Technical advances in PET imaging not only provide tools for investigating brain metabolism, blood flow changes, and other non-selective markers of neural activity, but also whole brain determinants of specific receptor-binding distributions in fully conscious humans. Such progress enables us to indirectly assess neurotransmitter changes associated with placebo analgesia. For example, it allows us to indirectly measure the release of endogenous opioids in the brain^[1,2,3] .

       Most importantly, technology even allows us to collect fMRI and PET data simultaneously, which provides a new tool when investigating brain activity and neurotransmitters^[2]. In addition, there are other brain imaging tools such as high resolution EEG, and MEG, which can provide accurate temporal resolution information.

       The literature on brain imaging studies in acupuncture research is rich and extensive. In this paper, we will only focus on some key points critical to acupuncture research, i.e. acupoint specificity, dissociating acupuncture from placebo effects, and using imaging as a biomarker / measurement and predictor for acupuncture treatment.

1　fMRI study on acupoint specificity

       A salient feature of Traditional Chinese Medicine (TCM) is that specific acupuncture points are theorized to have salubrious effects on distant target organ systems. When fMRI studies on acupuncture were beginning, many investigators attempted to use fMRI (typically, a block or event related design) to test the existence (correction) of acupoint specificity^{[1,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22]}. One typical example of such research is the fMRI study of the vision points, acupoints that can be used to treat eye disorders.

       In an early study published in 1998, Cho and colleagues^[23] proposed a conceptual relationship whereby acupoint specificity was mediated via central neural networks that included corresponding brain regions. Using fMRI to test their model, Cho’s group reported that acupuncture manipulation at acupoints on the leg known to benefit visual system disorders (UB 67-UB 60) could produce specific fMRI signal changes within the occipital lobes of the brain, whereas sham non-acupoints (NAP) could not.

       Cho’s study received much attention and triggered a rush of experimental replications. Several research groups^{[9,10,13,24,25,26,27,28]} attempted to replicate this work in various ways. Although most drew similar conclusions, their results sometimes varied in fMRI signal change directionality. The work of Gareus et al^[10] was a notable exception, reporting in contrast to Cho et al. that acupuncture at a vision-related acupoint (GB 37) could neither directly produce activation in the visual cortex and associated areas, nor modulate fMRI signal changes in the visual cortex evoked by calibrated visual stimulation. In another study^[26], we found electroacupuncture stimulation at both vision related acupoints (UB60 and GB37) and the NAP produced modest, comparable fMRI signal decreases in the occipital cortex, including the bilateral cuneus, calcarine fissure, and surrounding areas. There is no significant difference among the three points UB 60, GB 37 and NAP. We speculate that cross modal inhibition, produced by needling-evoked somatosensory stimulation, may account for our finding of BOLD signal decreases in the occipital cortex. More recently, a research group from China^[27,28] using a different experimental paradigm, found that the brain’s neural response to acupuncture stimulation at GB37 and a nearby NAP was significantly different.

       Despite their continuous investigation of this topic, Cho and colleagues published a retraction of their earlier results^[23], reporting that, "there is no point specificity, at least for pain and analgesia effects, and that we no longer agree with results in our PNAS article" ^[29].

       Rethinking the concept of the vision point could prove interesting. Based on Traditional Chinese Acupuncture theory, the meridians that are connected to the eyes should all be usable in treating eye disorders. In this case, the more than one hundred acupoints on the urinary bladder, gallbladder, small intestine, heart and liver meridians could all be used to treat eye disorders.

       In addition, it is worth noting that GB 37 (Chinese translation Guangming, brightness) has only been used to treat eye disorders in the past century. A literature search did not find any evidence of GB 37 use for treating eye disorders in ancient classic acupuncture books, including The Spiritual Axis (Ling Su); Systematic Classic of Acupuncture (Zhen jiu jia yi jing, Jin Dynasty 265-420); The Illustrated Classic of Acupuncture Points as Found on the Bronze Model (Tong ren shu xue zhen jiu tu jing, Song Dynasty 960-1279); Classic of Nourishing Life with Acupuncture and Moxibustion (Zhen jiu zi sheng jing, Song Dynasty 960-1279); Gatherings from Eminent Acupuncturist (Zhen jiu ju ying, Ming Dynasty 1368-1644); and Great Compendium of Acupuncture and Moxibustion (Zhen jiu da cheng, Ming Dynasty 1368-1644). This lack of support may imply that using acupoint GB 37 to treat eye disorders is a modern concept that we speculate derived from the Chinese name of the acupoint, brightness.

       In addition to the development on the function of acupoints, the brain activity itself is very complex, and often difficult to control in experimental settings. For this reason, subtle alterations in the experimental paradigm including instruction to the subjects and their cooperation, as well as details of experimental procedures may all influence study findings. It is well known in the field of fMRI that the attention level, general arousal, mood of each individual, and deqi sensation evoked during an acupuncture treatment^[30,31] while scanning can each significantly affect the final results of a study.

       Accordingly, subtle differences between fMRI signal changes evoked by stimulation at acupoints and NAPs may be masked by psychological conditions such as attention. Taking the case of vision-related acupoints as an example, the subtle fMRI signal difference between vision-related and non vision-related acupoints / non-acupoints, if the differences exist, may be obscured by other physiological phenomena such as cross-modal interaction.

       In summary, it has been more than a decade since Cho and colleagues first attempted to investigate acupuncture point specificity using fMRI^[23]. However, validation of the phenomenon through the brain remains undetermined. This suggests that the field faces many challenges due to the complexities of acupuncture, the brain, and fMRI.

2　Powerful placebo effect in acupuncture research and choices of controls for the study

       Acupuncture, as defined on the website of the National Center for Complementary and Alternative Medicine, refers to a family of procedures involving the stimulation of specific points on the body using a variety of techniques (http://nccam.nih.gov/health/acupuncture?nav=gsa).

       Despite its wide application in Eastern countries for thousands of years, clinical trials evaluating the efficacy of acupuncture treatment have yielded rather contradictory results due to large placebo effects^{[32,33,34,35,36,37,38]}. In part, these findings reflect the complexity of identifying the "active ingredients" of acupuncture so as to enable design of an appropriate sham condition^[39] .

       Because acupuncture and placebo may work in part by activating self-healing / regulation mechanisms, it is not surprising that studies find the two share common pathways. Using pain as an example, studies show that both endogenous opioids and cholecystokinin are involved in acupuncture analgesia^[40,41] and placebo analgesia^[42,43]. Under the umbrella of self-healing, the distinction between the two could be blurred too far. A more useful construct holds that acupuncture, which is based on the stimulation of acupuncture points in the human body, should be regarded for "bottom-up" modulatory effects, and the placebo for "top-down" modulatory effects based on previous learning, expectation, therapeutic alliance and other factors common to all placebo treatments.

       Brain imaging has the potential to illuminate these contributory mechanisms. We found in previous studies that sham acupuncture combined with positive expectancy can produce a significant placebo analgesia effect^[44] and that sham acupuncture combined with negative expectancy can produce significant nocebo hyperalgesia^[45]. In addition, we also found that enhanced expectancy can significantly enhance the acupuncture analgesia effect whereas diminished positive expectancy appeared to inhibit acupuncture analgesia^[46]. Although placebo and acupuncture analgesia are comparable in behavioral measurements as indicated by subjective pain rating reduction, the two are associated with different brain networks. More specifically, verum acupuncture primarily involved lower signal intensity changes in brain regions associated with pain intensity signaling^[46,47]. In contrast, sham acupuncture involved lower activity in brain regions associated with pain-related cognitive-affective signaling. This finding is consistent with a recent study which reported that verum acupuncture, but not sham acupuncture, produced short-term increases in the binding potential of μ-opioid receptors (MOR) in multiple pain and sensory processing regions, and long-term increases in MOR binding potential in some of the same structures^[3] .

       It is worth noting that this large expectancy effect is not something unique to acupuncture. Studies found that expectancy can significantly modulate the analgesic effect of pharmacological drugs as well^[48]. In a previous study^[49], investigators explored how expectancy could change the analgesic efficacy of a potent opioid in healthy volunteers. Their results showed that positive treatment expectancy substantially enhanced the analgesic benefit of remifentanil and negative treatment expectancy abolished remifentanil analgesia.

       In parallel, in light of the contradictory nature of clinical trials on the efficacy of acupuncture treatment, it is important to advance the investigation of the efficacy of acupuncture and the mechanism by which it functions. Double-blinded randomized controlled trials (RCTs) serve as a gold standard when comparing the effect of treatment with the effect of an inert control. However, determining the proper inert control for an RCT designed to evaluate the efficacy of acupuncture is methodologically challenging due to: (1) Difficulty of mimicking both the visual appearance of the acupuncture treatment device and the method of needle insertion (2) Challenge of controlling for all nonspecific factors involved in acupuncture treatment including the ritual effect (3) Double-blinding the acupuncturist^[39].

       Placebo controls for acupuncture studies generally fall under one of two categories: (1) sham acupuncture, in which the skin is punctured with real acupuncture needles at nonacupoint locations, shallowly at acupoint locations, or both and (2) placebo acupuncture, which utilizes nonpenetrating stimuli. The latter includes the Streitberger device, Park device, Japanese double blinded device, the foam device and other non-penetrating devices such as toothpicks. Each of these devices has their pros and cons. Please see reference^[39] for a detailed description and comparison of these devices.

       Taken together, the applications of fMRI can significantly enhance our ability to distinguish between the neurological effects underlying verum and sham acupuncture effects. Choosing an appropriate control is crucial for acupuncture research as there is no single placebo method / device that can be universally applied for all acupuncture studies; the choice of an acupuncture control must therefore be determined by the specific aim of the study.

3　Use brain imaging measurements as an objective biomarker

       In medical practice, many patients complain of the highly subjective experience, frequently using phrases such as "I am not feeling well" and "I feel a lot of pain on my lower back" . Under most circumstances, the physician / medical practitioner can only make judgments based on the reports of the patients, and there is no objective measurement for these personal experiences.

       Using pain experience as an example, although there are more patients who visit clinicians for pain related disorders than almost any other illness, the most commonly used metric for pain measurements is the patient’s subjective report using the visual analogue scale between 0-100. As a result, there exists a demand in both clinical practice and research for identification of the neural correlates associated with pain. Since the subjective experience depends on the central processes of the brain, brain imaging such as fMRI has been applied toward developing a biomarker for pain.

       Over the course of the past few decades, investigators have identified a brain network underlying the experience of pain and pain intensity encoding. This network involves brain regions including the primary and secondary somotasensory cortices, insula, and dorsal anterior cingulate cortex^[50,51,52]. Wager and colleagues^[53] recently developed an fMRI-based measure for pain on an individual level. First, they used machine-learning analyses to identify a pattern of fMRI activity across regions associated with heat-induced pain. They then tested the sensitivity and specificity of the brain response pattern to painful stimuli versus nonpainful warm stimuli in a new sample. Finally, they assessed specificity relative to social pain and the responsiveness of the measure to the analgesic agent remifentanil. This study demonstrated the feasibility of using fMRI to assess pain elicited by noxious heat stimuli in healthy subjects.

       In addition to a task-related fMRI study, another recent application of brain imaging to the investigation of the neural correlates associated with pain is the study of resting state functional connectivity and brain structure^[54,55,56]. It is believed that low-frequency components of spontaneous fMRI signals during rest can provide information about the intrinsic functional and anatomical organization of the brain. With these tools, investigators have found significant brain functional connectivity and structural changes in chronic pain patients as compared to matched healthy controls^[54,55,56]. Previous studies also suggested that acupuncture can modulate the functional connectivity^{[57,58,59,60,61,62,63,64]} as well as cortical thickness (after longitudinal treatment)^[65] .

       In a more recent study^[65], we investigated brain cortical thickness and the functional connectivity changes after acupuncture treatment in knee osteoarthritis patients. We found that after longitudinal treatment, cortical thickness in the left posterior medial prefrontal cortex (pMPFC) decreased significantly in the sham group across treatment sessions as compared with the verum acupuncture group. Resting state functional connectivity analysis using the left pMPFC as a seed showed functional connectivity between the left pMPFC and the key regions of the descending pain modulatory system (rostral anterior cingulate cortex, and periaqueductal gray)^[66] are significantly enhanced after verum acupuncture when compared with the sham acupuncture group. Taken together, the fMRI/MRI measurement can be used as marker to evaluate the treatment effect and predict the development of disease / treatment^[67,68].

       In summary, brain imaging tools have been used extensively in acupuncture research. On the one hand, the application of these tools has enhanced our understanding of acupuncture’s mechanism, but on the other hand, we should be very cautious of our interpretation of the results due to the unique and complicated nature of acupuncture as a medical system, which includes many branches and schools. In particular, studies or clinical trials can only test a specific acupuncture treatment protocol for a specific population cohort. Over-interpretation of a positive or negative result is not scientifically appropriate and can also damage the development of acupuncture research in the long term. We believe future studies should focus on 1) translational research that can utilize the findings from brain imaging studies to enhance acupuncture treatment effects, 2) elucidating the association between brain activity/connectivity/structural changes and the clinical outcomes, and 3) examining the causal relationship between the brain and behavioral effects.