Acupuncture has emerged as a common modality of alternative and complementary therapeutic intervention in the Western medicine. In spite of its public acceptance, an unequivocal scientific explanation regarding physiological and biological mechanisms underlying acupuncture has not been attained and awaits further investigations. One unresolved but fundamental question is whether acupuncture needling at certain acupoints can produce functionally specific effects in the brain compared to a sham or placebo control procedure.
Previous neuroimaging studies have revealed that acupuncture stimulation can elicit widespread cerebro-cerebellar brain regions [1–6], largely overlapping with the neural networks for both pain transmission and perception . These regions process information in circuits that can broadly be assumed to engage: the affective (amygdala, hippocampus), sensory (thalamus, primary (SI) and secondary (SII) somatosensory cortices), cognitive (ACC, anterior insula), and inhibitory (PAG, hypothalamus) processing during the experience of pain . Several studies on brain responses to acupuncture stimuli in patients with chronic pain or pain condition compared with controls have also found prominent signal attenuations in the amygdala and SI, as well as signal potentiations in the hypothalamus and motor-related areas [9–11]. Moreover, a very recent study has found that the underlying analgesia efficacy of acupuncture mainly involves the underlying molecular pathways, particularly by activating A1 receptor . This evidence has brought to light the fact that the central representation of a peripheral acupuncture signal may involve a network of neurons, which are widespread distributions across multiple levels of brain areas.
To date, most of neuroimaging studies have primarily focused on the spatial distribution of neural responses to acute effects of acupuncture. However, the acupuncture needling itself is not sufficient to produce its analgesia effects . Evidence from both human behavior and animal studies has indicated that a striking feature of acupuncture analgesia, in both human and animals, is its longevity--a delayed onset, gradual peaking and gradual returning [14–16]. For a typical 30-min acupuncture session, the pain threshold has a slowly increase tendency even outlasting the treatment . We infer that acupuncture procedure typically involves two administration steps: (1) needling stimulation in deep tissue with skin piercing and biochemical reaction to tissue damage, and (2) prolonged effects after the removal of acupuncture needle stimulation [13, 17–19]. It is also substantiated that the physical needling stimulus, as well as the delayed effect of acupuncture, can similarly activate many brain areas [18, 19]. Careful interpretations of acupuncture intervention depend on how to effectively characterize the nature of temporal variations underlying neural activities that give rise to hemodynamic responses, rather than how to simply detect the occurrence of such changes.
Conventional statistical fMRI analysis of acupuncture has typically adopted the hypothesis-based approach (general linear model, GLM), and mainly tested whether activity in a brain region is systematically related to some known input function [3, 5, 6, 20]. In other words, this model-based approach implicitly embodies specific assumptions or requires a priori knowledge about the shape of the time courses to be investigated. Since the temporal profile of acupuncture-associated response is difficult to specify in advance, the GLM approach is limited and may be susceptible to errors . Only recently, independent components analysis (ICA), using few a priori assumption, is applied to extract reliable patterns underlying the psychological activity of acupuncture [19, 21]. However, this method still lacks in accuracy to make direct inferences on whether a component (brain network) varies over time and when changes occur in certain time points. Great emphasis has, therefore, been given to understand temporal characteristics of these spatially defined brain regions, with considerations for how multiple levels of their dynamic activities in concert cause the processing of acupuncture.
Built upon our previous studies [17, 18, 22, 23], we have formulated a hypothesis that distinct, time-dependent changes elicited by acupuncture are mirrored by the temporal responses observed within the wide brain networks, which has been suggested to participate in different stages of acupuncture process. To address this question, we adopted the control theory based approach namely change-point analysis, in which a hierarchical exponentially-weighted moving average (HEWMA) approach was used to make direct inferences on acupuncture-related activities [24, 25]. This method can also effectively deal with high individual variabilities of neural responses induced by acupuncture .