresholded at Z > 4.3 with a corrected cluster significance level of P = 0.05. Clusters have been overlaid onto a standard brain (MNI152) shown
in orthographic view. The crosshairs in each panel show the position of the voxel with the highest Z score in the cluster with the highest statistical significance. The
number at the bottom right corner of the axial images gives the z direction coordinate of the selected MNI152 standard brain slice. R: right brain hemisphere; L: left
brain hemisphere.

In this study, deactivations elicited by acupuncture were
more marked than the activations, a finding that is consistent
with other neuroimaging studies of acupuncture [11,12]. The
predominance of deactivations has been speculatively linked
to acupuncture’s putative therapeutic role [11]. It is not clear
whether such deactivations might involve a reduction of exci-
tatory inputs or a reduction of inhibitory inputs or some
combination of both [7]. An argument has been made that neg-
ative BOLD responses can be explained by the concept of a
“vascular steal” from nearby areas where there are activations
[20]. However, we do not observe any obvious pattern of BOLD
responses which indicate that areas of activation are coupled
with adjacent areas of deactivation. An alternative interpreta-
tion is that the brain can have a resting state that is a “default
mode”, which involves a baseline level of sustained brain activ-
ity that is inhibited during tasks [8,19]. Building on these studies,
further research is needed to establish the meaning and role of
the extensive deactivations associated with acupuncture that we
observed in this study for both depths of needling.
In this study, it was noted that some of the activations and
deactivations were located to the white matter of the brain
(Fig. 2). One explanation for this is that large motion arte-
facts may have been present in the fMRI scans. However,
this is unlikely as movement displacements were well within
acceptable limits. Another possibility is that registration of the
fMRI images from each participant onto a common standard
brain causes activations/deactivations to be shifted to the white
matter. It should be noted that many fMRI studies, including
those related to acupuncture [12], observe white matter activa-
tions/deactivation.
The current investigation only examined and analysed fMRI
responses to the needling at different depths, and these only
within the relatively short-term time span of a block design. It
could be argued that possible differences between the two depths
of needling may be too subtle for detection in fMRI, or that
the differences might only emerge over time when the longer-
term impact of acupuncture is being investigated. Research into
endorphin release, for example, indicates that acupuncture’s
physiological effects can continue for many hours [9] and clin-
ical benefits might be sustained over long periods [24]. More
accurate modelling of clinical effects may need to involve six
or more acupuncture sessions [6] Also, the participants in this
study were healthy individuals and it is possible that superficial
and deep acupuncture could potentially have different effects
when used to treat people with pathology.
There is a question as to whether the BOLD responses that
we have observed in this study were a response to the experience
of pain rather than to the acupuncture. Painful stimuli are known
to involve processing in the somatosensory cortex, the insula
and the anterior cingulate cortex [25]. In addition, experimental
research has generally shown that pain is primarily associated
with activations rather than deactivations [2,10]. Our results,
with patterns of BOLD response affecting different areas than
commonly seen in pain research, is a finding supported by the
data of Hui and colleagues [11,12]. Like these investigators, we
also observed a predominance of deactivations over activations.
 H. MacPherson et al. / Neuroscience Letters 434 (2008) 144–149
These findings taken together suggest that the responses to
acupuncture are different from those to be expected from pain
alone.
This evidence is consistent with the equivalent therapeutic
outcomes that are claimed by proponents of both Japanese
(superficial needling) and Chinese styles of acupuncture (deep
needling). Further research in the brain imaging of acupuncture
is needed to explore the response to needle stimulation at other
acupuncture points as well as non-acupoints, to compare the
stimulation of the needle versus no stimulation, to evaluate
responses of patients receiving acupuncture for specific condi-
tions and identify the impact of non-penetrating sham needles
[22]. To enhance clinical relevance, there is also a need to
correlate changes in brain images with changes in outcome for
specific symptoms.
Our fMRI experimental results indicate that superficial and
deep acupuncture needling are associated with imaging patterns
that have no significant differences. This evidence is consistent
with the equivalent therapeutic outcomes that are claimed by
proponents of Japanese and Chinese styles of acupuncture that
utilise superficial and deep needling, respectively.
Acknowledgements
Acknowledgements are due to the 17 participants. The assis-
tance from Iain Lamb and the various support personnel at the
York Neuroimaging Centre is gratefully acknowledged. Others
who have provided insight and support for this study include
Professor Trevor Sheldon, Professor David Torgerson and Anne
Burton, University of York, and Professor Robert Turner, Uni-
versity College London.
References
[1] C.F. Beckmann, M. Jenkinson, S.M. Smith, General multilevel linear mod-
eling for group analysis in FMRI, Neuroimage 20 (2) (2003) 1052–1063.
[2] U. Bingel, M. Quante, R. Knab, B. Bromm, C. Weiller, C. Buchel, Sub-
cortical structures involved in pain processing: evidence from single-trial
fMRI, Pain 99 (1/2) (2002) 313–321.
[3] S. Birch, R. Felt, Understanding Acupuncture, Churchill Livingstone, Edin-
burgh, 1999.
[4] P. Deadman, M. Al-Khafaji, A manual of acupuncture, J. Chin. Med. Publ.
(1998).
[5] D.M. Eisenberg, R.B. Davis, S.L. Ettner, S. Appel, S. Wilkey, M. Van
Rompay, R.C. Kessler, Trends in alternative medicine use in the United
States, 1990–1997: results of a follow-up national survey, JAMA 280 (18)
(1998) 1569–1575.
[6] J. Ezzo, B. Berman, V.A. Hadhazy, A.R. Jadad, L. Lao, B.B. Singh, Is
acupuncture effective for the treatment of chronic pain? A systematic
review, Pain 86 (3) (2000) 217–225.
[7] P.H. Ghatan, J.C. Hsieh, K.M. Petersson, S. Stone-Elander, M. Ingvar,
Coexistence of attention-based facilitation and inhibition in the human
cortex, Neuroimage 7 (1) (1998) 23–29.
[8] D.A. Gusnard, M.E. Raichle, M.E. Raichle, Searching for a baseline: func-
tional imaging and the resting human brain, Nat. Rev. Neurosci. 2 (10)
(2001 Oct) 685–694.
[9] J.S. Han, Acupuncture and endorphins, Neurosci. Lett. 361 (1–3) (2004)
258–261.
[10] C. Helmchen, C. Mohr, C. Erdmann, D. Petersen, M.F. Nitschke, Differen-
tial cerebellar activation related to perceived pain intensity during noxious
thermal stimulation in humans: a functional magnetic resonance imaging
study, Neurosci. Lett. 335 (3) (2003) 202–206.
149
[11] K.K. Hui, J. Liu, N. Makris, R.L. Gollub, A.J. Chen, C.I. Moore, D.N.
Kennedy, B.R. Rosen, K.K. Kwong, Acupuncture modulates the limbic
system and subcortical gray structures of the human brain: evidence from
fMRI studies in normal subjects, Hum. Brain Mapp. 9 (1) (2000) 13–25.
[12] K.K. Hui, J. Liu, O. Marina, V. Napadow, C. Haselgrove, K.K. Kwong,
D.N. Kennedy, N. Makris, The integrated response of the human cerebro-
cerebellar and limbic systems to acupuncture stimulation at ST 36 as
evidenced by fMRI, Neuroimage 27 (3) (2005) 479–496.
[13] M. Jenkinson, P. Bannister, M. Brady, S. Smith, Improved optimization for
the robust and accurate linear registration and motion correction of brain
images, Neuroimage 17 (2) (2002) 825–841.
[14] M. Jenkinson, S. Smith, A global optimisation method for robust affine
registration of brain images, Med. Image Anal. 5 (2) (2001) 143–156.
[15] J. Kong, L. Ma, R.L. Gollub, J. Wei, X. Yang, D. Li, X. Weng, F. Jia,
C. Wang, F. Li, R. Li, D. Zhuang, A pilot study of functional magnetic
resonance imaging of the brain during manual and electroacupuncture
stimulation of acupuncture point (LI-4 Hegu) in normal subjects reveals
differential brain activation between methods, J. Altern. Complement Med.
8 (4) (2002) 411–419.
[16] H. MacPherson, A. Asghar, Acupuncture needle sensations associated with
de qi: a classification based on experts’ ratings, J. Altern. Complement Med.
12 (7) (2006 Sep) 633–637.
[17] V. Napadow, N. Makris, J. Liu, N.W. Kettner, K.K. Kwong, K.K. Hui,
Effects of electroacupuncture versus manual acupuncture on the human
brain as measured by fMRI, Hum. Brain Mapp. 24 (3) (2005) 193–
205.
[18] J. Park, A. White, C. Stevinson, E. Ernst, M. James, Validating a new non-
penetrating sham acupuncture device: two randomised controlled trials,
Acupunct. Med. 20 (4) (2002) 168–174.
[19] M.E. Raichle, A.M. MacLeod, A.Z. Snyder, W.J. Powers, D.A. Gusnard,
G.L. Shulman, A default mode of brain function, Proc. Natl. Acad. Sci.
U.S.A. 98 (2) (2001 Jan 16) 676–682.
[20] A. Shmuel, E. Yacoub, J. Pfeuffer, P.F. Van de Moortele, G. Adriany, X. Hu,
K. Ugurbil, Sustained negative BOLD, blood flow and oxygen consumption
response and its coupling to the positive response in the human brain,
Neuron 36 (6) (2002) 1195–1210.
[21] S.M. Smith, Fast robust automated brain extraction, Hum. Brain Mapp. 17
(3) (2002) 143–155.
[22] K. Streitberger, J. Kleinhenz, Introducing a placebo needle into acupuncture
research, Lancet 352 (9125) (1998) 364–365.
[23] K.J. Thomas, P. Coleman, J.P. Nicholl, Trends in access to complementary
or alternative medicines via primary care in England: 1995–2001 results
from a follow-up national survey, Fam. Pract. 20 (5) (2003) 575–577.
[24] K.J. Thomas, H. MacPherson, L. Thorpe, J. Brazier, M. Fitter, M.J. Camp-
bell, M. Roman, S.J. Walters, J. Nicholl, Randomised controlled trial of
a short course of traditional acupuncture compared with usual care for
persistent non-specific low back pain, BMJ 333 (7569) (2006) 623–626.
[25] R.D. Treede, D.R. Kenshalo, R.H. Gracely, A.K. Jones, The cortical rep-
resentation of pain, Pain 79 (23) (1999) 105–111.
[26] M.W. Woolrich, T.E. Behrens, S.M. Smith, Constrained linear basis sets
for HRF modelling using Variational Bayes, Neuroimage 21 (4) (2004)
1748–1761.
[27] M.W. Woolrich, B.D. Ripley, M. Brady, S.M. Smith, Temporal autocor-
relation in univariate linear modeling of FMRI data, Neuroimage 14 (6)
(2001) 1370–1386.
[28] K.J. Worsley, A.C. Evans, S. Marrett, P. Neelin, A three-dimensional sta-
tistical analysis for CBF activation studies in human brain, J. Cereb. Blood
Flow Metab. 12 (6) (1992) 900–918.
[29] M.T. Wu, J.C. Hsieh, J. Xiong, C.F. Yang, H.B. Pan, Y.C. Chen, G. Tsai,
B.R. Rosen, K.K. Kwong, Central nervous pathway for acupuncture stim-
ulation: localization of processing with functional MR imaging of the
brain-preliminary experience, Radiology 212 (1) (1999) 133–141.
[30] B. Yan, K. Li, J. Xu, W. Wang, K. Li, H. Liu, B. Shan, X. Tang, Acupoint-
specific fMRI patterns in human brain, Neurosci. Lett. 383 (3) (2005)
236–240.
[31] S.S. Yoo, E.K. Teh, R.A. Blinder, F.A. Jolesz, Modulation of cerebellar
activities by acupuncture stimulation: evidence from fMRI study, Neu-
roimage 22 (2) (2004) 932–940.