The following seven databases were searched from inception to December 31, 2020: PubMed, EMBASE, Cochrane database, China National Knowledge Infrastructure (CNKI), Chinese Biomedical Literature (CBM), Chongqing VIP Database (VIP, Chinese Database), and Wanfang Database (WF, Chinese Database). The language was restricted to English or Chinese. The keywords were a combination of “low back pain,” “acupuncture,” and “neuroimaging technologies.” The search strategy for each of the electronic databases queried is shown in Supplementary Table 1. Additional studies were found by screening references of included articles and relevant reviews.

Studies were eligible for inclusion if they (1) were peer-reviewed original research conducted in LBP patients, (2) applied acupuncture as the intervention, and (3) used structural and functional magnetic resonance imaging (sMRI, fMRI), positron emission tomography (PET), electroencephalography (EEG), and functional near-infrared spectroscopy (fNIRS), etc.

We excluded the experimental pain model, conference abstracts, case reports, protocols, animal study, monograph, and reviews.

Study selection, data extraction, and assessment of risk of bias were performed independently by two investigators (XD and YQ). Two Cochrane's risk of bias tools (Sterne et al., 2016, 2019) were utilized for evaluating the risk of bias of included randomized controlled trials (RCTs) and non-RCTs, respectively. Discrepancies were addressed by discussion or consulting a third party (PM).

For the eligible studies, we retrieved the following: (1) basic information: first author's name, year of publication, country, and diagnosis; (2) methodology: characteristics and sample size of participants, study design, acupuncture intervention (treatment courses, acupoints, manipulation modality, and needle sensation), imaging modality, and analysis methods; and (3) brain imaging data.

There is a great difference in the methodology issues of the included studies, and a meta-analysis was considered inappropriate. A descriptive analysis was presented to summarize (1) the brain alterations of LBP patients, (2) acupuncture-induced brain alterations, (3) brain functional alterations induced by MA and EA, and (4) acupuncture-related brain activities in resting-state and task-state fMRI.

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram (Page et al., 2021) of literature search and screening process is shown in Figure 1. A total of 310 results were yielded from the databases. After deduplication and screening phases, 19 studies (Ji et al., 2007; Junhai et al., 2008; Yongsong et al., 2011; Li et al., 2012, 2014; Guoqiang et al., 2014; Jialiang et al., 2014; Tao et al., 2015; Lin and Xianmo, 2017; Yijun et al., 2017; Makary et al., 2018; Lee et al., 2019; Tu et al., 2019; Xiang et al., 2019; Yan et al., 2019; Kim et al., 2020; Lan et al., 2020; Liu et al., 2020; Yu et al., 2020) between 2007 and 2020 were incorporated in this review (Table 1). Among the articles, 14 studies were performed in China (Ji et al., 2007; Junhai et al., 2008; Yongsong et al., 2011; Li et al., 2012, 2014; Guoqiang et al., 2014; Jialiang et al., 2014; Tao et al., 2015; Lin and Xianmo, 2017; Yijun et al., 2017; Xiang et al., 2019; Yan et al., 2019; Lan et al., 2020; Liu et al., 2020), three were conducted in the USA (Tu et al., 2019; Kim et al., 2020; Yu et al., 2020), and two were conducted in South Korea (Makary et al., 2018; Lee et al., 2019). Five kinds of diseases were involved in these studies: chronic low back pain (Li et al., 2014; Tu et al., 2019; Xiang et al., 2019; Yan et al., 2019; Kim et al., 2020; Yu et al., 2020), non-specific low back pain (Makary et al., 2018; Lee et al., 2019), lumbar disc herniation (Yongsong et al., 2011; Guoqiang et al., 2014; Jialiang et al., 2014; Tao et al., 2015; Lin and Xianmo, 2017; Yijun et al., 2017; Lan et al., 2020), sciatica (Ji et al., 2007; Li et al., 2012; Liu et al., 2020), and low back and leg pain (Junhai et al., 2008).

Brain imaging data of frequently reported brain regions are summarized in Table 2.

In this review, five studies reported that LBP patients showed cerebral functional and structural alterations compared with HC. The structural alterations mainly include the elevated GVM in the SI and reduced FA in SI–back regions. The functional changes usually manifest as reduced functional connectivities of DMN; higher ReHo values in the bilateral inferior temporal gyrus, left STG, and left superior parietal gyrus; and lower ReHo values in bilateral postcentral gyrus, bilateral superior frontal gyrus, and right supplementary motor area. Furthermore, a systematic review on neuroimaging studies of LBP patients indicated that brain regions such as the ACC, insula, mPFC, and cerebellum were involved in the central pathological changes of LBP patients (Kregel et al., 2015). These studies mapped the functional and structural alterations in LBP patients and provided the potential target for exploring the central responses to acupuncture stimulation in LBP patients.

For LBP patients, acupuncture can elicit widely cerebral responses. These brain regions included the PFC (middle frontal gyrus and superior frontal gyrus), precentral gyrus, ACC, PCC, insula, thalamus, postcentral gyrus, putamen, precuneus, angular gyrus, parahippocampus, PAG, rostral ventral medulla (RVM), posterior inferior parietal lobe, and cerebellum regions and were mainly distributed in the “pain matrix,” DMN, salience network (SN), and descending pain modulatory system (DPMS) (Figure 3). Previous studies have reported the abnormal connectivity in DMN and SN in pain disorders (Zhao et al., 2017; Tu et al., 2020). In this review, several studies found that acupuncture can positively regulate the function of DMN and SN (Li et al., 2012, 2014; Lee et al., 2019; Xiang et al., 2019; Yan et al., 2019). Multiple neuroimaging studies suggested that modulating the activity of DMN is an important mechanism of acupuncture therapy (Deng et al., 2016; Fu et al., 2017; Sun et al., 2021) and regulating the cerebral function of the “pain matrix” is the common characteristic of acupuncture analgesia (Zhao et al., 2014; Shen et al., 2019). Thus, improving abnormal structural and functional activities in the brain of the LBP patient is an important mechanism of acupuncture treatment for LBP. In addition, abundant evidence suggests that modulating the DPMS, comprising the PAG and RVM, is one of the mechanisms of acupuncture analgesia (Chen et al., 2015; Li et al., 2016).

Among dozens of brain regions which responded to acupuncture for treating LBP, the PFC, insula, cerebellum, SI, and ACC were the main reported brain regions. The PFC and ACC are not only the main nodes of the “pain matrix” but also the key regions in DMN. They participate in pain modulatory, pain anticipation, affective, and cognitive processing (Tracey and Mantyh, 2007; Qin et al., 2008). It is interesting that Makary et al. reported that activation in ACC was unique to verum acupuncture and that activation in the PFC was observed in the sham acupuncture (Makary et al., 2018) and significantly correlated with the belief in acupuncture effectiveness score (Makary et al., 2018). Actually, the PFC is believed to be linked with expectancy-related modulation of pain processing (Casey, 1999). As a key region in the pain matrix and SN, the insula plays an important role in the cognitive and affective perception of pain (Kong et al., 2006), manifests a significant activation in chronic pain (Ihara et al., 2019) and experimental pain (Peyron and Fauchon, 2019), and is widely involved in acupuncture analgesia (Chae et al., 2013). In this review, nine studies identified the participation of insula in acupuncture for treating LBP. The cerebellum plays an important role in sensorimotor and vestibular control and also participates in cognition, autonomic, and emotional control (Schmahmann, 2019). Hui et al. reported that activation of the cerebellum elicited by pain occurred during the acupuncture stimulations (Hui et al., 2005). The SI cortex is a major component in the pain matrix participating in pain localization and discrimination. Kim et al. found that 4 weeks of acupuncture treatment for LBP can normalize the anatomic alterations of both gray matter (GM) and white matter (WM) in SI including decreasing the GVM and increasing the FA and AD. A similar structural modulation in SI was reported in a neuroimaging study on acupuncture for treating carpal tunnel syndrome (Maeda et al., 2017).

Generally, the majority of acupuncture-neuroimaging studies focus on the functional changes resulting from acupuncture stimulation, while a few studies centered on the structural alteration induced by acupuncture intervention. Among the 19 studies included in this review, only two studies reported the improvements in altered GM/WM of LBP patients. This phenomenon of emphasizing the function over structure in acupuncture-neuroimaging studies is closely related to the characteristic that acupuncture is good at regulating functional abnormalities. The few studies on acupuncture influencing the brain structure provide valuable evidence for the promotion of structural plasticity in the central nervous system (CNS) by acupuncture. Future studies could pay more attention to the structural plasticity in the CNS induced by acupuncture and take the accumulation effect of long-term acupuncture treatment and the cerebral structural changes in physiological period into consideration.

MA and EA are the two main modalities in acupuncture clinic practice. The advantages of MA are traditional and convenient, while the strong point of EA is quantifiable. Some studies hold that EA was more effective than MA in analgesia (Kong et al., 2002), while some studies hold that MA produced a better-sustained effect than EA (Schliessbach et al., 2011). In this review, among the 17 studies on acupuncture regulating the cerebral function in LBP patients, 13 studies selected MA as acupuncture intervention and four studies used EA. Either MA or EA could elicit the cerebral activity changes in the PFC, insula, SI, and ACC. These regions all belong to the “pain matrix” and the high-frequency brain regions in acupuncture-neuroimaging for pain. It is a pity that there was no published study which investigated the similarities and differences in cerebral responses between MA and EA in LBP patients. Some studies on healthy subjects have shown that both EA and MA could activate the PFC and insula (Kong et al., 2002; Jiang et al., 2013), EA produces more widespread fMRI signal changes than MA (Napadow et al., 2005), and EA can produce more activation and less deactivation than MA (Li et al., 2003; Schliessbach et al., 2011). These studies on healthy subjects suggest that different brain mechanisms may be recruited during MA and EA (Li et al., 2003).

Resting state and task state are the two main study designs in fMRI. The advantage of ts-fMRI is that it can directly reflect the effects of an explicit task in the brain, while the lack of tasks makes rs-fMRI quite simple in experimental design and easy to cooperate with patients. In this review, 13 studies used resting-state fMRI, and five studies used task-state fMRI. The common brain regions that were induced by acupuncture in resting-state and task-state fMRI were the PFC, insula, SI, SII, brainstem, and ACC. Generally, ts-fMRI is used to evaluate the instant effect of acupuncture analgesia, and rs-fMRI is applied to evaluate the cumulative effect of acupuncture. Notably, the angular gyrus was altered when evaluating the cumulative effect of acupuncture, but not when evaluating the instant effect. The angular gyrus is located in the anterolateral region of parietal lobe and is an important part of the DMN. Liu et al. (2021) found that there was a significantly greater ReHo value increase in the angular gyrus after the 12th acupuncture sessions compared with after the first acupuncture session in migraine patients, which indicated that the cumulative effect of acupuncture is more extensive and significant than the instant effect. This review showed that in the neuroimaging studies of acupuncture treatment for LBP, researchers not only pay attention to the instant effect of acupuncture but also pay attention to the cumulative effect of acupuncture.

In this review, five out of 13 rs-fMRI studies used ReHo as analytical approach. As an example of functional segregation, Reho has the advantages of simplicity, stability, and repeatability. However, since the brain is an integrated network rather than an isolated cluster, a functional integration approach is more popular nowadays. Therefore, future studies combining functional segregation with functional integration approach will help to better clarify the mechanism of acupuncture treatment for LBP.