A Narrative Review of Neuroimaging Studies in Acupuncture for Migraine

Acupuncture has been widely used as an alternative and complementary therapy for migraine. With the development of neuroimaging techniques, the central mechanism of acupuncture for migraine has gained increasing attention. This review aimed to analyze the study design and main findings of neuroimaging studies of acupuncture for migraine to provide the reference for future research. The original studies were collected and screened in three English databases (PubMed, Embase, and Cochrane Library) and four Chinese databases (Chinese National Knowledge Infrastructure, Chinese Biomedical Literature database, the Chongqing VIP database, and Wanfang database). As a result, a total of 28 articles were included. Functional magnetic resonance imaging was the most used neuroimaging technique to explore the cerebral activities of acupuncture for migraine. This review manifested that acupuncture could elicit cerebral responses on patients with migraine, different from sham acupuncture. The results indicated that the pain systems, including the medial pain pathway, lateral pain pathway, and descending pain modulatory system, participated in the modulation of the cerebral activities of migraine by acupuncture.


Introduction
Migraine is a chronic neurological disease characterized by recurrent headaches and neurological symptoms [1]. As the third most prevalent disorder and the seventh highest specific cause of disability worldwide [2,3], migraine not only affects the quality of life (QoL) and productivity of patients [4] but also results in a high financial burden to society [5]. Recently, acupuncture has been widely used as an alternative or complementary treatment for migraine with its efficacy and safety. Using the neuroimaging techniques to explore the underlying mechanism of acupuncture for migraine has also attracted the increasing interest of investigators.
Since the first neuroimaging study about acupuncture for migraine was published in 2008, nearly 30 studies have been emerged to improve the understanding of acupuncture for migraine with real-time visualized evidence. However, the designs and results of these studies showed significant differences. For example, some research studies concentrated on study designs from different perspectives, such as different manipulation modalities of acupuncture and different subtypes of migraine; several studies applied different neuroimaging scanning techniques or used different analytical methods. e various study designs may partly contribute to different results which affect the clinical application to some degree. erefore, this review aimed to analyze the study design and summarize the main results of the published neuroimaging studies in acupuncture for migraine to deepen the understanding of the cerebral mechanism of acupuncture for migraine and provide reference and guidance for future research.

Searching Strategy.
Data collections were conducted in three English language databases (PubMed, Embase, and Cochrane Library) and four Chinese language databases (Chinese National Knowledge Infrastructure, CNKI; Chinese Biomedical Literature database, CBM; the Chongqing VIP database, VIP; and the Wanfang database, WF) from database inception up to 31st Dec 2020, with the combination of the main keywords: neuroimaging, acupuncture, and pain; studies about migraine were selected. e retrospective searching was performed on reference lists of included articles. Details of search terms were depicted and modified for each database in Supplementary Table 1. e article was included if [1] the studies were original articles, [2] the patients were diagnosed with migraine, [3] acupuncture was the intervention method, and [4] neuroimaging techniques were used to investigate the cerebral changes elicited by acupuncture (Supplementary Table 1). e article was excluded if [1] it was a duplicate article or [2] it was the irrelevant article, review, case report, protocol, only abstract, editorial letter, retracted article, and animal experiment. Studies not fulfilling each of the above criteria were excluded (Supplementary Table 1).
Supplementary Figure 1 shows the flow diagram of the literature search and screening process.

Data Extraction and Analysis.
e published year, the institution of corresponding authors (Supplementary Table 2 and Supplementary Figure 2), study design (types of migraine, interventions (manipulation modality and course of treatments), types of control, and clinical scales of migraine) (Supplementary Table 3), and the related neuroimaging information (neuroimaging scanning techniques, analytical methods of neuroimaging data, and the results of brain regions involved in acupuncture analgesia for migraine) (Supplementary Table 4) were extracted.

Results
Twenty-eight original articles were included in this review.

Discussion
Since neuroimaging techniques were used to explore the mechanisms of acupuncture, the central mechanism of acupuncture for migraine has gradually drawn the attention of investigators. Since the first study explored the central mechanism of acupuncture for migraine, 28 studies have been published during the past 15 years. From the beginning, these articles simply concentrated on the self-control design of the cerebral activities, gradually developed to explore the different cerebral activities of the acupuncture efficacy (such as VA and SA), and now paid attention to the different cerebral activities of acupuncture for subtypes or stages of the migraine. ese articles significantly improved the understanding of acupuncture for migraine with realtime visualized evidence.

e Study Design of Acupuncture-Neuroimaging Studies for Migraine.
e design of the acupuncture-neuroimaging study for migraine could be divided into four categories according to the contents.
First, at the early stage, these studies [8-10, 18-21, 28, 29] mainly designed the self-controlled trial to investigate whether acupuncture could elicit cerebral responses on patients with migraine, accounting for more than one-third of studies. Although these studies only set one group to observe the cerebral signal changes comparing the brain activities at the baseline with that at the end of acupuncture therapy, these studies provided preliminary references for future studies. However, this kind of design had limitations what factors (such as the efficacy of acupuncture and moxibustion, placebo effect of acupuncture, or the natural course of migraine) caused the alterations of cerebral activities were still unclear.
Second, people gradually tended to investigate whether the cerebral changes elicited by acupuncture were different from those by SA, to explore the different cerebral activities between the acupuncture efficacy and placebo efficacy. e majority of these studies selected functional connectivity (FC) to analyze the neuroimaging data. FC has developed to the popular direction of the neuroimaging analytical method in recent years. However, few kinds of research concentrated on the depth, the intensities of acupuncture stimulation, or other factors that influenced the efficacy of acupuncture for migraine. e application of FC based on the specific ROI of the brain, rather than the whole brain, is a substantial limitation of these studies. us, it is valuable to explore the central mechanism of acupuncture for migraine from different perspectives including the influencing factors and large-scale whole brain networks.
ird, some studies investigated the central mechanism of acupuncture therapy for some subtypes (menstrual migraine or some stages of migraine). at means investigators mostly focused on the central mechanisms of acupuncture for specific subtypes of migraine. It has been proved that the  cerebral activities of the ictal or interictal phase of the migraine existed differently [34][35][36]. e migraine with aura, chronic migraine, different phases of migraine was hardly explored. erefore, future studies can explore other subtypes of migraine and apply the experimental pain to model the different phases of migraine to fully explore the cerebral responses to acupuncture for migraine.
Fourth, with the application of the machine learning (ML) algorithm in neuroimaging studies, people gradually began to predict the efficacy of acupuncture for migraine with neuroimaging data by the multivariate pattern analysis method. In this review, four studies applied ML to explore the structural and functional properties of the brain whether it can contribute significant information to predict the efficacy of acupuncture for migraine at the individual level [15,16,22,24]. Unlike the univariate group-level statistics, ML can predict the efficacy of acupuncture for migraine with valid multivariate characteristics and provide specific clinical guidance at the individual level, which has gradually attracted increasing attention by researchers in recent years. erefore, integrating ML and neuroimaging techniques will be the hot direction to identify neural "biomarkers" from neuroimaging data automatically at the individual level and predict the acupuncture efficacy for migraine.
Of the included studies, VAS was the most used scale for pain assessment for migraine. Besides, the clinical efficacy of migraine mostly focused on symptoms such as the frequency, intensity, and attack duration of migraine. Only five studies applied emotional scales and two studies applied disease-specific scales, MSQ, to assess the emotion and QoL of migraine patients. e pain experience is multifaceted and complex, encompassing multiple dimensions including psychological, cognitive, perception, affective, and behavior [37]. Except for the pain, the impaired functioning, psychological responses, and participant ratings of overall improvement are also associated with migraine [38][39][40]. erefore, the multidimensional characteristics of migraine such as psychological, physical functioning, and participant ratings of overall improvement should be assessed simultaneously to explore the central mechanisms of the analgesic efficacy of acupuncture for migraine. including MRI, PET, and H-MRS were applied. fMRI was the most commonly used technique to explore the cerebral activities of acupuncture for migraine. fMRI indirectly measures brain activity based on the blood oxygenation level-dependent effect [41], PET-CT relies on the exogenous tracer for measuring the functional and metabolic changes [42], and H-MRS explores the neurochemical changes of the brain [43]. Only one study combing DTI and fMRI to explore the cerebral alterations of acupuncture for migraine [20] ofthe included studies. Multimodal neuroimaging can integrate the strengths of each imaging modality and make the data cross-validation. erefore, future studies can combine the multimodal neuroimaging techniques to detect the dynamic brain activity of acupuncture for migraine to make the results complementary and cross-validation. For example, combine molecular imaging such as H-MRS with fMRI to detect cerebral activities from the molecular and functional levels. H-MRS can be a complementary neuroimaging technique to fMRI, which can identify static as well as dynamic levels of specific brain neurotransmitters that likely relate to the fMRI signal involved in pain processing and modulation of acupuncture for migraine [44]. Besides, combing fMRI with electroencephalography can provide complementary information of brain activity (hemodynamic and electrophysiological, respectively) with the spatiotemporal resolution, which can be applied to detect the dynamic cerebral activities of acupuncture for the whole migraine attacks.

Pain Pathways Participating in the Acupuncture for Migraine.
e neuroimaging studies about acupuncture for migraine demonstrated that acupuncture modulates a widely distributed network of brain areas including the frontal lobe, temporal lobe, occipital lobe, parietal lobe, diencephalon, brainstem, and cerebellum ( Figure 2). According to the high frequency of altered brain regions, the important pain pathways related to the acupuncture for migraine of the included studies were clarified (Figure 2).

e Medial Pain System Participating in the Cerebral
Responses to Acupuncture for Migraine. In this review, the included studies have suggested that the regions of the medial pain system, including the anterior cingulate cortex (ACC), insula, and middle frontal gyrus (MFG), are participating in the cerebral responses to acupuncture for migraine (Figure 2(b), orange nodes). e medial pain pathway projects to the limbic system including the ACC and insula via the medial thalamus participating in the process of motivational-affective aspects, cognitive-evaluative aspects, and memory of pain [45,46]. Chronic pain can lead to persistent emotional disorders such as anxiety, fear, and influence brain processing on many levels [47][48][49]. Cumulative evidence suggested that ACC contributed to the response to the process of anxiety and fear of pain [50,51] and insula involved in the processing of the emotional motivation dimension of pain [52,53]. MFG is often referred to as the dorsolateral prefrontal cortex (dlPFC) [54], dlPFC is a core putative target for modulation of pain-related fear responses [55]. Previous studies suggested that the altered function of the regions of the medial pain system participates in the pathology of migraine [56,57]. In this review, the key regions of the medial pain system are involved in the modulation of acupuncture for migraine, which indicated that acupuncture may exert its therapeutic efficacy on migraine by modulating the medial pain system.

e Lateral Pain System Participating in the Cerebral
Responses to Acupuncture for Migraine. e thalamus cortex is known to be a relay center for the processing of nociceptive inputs, and the thalamus can be divided into medial and ventrobasal parts [58]. Pathways from the medial thalamus project diffusely to wide areas of the cortex and together make up the medial pain system. e ventrobasal thalamus receives input from the ascending tract, sends fibers to the primary and secondary somatosensory areas of the cerebral cortex, and makes up the lateral pain system. e postcentral gyrus (PoCG) was a part of secondary somatosensory areas of the cerebral cortex. e lateral pain system was where refined localization and discrimination of pain occurred. e altered thalamus and PoCG involved in the lateral pain system pointed to abnormal pain processing and modulation in migraine [59][60][61]. After summarizing the neuroimaging results of acupuncture for migraine, thalamus and PoCG were the high-frequency regions that can be regulated by acupuncture for migraine. erefore, the lateral pain pathway (Figure 2(b), yellow nodes) processing the localization and discrimination of pain was also the important region of acupuncture for migraine.
Besides, the results indicated that the key regions of default DMN [62] (Figure 2(b), green nodes), including the media prefrontal cortex, cingulate cortex, precuneus, medial temporal lobe, insula, and hippocampus, are involved in the cerebral responses to acupuncture for migraine. DMN is associated with attention, memory, prospection, and selfreferential processing [63][64][65]. Recent studies showed that altered DMN is associated with many chronic pain conditions [66][67][68], and it is the primary network affected by chronic pain [69]. Previous studies have also found that DMN plays an important role in pain modulation of migraine [70][71][72]. In this review, most regions of DMN are involved in the modulation of acupuncture for migraine, which indicated that acupuncture may exert its therapeutic effects by modulating the distributed regions for migraine.

e Descending Pain System
Participating in the Cerebral Responses to Acupuncture for Migraine. Disruption of the balance of descending pain modulation associated with chronic pain has been confirmed [73,74]. e well-known descending pain systems majorly contained periaqueductal gray (PAG) and rostral ventromedial medulla (RVM). PAG is the primary region in controlling the modulation of the descending pain modulatory system in the brainstem [75,76] and affects descending pain modulation primarily through its reciprocal connections with RVM [77]. e descending pain system is significantly implicated in the dysfunction of the descending pain modulatory pathways in Pain Research and Management migraine [78][79][80]. e PAG and RVM were activated after the treatment of acupuncture in this review, which suggested that the functions of the descending pain system were not only a response to migraine but also an underlying brain pathway that was perhaps at the core of modulating the pain of migraine.
Further analysis of the relation between the study designs and results indicated that the medial pain pathway was the core altered region during VA vs. SA control modality about the long-term effect [6-8, 12, 13, 16, 26, 27, 33]. e results indicated that the role of the medial pain system, which mainly participated in the pain-related emotional response, was one of the key modulations of acupuncture for migraine. Additionally, the results about adopting one treatment session to explore the immediate-efficacy acupuncture for migraine [9,10,12,14,18,19,21,26] mainly concentrated on the lateral pain system. e results manifested that the lateral pain system was another key pathway processing the sensation of pain involving the modulation for acupuncture. In other words, the pain experience has a multidimensional nature. e modulation of acupuncture was multimodal rather than unidimensional, which contained the medial pain system for processing the emotion loop, lateral pain system for processing the sensory loop, and descending pain system for inhibiting pain.

Conclusion
is review mainly overviewed the study design and main findings of acupuncture for migraine by neuroimaging techniques. is review manifested that acupuncture could elicit cerebral responses on patients with migraine and different from SA. Current studies began to use the ML to predict the efficacy of acupuncture for migraine. In the future, researchers can promote the integration of clinical and neuroimaging data with a bigger sample size to predict the efficacy of acupuncture for migraine and provide specific clinical guidance at the individual level. Data Availability e datasets of this study can be requested from the first author at mapeihong@stu.cdutcm.edu.cn.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
Peihong Ma, Xiaohui Dong, and Yuzhu Qu contributed equally to this article. Fang Zeng designed the study.Peihong Ma, Xiaohui Dong, and Yuzhu Qu participated in screening studies. Yuke Teng and Kunnan Xie extracted the data from the included studies. Zhaoxuan He, Shirui Cheng, and Zilei Tian analyzed the data. Peihong Ma, Ruirui Sun, and Tao Yin drafted the manuscript. Siyi Yu and Fang Zeng revised the draft. All authors have read and approved the publication of the final manuscript. Acknowledgments e authors thank Hengjun Liu, Tongji University for drawing the picture.
is study was supported by grants from the National Natural Science Foundation of China (Grant no. 81973960), Sichuan Province Scientific and Technological Innovation Team for Youths (Grant no. 2019JDTD0011).

Supplementary Materials
Supplementary  Table 4.