Corticomuscular coherence and its non-invasive modulation in stroke applications: a narrative review

Abstract

Neural activities are widely observed in the central nervous system and are essential for movement execution. Corticomuscular coherence (CMC) represents the coherence of oscillatory communication between motor cortex and peripheral muscles. Motor function impairments caused by stroke often involve abnormal CMC patterns. This review aimed to investigate CMC and how to modulate them for post-stroke functional recovery. We first introduced the origins of CMC in movement execution. We then explored how stroke affects CMC. Based on these abnormal CMC patterns, we summarized potential non-invasive neuromodulation strategies. A deeper understanding of CMC in beta band (∼15-30 Hz) could clarify the physiological mechanisms underlying movement and aid in improving stroke rehabilitation.

1. Introduction

Neural oscillations depict rhythmic and synchronized signal patterns, which were first reported by Caton in 1875. Later, in 1924, Berger quantified alpha oscillations, ranging from 8 to 12 Hz, and was the first to associate them to sleep and wakefulness states (Tassinari, 2019). Neural oscillations recorded by electrophysiological signals (e.g., EEG/MEG/EMG) involve synchronization and phase locking of neuronal activities, which likely enhances communication among neurons and facilitate coordinated information processing across brain regions. Fries proposed the "communication through coherence" theory, which suggests that strong effective connectivity requires rhythmic synchronization within pre- and postsynaptic groups as well as coherence between them (Fries, 2015). Corticomuscular coherence (CMC), reflecting the functional coupling between the motor cortex and muscles during movement, is a key concept within this framework (Liu et al., 2019). CMC was first observed as significant coherence within the beta-frequency band (15–30 Hz) between sensorimotor cortex and contralateral muscle during sustained muscle contractions in both monkey and human (Baker et al., 1997; Conway et al., 1995). Following the initial discovery, studies proposed that rhythmic neural activity originating in the primary motor cortex (M1) and corticospinal tract mediates corticomuscular coherence, given that corticospinal cell activity does encode motor cortical oscillations (Baker et al., 2003). However, the sole pathway theory was challenged by the loop theory, which posits that corticomuscular coherence is generated by a loop between cortex and the periphery, involving not just descending (motor) propagation, but also ascending (sensory) transmission. Stronger CMC indicates that the brain and muscles are working in harmony during movement, whereas weakened CMC reflects poor coordination between cortical signals and muscle output, leading to impaired motor performance. Specifically, during movement execution, bidirectional CMC in both afferent and efferent directions reflect the communication pattern of the central nervous system sending motor commands and receiving afferent sensory feedback from the muscles.

CMC has emerged as a promising non-invasive biomarker for probing motor network dysfunction across a spectrum of neurological and psychiatric disorders. One condition where CMC is significantly disrupted is in stroke. Stroke is a disease characterized by the death of brain neurons due to the blockage or rupture of cerebral blood vessels and is one of the leading causes of adult disability worldwide (Gorelick, 2019). According to World Stroke Organization, there are 12 million new strokes, 62% of which are ischemic while 28% are hemorrhagic (Feigin et al., 2022). Every year 6.5 million stroke patients die as a result, with more than 5 million new stroke survivors living with disability. The sequelae of stroke include consciousness, motor, and sensory impairments, and more than 86% of survivors suffer from motor impairment or motor combined with other impairments (Gittins et al., 2020). Stroke induces abnormal CMC patterns (Aikio et al., 2021; Liu et al., 2019). This disruption impairs the brain's ability to produce the organized and synchronized signals needed for effective motor control, leading to motor deficits such as muscle weakness, impaired coordination, and difficulty performing voluntary movements. Conventional rehabilitation treatments often fall short of achieving full recovery, with more than half of stroke survivors suffering from moderate to severe motor disabilities, such as hemiplegia and spasticity (Hendricks et al., 2002). Pathological alterations in CMC magnitude, frequency, and topography are often correlated with the severity of motor impairments, such as weakness, bradykinesia, or tremor (Liu et al., 2022; Rossiter et al., 2013). This clinical relevance has catalyzed a pivotal shift in the field, from observation to intervention, posing a critical question: if aberrant CMC is a feature of motor pathology, can we therapeutically modulate this coherence to restore or enhance motor function? This question has opened up exciting translational opportunities and spurred the development of novel neurotechnological approaches.

Therefore, this review aims to systematically summarize and critically evaluate the current landscape of non-invasive modulation techniques used to influence CMC. Recent advancements in non-invasive neuromodulation techniques such as tES,ultrasound, have shown potential in modulating CMC in both animal models and clinical settings (Kudo et al., 2022; Wang et al., 2019). These methods aim to either excite or inhibit specific brain regions to promote cortical plasticity, helping the brain reorganize neural connections to compensate for damaged motor areas. By stimulating the brain or peripheral nervous system, neuromodulation can strengthen neural circuits compromised by stroke, facilitating improved motor function and coordination. Targeting abnormal stroke-induced CMC may improve the efficiency of motor function recovery.

In this context, understanding the relationship between CMC and neuromodulation provides valuable insights into the mechanisms underlying motor deficits and their potential treatments. Modulating CMC through neuromodulation offers a potential promising strategy for improving motor recovery in stroke patients. This narrative review focuses on:

• 1.The origins and quantification of CMC,
• 2.The effects of stroke on CMC patterns,
• 3.Non-invasive neuromodulation techniques for modulating CMC. The ultimate goal is to expand our understanding of how CMC modulation can enhance post-stroke motor function recovery, as shown in Fig. 1.

Fig. 1.

CMC enables people to perform voluntary movement, while stroke disrupts the CMC patterns and leads to motor impairments. Modulation strategies provide potential solution for movement disorder recovery.

We conducted our initial literature search using PubMed and Web of Science. For exploring CMC generation, function in movement, the key words included: corticomuscular coherence; movement; sensorimotor; beta band/oscillations; gamma band/oscillations; static, isometric; dynamic, isotonic; movement onset/stop; movement coordination. To investigate stroke-induced CMC changes, the keywords included: stroke; corticomuscular coherence; directed corticomuscular coherence; motor function. For non-invasive neuromodulation in movement oscillations studies, key words included: neuromuscular electrical stimulation; transcranial directed current stimulation; transcranial alternate current stimulation; transcranial electrical stimulation; transcranial magnetic stimulation; ultrasound stimulation; neuromodulation; oscillations; cortcomuscular coherence.

2. Origin and quantification of corticomuscular coherence

Over the past hundred years, the understanding of brain oscillations has expanded significantly, with different oscillation frequencies now recognized to be associated with various brain functions. Neural oscillations observed in the central nervous system are thought to arise from rhythmic or synchronous activity among large populations of neurons. Coherent oscillations among neuronal groups create temporal windows of communication, facilitating effective interaction and providing a flexible pattern of communication structure (Fries, 2005). This structure enables distinct or distant neural groups to coordinate precisely via specific phase relationships within one or multiple oscillatory frequency bands (Whittington et al., 2018). During movement execution, communication between the central nervous system and peripheral muscles can be represents through CMC, which denotes the functional coupling between the brain's motor cortex and peripheral muscles (Mima and Hallett, 1999). Currently, it is believed that the source of CMC originates from the cortex. Oscillations generated in the motor cortical circuit propagate motor intention and planning information downward to spinal motor neurons, activating motor units to produce motor unit action potentials. This stimulation leads to muscle fiber contraction and the generation of electromyogram (EMG) activity (Baker and Lemon, 1998; Williams and Baker, 2009).

In recent years, the theory of sensorimotor loop integration has been proposed, suggesting that the descending corticospinal tract is unlikely to be the sole pathway in generating CMC. More evidence suggests that a loop between the cortex and periphery, combined with both descending (motor) propagation and ascending (sensory) pathways, is involved in generating CMC. Oscillatory synchronization between brain regions involved in sensation and movement promotes sensorimotor integration, facilitating efficient feedback processing during motor execution. In the ascending pathway, oscillatory information encoding the state of muscles, joints, and other proprioceptors is transmitted to the central nervous system for the correction of motor output commands, enabling precise motor control. Studies in non-human primates and clinical settings revealed have indicated an interaction between descending and ascending pathways (Tsujimoto et al., 2009; Witham et al., 2011). The oscillations in the afferent pathway implied that central oscillations pass around a peripheral feedback loop and may be involved in sensorimotor integration (Baker et al., 2006). Currently, two potential functions are attributed to neural oscillations in the sensorimotor loop: 1) Neural oscillations serve as "probe pulses": by sending known signals to muscles and measuring sensory input, they are used to assess peripheral status within the sensorimotor cortex; 2) They contribute to the promotion of stable motor states. Neural oscillatory activity enhances responses to peripheral feedback in the presence of disturbances, thus maintaining stable motor control by amplifying feedback to disturbances (Gilbertson et al., 2005).

CMC can be quantified by different approaches which reveal distinct aspects of the underlying neurophysiological relationship. The choice of method depends heavily on the specific research question, whether it is about the strength of coupling, its directionality, or its underlying linear or non-linear nature. Here, we review the primary methods used in the literature. The most stablished and widely used method for quantifying CMC is Magnitude-Squared Coherence (MSC) (Liu et al., 2020), which measures the non-directional linear correlation between two signals in frequency domain as following,

$$Cxy\left(f\right)=\frac{\mid Pxy\left(f\right){|}^2}{Pxx\left(f\right)Pyy\left(f\right)}$$ (1)

Where $x$ and $y$ are EEG/EMG respectively, and $Pxx\left(f\right)$, $Pyy\left(f\right)$, $Pxy\left(f\right)$ are auto-spectral densities or cross-spectral densities at frequency $f$. To quantify the directional information flow between cortex and muscles in the descending and ascending pathways, Granger causality (GC) was introduced. GC is a statistical method for determining whether one time-series can be used to forecast another, $x_2$ “Granger causes” $x_1$ if the inclusion of past observations of $x_2$ reduces the variance of $x_1$, compared with only the inclusion of previous observations of x1. The GC was estimated as following (Gao et al., 2021),

$$x_1\left(t\right)=\sum _{i=1}^T{a}_{11,i}{x}_1\left(t-i\right)+\sum _{i=1}^T{a}_{12,i}{x}_2\left(t-i\right)+{\epsilon }_1\left(t\right)$$ (2)

$$x_2\left(t\right)=\sum _{i=1}^T{a}_{21,i}{x}_1\left(t-i\right)+\sum _{i=1}^T{a}_{22,i}{x}_2\left(t-i\right)+{\epsilon }_2\left(t\right)$$ (3)

$$g{c}_{2\to 1}=\ln \frac{var\left({\epsilon }_{o,1}\right)}{var\left({\epsilon }_1\right)}$$ (4)

where $a$ is the coefficient, ${\epsilon }_1$ and ${\epsilon }_2$ are the prediction errors from the full model, ${\epsilon }_{o,1}$ is derived from the model omitting the contribution of $x_2$. Recognizing that neural systems are inherently non-linear, information-theoretic measures were to provide a more complete assessment non-linear connectivity. Mutual information (MI) employs the entropy of high-order statistics to estimate uncertainty and non-linear connectivity as following (Tan et al., 2022),

$$H\left(X\right)=\sum _{i=1}^{n}p\left(x_i\right){\log }_{2}p\left(x_i\right)$$ (5)

$$MI\left(X,Y\right)=H\left(X\right)+H\left(Y\right)-H\left(X,Y\right)=\sum _{j=1}^m\sum _{i=1}^{n}p\left(x_i,y_i\right){\log }_2\left(\frac{p\left(x_i,y_i\right)}{p\left(x_i\right)p\left(y_i\right)}\right)$$ (6)

Where $H\left(X\right)$ is the entropy of signal $X$, and $H\left(X,Y\right)$ is the conditional entropy of $X$ and $Y$. To further quantify the directed transfer of information in non-linear context, transfer entropy (TE) was applied. It quantifies whether the past of the signal $X$ reduces uncertainty about the future of the signal $Y$, beyond what signal $Y$ own past can predict, calculated as following (Liu et al., 2021a),

$$T{E}_{X\to Y}=H\left(y_{t+u}{y}_t^n\right)-H\left(y_{t+u}{x}_t^m,y_t^n\right)={\sum }_{y_{t+u},y_t^n,x_t^m}P\left(y_{t+u},y_t^n,x_t^m\right)\log \left(\frac{P\left(y_{t+u},x_t^m,y_t^n\right)}{P\left(y_{t+u},y_t^n\right)}\right)$$ (7)

where $t$ is the discrete point of time series, $u$ represents the prediction time, m and n represents the dimensional number of $x$ and $y$. $H\left(y_{t+u}{y}_t^n\right)$ represents the conditional entropy of $Y$ based on the past of $Y$. $H\left(y_{t+u}{x}_t^m,y_t^n\right)$ denotes the conditional entropy of $Y$ based on the past of $X$ and $Y\text{.}$ Transfer spectral entropy (TSE) was introduced as a frequency domain extension of the mode-free TE method, Compared with the GC, TSE performs more optimally in detecting the nonlinear characteristics of information transfer (Liu et al., 2021a).

In summary, the choice of computational methods is an essential step in CMC research. While MSC remains the standard for assessing the strength of linear coupling, advanced methods like GC and TE provide deeper insights into the directionality in linear and non-linear aspects of the connectivity between the cortex and muscles.

3. Stroke-induced abnormal corticomuscular coherence patterns

The execution of movements requires the synergistic coordination of descending motor commands from the central nervous system and ascending sensory feedback from muscles, with neural oscillatory activity playing a crucial role. Motor impairments resulting from stroke are associated with disruptions in the functional connectivity between the central nervous system (CNS) and peripheral muscles, affecting both the descending motor command and ascending sensory feedback pathways. Consequently, stroke survivors often exhibit motor dysfunction, making it difficult to achieve precise motor control. Detailly, stroke disrupts both the descending motor control commands and the processing of sensory feedback leads to abnormal CMC pattern. Studies on stroke-induced changes in CMC patterns, particularly in the beta band, are listed in Table 1.

Table 1.

Stroke-induced CMC changes.

Reference	Subjects	Frequency range and quantification method	Task and muscles	Results
Larsen et al. (2017)	19 acute and subacute ischemia stroke +18 healthy	Beta (15–30 Hz); MSC	Thumb and index finger press; adductor pollicis (ADP) abductor pollicis brevis (APB); first dorsal interosseous (FDI)	Stroke reduced CMC compared to controls
(Katheri et al., 2014)	11 acute cortical and subcortical stroke+ 14 healthy	Beta (13–35 Hz); MSC	Thumb press; APB	Stroke reduced CMC frequency in affected side
Liu et al. (2021b)	10 chronic subcortical stroke +10 healthy	Beta (15–30 Hz), gamma (30-45 Hz); Extended Partial Directed Coherence	Arm flexion and extension; flexor digitorum superficialis, flexor carpi urinaris, flexor carpi radialis, extensor carpi ulnaris, extensor carpi radialis longus, biceps brachii (BB), triceps brachii (TB) and deltoideus triangularis (DT)	Stroke patients had significantly reduced CMC in both descending and ascending pathway
Krauth et al. (2019)	5 subacute ischemia stroke+7 healthy	Beta (12–30 Hz); MSC	Wrist extension; wrist extensor musculature	CMC Is reduced acutely post-stroke and increased during recovery
Delcamp et al. (2022)	24 chronic ischemia stroke + 22 healthy	Beta (13–31 Hz); MSC	Elbow extension; TB, BB, brachialis and brachioradialis	Stroke increased CMC in beta band
Chen et al. (2018)	16 chronic ischemia/hemorrhage stroke + 8 healthy	Beta (16–32 Hz), gamma (30–45 Hz); MSC	Uplift and maintain arm; DT, BB	Abnormal increased CMC in DT and reduced CMC was BB for stroke
Aikio et al. (2021)	29 ischemia stroke +22 healthy	Beta (13–30 Hz); MSC	Isometric contraction; extensor carpi radialis	Stroke reduced CMC peak frequency in both hemispheres and strength in affected hemisphere
Zhou et al. (2021)	14 chronic ischemia and hemorrhage stroke +11 healthy	Beta (13–35hz); GC	Isometric finger extension at 20% MVC; extensor digitorum, flexor digitorum, TB, BB	Stroke altered descending dominance, increased ascending feedbacks and prolonged descending conduction time in upper extremity.
Gao et al. (2018)	5 subacute stroke + 7 healthy	Beta (15–35 Hz); TE	Grip and elbow flexion, FDS, flexor, musculus biceps brachii, triceps	Stroke increased corticomuscular coupling in bi-direction

Stroke-induced abnormal CMC patterns varied in previous studies while most of studies reported with reduction in CMC caused by stroke, then recovery with motor function recovery. The different response may be attributed to stroke periods. Acute and subacute ischemia stroke survivals exhibited less CMC amplitude compared to healthy controls (Larsen et al., 2017), and lower CMC peak frequency in acute cortical/subcortical stroke survivals (Katheri et al., 2014). In subacute ischemia stroke survivals performing wrist extension tasks, decreased CMC gradually increases over time and with functional recovery (Krauth et al., 2019). Other study reported the CMC enhancement has been associated with improvements in upper extremity function as assessed by the Fugl-Meyer Assessment score (Khademi et al., 2022). The improvement in beta-band CMC in stroke patients is related to tactile sensitivity and hand function recovery (Aikio et al., 2021). In acute and subacute ischemia stroke survivals, they exhibited reduced CMC compared to healthy controls (Larsen et al., 2017), however the hand motor function improvement is not necessary with CMC changes. The recovery of hand strength after stroke is associated with increased motor-evoked potential amplitudes and decreased thresholds (Thickbroom et al., 2002).

The muscle type may be a key factor in CMC changes. In the uplift and maintain arm task, the CMC pattern varied in deltoid and biceps brachii muscles, where increased CMC in deltoid and reduced CMC was biceps brachii for chronic ischemia and hemorrhage stroke survivals (Chen et al., 2018). Study also reported the increased beta band CMC in both antagonist and agonist muscles in elbow extensions tasks (Delcamp et al., 2022) in chronic ischemia stroke survivals. Based on symbolic transfer entropy analysis, greater strength of the bi-directional connectivity (EEG-to-EMG and EMG-to-EEG) was found in subacute stroke survivals compared to healthy controls, especially in beta band during the upper limb movement (Gao et al., 2018).

Overall, the varying CMC responses induced by stroke can be attributed to factors such as stroke subtypes, location, duration, task types, and muscle types. These differences in CMC responses may also indicate the need for distinct rehabilitation strategies. In cases of reduced CMC, this decrease may be due to a lack of motor commands, commonly observed in acute or subacute phases. Rehabilitation during these phases should focus on re-establishing connections between the central nervous system (CNS) and peripheral muscles, emphasizing strengthening motor commands and activating muscles to prevent atrophy. This approach aims to develop new motor control strategies, potentially leading to increased CMC as motor function recovers. In the chronic phase, stroke survivors may exhibit higher or similar CMC levels, suggesting compensation by the brain and muscles during movement execution. These enhancements may involve recruiting additional brain areas for movement control, which depends on whether the lesion is subcortical or cortical and whether it includes the primary motor cortex. For individuals with abnormally high CMC parameters, rehabilitation should concentrate on executing movements with proper muscle activation rather than merely activating as many muscles as possible, akin to the refinement stage of motor learning. Thus, understanding CMC patterns may guide the development of more precise and individualized rehabilitation treatments to enhance motor function recovery.

To evaluate disruptions in dynamic information transmission in both the descending and ascending pathways, directed corticomuscular coherence (dCMC) has been utilized (Liu et al., 2019). Studies have shown that in healthy individuals, during the execution of movements, the dominant oscillations occur in the descending pathway (Mima et al., 2001). Oscillations in the beta frequency band in the descending pathway are related to the transmission of information from motor unit pools and are positively correlated with muscle coordination, enhancing the descending pathway can improve the precise control and output of force (Spedden et al., 2019). The ascending pathway of cortico-muscular oscillations is mainly used for sensory feedback. Healthy individuals can adjust and enhance the ascending sensory feedback pathway to maintain stability and balance (G. Wang et al., 2024). Functional impairments in the central nervous system of stroke patients may reduce the output of descending neural drive control. For example, distal muscles lose the dominant mode of oscillations in the beta frequency band in the descending pathway, conversely, an enhancement of ascending dCMC was observed in a less stable state, suggesting impaired movement control after stroke and a compensatory role of the ascending pathway in stabilizing voluntary movement (Zhou et al., 2021). During the execution of arm extension paradigms, the strength of the ascending and descending connections between the CNS and peripheral muscles is significantly lower in stroke patients compared to healthy subjects (Liu et al., 2021b). The reduction in descending drive leads to a trend of co-modulation between beta frequency band oscillations in the cortico-muscular coupling and the muscles of the paralyzed limb, resulting in simplified control between paralyzed limb muscles, reducing the complexity and coordination of neural-muscular control, leading to loss of motor function (Delcamp et al., 2023). Some studies have found that post-stroke motor impairments may be accompanied by a reduction in the transmission and processing of ascending pathway information. After stroke, there is a change in the dominant mode of neural-muscular coupling, with increased ascending sensory feedback compensating for the lack of descending control commands in distal limbs, thereby achieving stable motor output (Zhou et al., 2021). Disruptions in proprioceptive feedback weaken oscillations in the beta frequency band, resulting in weakened cortico-muscular oscillations in both ascending and descending pathways in stroke patients, making it difficult to achieve stable motor output (Liu et al., 2021b).

4. Non-invasive neuromodulation in corticomuscular coherence modulation

The goal of motor function recovery is to reestablish connections or promote the existed neural-muscular connections into normal state to restore function. Recently, non-invasive neuromodulation has become a potential tool to modulate oscillations and corticomuscular connection. The principles of neuromodulation base on Hebbian plasticity, that is, to induce a causal timing between the firing of two neurons and induce spike-timing-dependent plasticity (STDP) (Bi and Poo, 1998). Generally, stimulation interventions designed to leverage STDP to potentiate cortical or corticospinal interconnections aim at increasing the coincidence of action potentials in pre- and postsynaptic neuronal populations with a positive sequential timing (pre before post) in accordance to the Hebbian principles of causal association (Ting et al., 2021). The non-invasive neuromodulation paradigms can be divided into: 1) peripheral muscle modulation, 2) central nervous system modulation according to the stimulation site.

4.1. Neuromuscular electrical stimulation

For peripheral muscle modulation, the target focuses on peripheral muscle and produce contractions of paralyzed or paretic muscles by electrical current, such as neuromuscular electrical stimulation (NMES). The NMES plays a role as neuroprosthesis to improves ambulation function of stroke survivors (Chae et al., 2008). The parameters of NMES are key point in stimulation effects. Stimulation frequency modulate the effects, studies reported the NMES in 10 Hz applied to biceps brachii caused sensory response reduced corticomotor responsiveness while 30 Hz stimulation with muscle contraction promote corticomotor responsiveness (Chipchase et al., 2011). In abductor pollicis brevis, stimulation from 20 to 100 Hz revealed the corticomotor excitability is positive correlated to stimulation amplitude, In addition to activation of motor pathway, NMES also activate the ascending afferent volley directly and encapsulate secondary reafference form the muscle contraction, which probably due to a combination of motor and sensory effects (Schabrun et al., 2012). For corticomuscular coherence, study reported even a short term NMES can result in improvement of CMC in wrist flexion and verified the immediate effects of NMES, which may be the basis of long-term plasticity induced by NMES (Xu et al., 2018).

In steady-hold thumb flexion at 50% maximal voluntary contraction (MVC) task, 40 min NMES lead to significant improvement in gamma band corticomuscular coherence in both healthy and stroke subjects, meanwhile, the steadiness in sustained contraction also increased, which is related to NMES-induced strong sensory input and enhanced sensorimotor integration (Lai et al., 2016). Similar improvement also observed in lower limb task, when applying NMES to pedaling in chronic stroke survivors, both the paretic side descending and ascending cortico-muscular pathways connectivity were increased after NMES-training. These studies implied potential applications of NMES in modulating CMC during post-stroke motor rehabilitation to facilitate recovery.

However, difficulties have been found in NMES alone to precisely activate groups of muscles for dynamic and coordinated limb movements with desired accuracy in kinematics, for example, speeds and trajectories (Rong et al., 2017). Furthermore, faster muscular fatigue would be experienced when using NMES with intensive stimuli, in comparison with the muscle contraction by biological neural stimulation (Ibitoye et al., 2014).

4.2. Transcranial direct current stimulation (tDCS)

tDCS is non-invasive neuromodulation method compromise two kinds of electrodes, anode and cathode to deliver low amplitude, direct current to interest target. It induced depolarization or hyperpolarization of resting membrane potentials, thereby modulating cortical excitability (Zaghi et al., 2010). tDCS stimulation is likely to modulate glutamate transmission and gamma-aminobutyric acid transmission in the cortex, as well as the activities of dopamine, serotonin, and acetylcholine transmissions in the central nervous system, which further modulate the balance between excitatory and inhibitory inputs (Yamada and Sumiyoshi, 2021). In detail, the pyramidal neurons located under the anodal electrode are suggested to increase in excitability through depolarization of both the soma and the afferent axons while the pyramidal neurons located under the cathodal electrode decrease in excitability through hyperpolarization of the soma and afferent axons (Nitsche and Paulus, 2000; Reato et al., 2013).

It has also been reported that the direction of cortical modulation depends not only on the polarity of electrodes but also on the type and the spatial orientation of neurons as well as the stimulation intensity. For example, under certain parameters the neurons in the deeper layers of the neocortex can be activated by cathodal and inhibited by anodal stimulation, possibly as a result of the inversion of current flow associated with the neuron's spatial orientation (Creutzfeldt et al., 1962). Also, study reported that high current intensities are required to activate pyramidal cells, whereas weak stimulation is enough to activate nonpyramidal neurons (Purpura and McMurtry, 1965). Since weak direct currents led to subtle membrane depolarization, rather than direct excitable, it will make the synchronized neurons more sensitive or responsive to inputs (either excitatory or inhibitory). Simulation study applied tDCS in large-scale network model and found tDCS sharpens and shifts the frequency distribution towards higher frequency, as a consequence of polarization (Kunze et al., 2016). It provides the potential of tDCS in oscillation modulation. Here we review the oscillation modulated by tDCS in Table 2.

Table 2.

Oscillation modulated by tDCS.

Reference	Subjects	Stimulation pattern	Electrode	Location	Parameters	Task	Results
Wilson et al. (2018)	35 healthy	Anodal tDCS	5∗7cm2 sponge	Midline occipital cortex/right prefrontal cortices	2 mA, 20 min	Visual task	tDCS to the occipital increased the amplitude of local gamma oscillations,
McDermott et al. (2019)	48 healthy	Anodal, cathodal	5 × 7 cm2 sponge	Midline occipital cortex/right supraorbital cortices	2 mA, 20 min	Visual flanker task	Anodal stimulation reduced theta oscillations
Dutta and Chugh (2012)	10 healthy	Anodal	3∗3 cm2/5∗7 cm2 saline-soaked sponge electrode	Motor cortex/forehead above the contralateral orbit.	2 mA, 10 min	Standing postural steadiness	Anodal tDCS induced CMC in tibialis anterior muscle during quiet standing
Hou et al. (2022)	17 healthy	Anodal	35 cm2 sponge electrode	M1/forehead medial above the nasion	2 mA, 15 min	Dynamic and static postural stability	Anodal tDCS over M1 has an immediate improving effect on static postural stability and dynamic performance in young healthy adults
Bao et al. (2019)	30 stroke	HD anodal, cathodal	Ag/AgCl electrodes	Center: ipsilesional motor cortex	1 mA, 10 min	Wrist isometric contraction	Anode HD-tDCS promotes peak CMC of isometric contraction tasks in the broadband of interest (8–48 Hz) but not in the individual EEG bands (alpha, beta, and low gamma)
Zhan et al. (2023)	24 healthy	HD anodal	Ring-type electrodes (3.5 fcm2)	Primary motor cortex	2 mA, 20 min	Ankle dorsi-plantarflexion	Anode HD-tDCS increased beta and gamma band CMC
Fan et al. (2023)	23 stroke	Anodal		Primary motor cortex, contralateral brain area FP2	2 mA, 20 min	Grasp	tDCS increased beta and gamma band CMC in patients

For cortical oscillations, studies have found the tDCS’s modulation effects in multiple cortexes, like in occipital, prefrontal, and motor cortex. The anode tDCS could increase occipital gamma activity during visual stimulus through modulating GABAergic activity (Kujala et al., 2015; Wilson et al., 2018). Besides single target modulation, 20min anodal tDCS modulated task-related oscillations and spontaneous activity across multiple cortical areas, both near the electrode and in distant sites that were putatively connected to the targeted regions (McDermott et al., 2019). These results suggest the potential in transsynaptic oscillation modulations. For the modulated oscillation frequency range, tDCS influenced wide range of dominant peak oscillatory episodes, include continuous range like 0-15 Hz and separate range like alpha band and delta band (<4 Hz) (Kunze et al., 2016), rather than in the specific cortical frequencies for resting-state EEG, these results suggests tDCS modulates oscillatory activities in the wider band but not in the single bands (Luft et al., 2018).

In motor cortex, study reported apply tDCS for 3 consecutive days in the subacute period enhanced BDNF (brain-derived neurotrophic factor) expression and dendritic spine density in the peri-infarct motor cortex, along with increasing functional connectivity between motor and somatosensory cortices in beta, alpha and gamma bands (Longo et al., 2022). Research indicates that high-density direct current stimulation can effectively activate the motor cortex, thereby enhancing the neural drive in the corticospinal pathways and improving the neural-muscular coupling (Zhan et al., 2023). In healthy subjects, tDCS showed modulation effects in both steadiness and dynamic tasks (Dutta and Chugh, 2012, Hou et al., 2022). In chronic stroke patients, 10 min anode HD-tDCS centered in C3 area showed significant ability in modulating CMC in wide frequency range (8-48 Hz) (Bao et al., 2019). The modulation effects in stroke patients faded in 30 min, while in healthy subjects, the modulation effects were prominent in 30min post-stimulation (Kuo et al., 2013), which suggested the narrow modulation window caused by stroke. Combining tDCS with rehabilitation training in stroke patients showed tDCS increases the rehabilitation effects caused by training, and gained more CMC improvements especially in beta and gamma bands (Fan et al., 2023). These results implied the potential of taking CMC as modulated target in stroke rehabilitation.

However, large-scale clinical trials have increasingly shown that modulating cortical excitability does not effectively enhance motor function in post-stroke rehabilitation. A multicenter randomized clinical trial found no significant improvement in motor function when stroke patients underwent randomized tDCS anodal stimulation to activate the affected motor cortex or tDCS cathodal stimulation to inhibit the healthy cortex (Hesse et al., 2011). A meta-analysis of randomized controlled trials on the efficacy of tDCS after stroke showed limited evidence of its effectiveness in upper limb motor rehabilitation (Elsner et al., 2017). Further studies on how tDCS parameters influenced motor function recovery are necessary.

4.3. Transcranial alternating current stimulation (tACS) and oscillatory transcranial direct stimulation (otDCS)

Recently, transcranial alternating current stimulation has been introduced to directly modulate the ongoing rhythmic brain activity by the application of oscillatory currents. The tACS delivers sinusoidal current at a set frequency which can modulate ongoing neural oscillations by artificially inducing up-and-down states similar to patterns exhibited by endogenous rhythms. Its effects mainly manifest as neural oscillation entrainment or resonance, enabling the modulation of endogenous oscillatory rhythmic neural activity. According to the principles of timing-dependent plasticity, during stimulation, synapses in circuits that resonate at frequencies similar to the input are strengthened (Helfrich et al., 2014; Nasr et al., 2022). Huang et al. reported tACS entrains alpha oscillations and the fast-spiking inhibitory interneurons exhibit a stronger entrainment response to tACS in both the ferret experiments and the computational model, likely due to their stronger endogenous coupling to the alpha oscillation (Huang et al., 2021). Zaehle et al. delivered tACS over the occipital cortex of 10 healthy participants and found the stimulation elevated the endogenous alpha power in parieto-central area, which was related to the neural resonance active by stimulation, synapses of those circuits that have a resonance frequency similar to that of the repetitive input are strengthened during stimulation (Zaehle et al., 2010).

With the development of neuromodulation, tACS has shown the potential in motor function modulation. Studies showed the potential of tACS in modulating brain oscillations to facilitate the post-stroke rehabilitation. In non-human primate stroke model, Ganguly et al. found low-frequency epidural alternating current stimulation (ACS) increased co-firing within task-related ensembles in perilesional cortex and improved grasp dexterity (Khanna et al., 2021). It attributed the recovery to the simulated ACS drove ensemble co-firing and enhanced propagation of neural activity through parts of the network with impaired connectivity, suggesting a mechanism to link increased co-firing to enhanced dexterity. The frequency of tACS may influence the effects in post-stroke rehabilitation. By comparing the effects of 10 Hz and 20 Hz tACS in chronic stroke patients, the frequency-specific modulation was found. For stimulation, more functional interaction between sensorimotor and motor control regions was activated while no similar phenomenon was found under 10 Hz stimulation situation (Yuan et al., 2022). These studies suggest the importance of parameters in tACS in stroke rehabilitation.

Except the oscillation modulation, tACS showed ability in cortico-muscular oscillation modulation. Continuous intervention of the beta frequency band in the primary motor cortex with tACS after fine motor learning can increase the strength of beta band cortico-muscular coupling oscillations, promoting motor learning and consolidation. Stimulation during rest selectively enhances corticospinal excitability, whereas stimulation during movement selectively modulates corticospinal excitability related to executing motor actions (Arai et al., 2011). During isometric contraction, 10 Hz tACS significantly reduced low gamma band CMC, the findings suggest the modulation effects of tACS may not limited in the stimulation frequency range. Although tACS depend on the endogenous power of oscillations, the internal cross-frequency interplay between alpha and low gamma band activity may play an essential in leading to a cross frequency modulation (Neuling et al., 2013; Wach et al., 2013). In Parkinson’s disease (PD) subjects, the beta band CMC during isometric contraction and amplitude variability during finger tapping were reduced after 20 Hz tACS of primary motor cortex, while no similar phenomenon was found in healthy subjects. The decreased CMC was related to the inhibition of cortical drive, which may be related to the connectivity between M1 and the contralateral active muscles, or local cortical oscillatory entrainment during frequency-specific stimulation modulation (Krause et al., 2014).

Oscillatory tDCS (otDCS), which includes elements of DC and AC, showed the ability in simultaneously modulates the potential and oscillation activity of neuronal membranes (Herrmann et al., 2013). Groppa et al. revealed the 0.8 Hz otDCS could increase or decrease the cortico-spinal excitability with different polarity, However, these effects do not differ from control conditions utilizing tDCS, suggesting that the DC portion of the stimulation currents caused the observed effects (Groppa et al., 2010). While recently, Kudo et al. applied individualized beta-band otDCS over the primary motor cortex in healthy individuals, and found only otDCS enhanced corticomuscular coherence and corticospinal excitability while tACS and tDCS had no effects on CMC or MEPs (Kudo et al., 2022). These findings suggested the importance of delivering stimulation based on internal state and implied the potential function of otDCS in modulating corticospinal communication, which is essential in post-stroke motor function recovery.

4.4. Transcranial magnetic stimulation

Transcranial magnetic stimulation works in different way compared to TES, it delivered a strong electromagnetic field to generate current. TMS can depolarize membrane potential and primarily causes intracortical excitation of myelinated axons at their terminals to trigger action potential. These effects are addressed by producing motor response, while TES are usually sub-threshold stimulation. TMS can modulate the neuronal circuitry in the motor cortex, leading to instantaneous generation of CMC. This effect was caused either by direct activation of corticospinal cells or by activation of local neuronal circuitries in the motor cortex (Liv Hansen and Bo Nielsen, 2004). This modulation effects also depend on internal cortical oscillations, where with single TMS, stronger CMC was found when larger MEPs were triggered (Keil et al., 2014).

Repetitive transcranial magnetic stimulation (rTMS) refers to applying recurring TMS pulses. The neuromodulatory effects depend upon several stimulation parameters such as frequency, intensity, duration. Broadly, rTMS in high frequency (>1 Hz) increases the cortical excitability, and low frequency (<1 Hz) TMS depresses the cortical excitability. In stroke survivals, 1 Hz rTMS applied to unaffected hemisphere for 15 min led to reduction in CMC which further facilitating inter-hemispheric re-balance (L. Wang et al., 2024). Besides, for a short-term 15 min 1 Hz rTMS in stroke patients, it induced coherence changes beyond the stimulation side and inhibition of the coupling between motor cortex and effector muscles (Liu et al., 2022; Tan et al., 2022). For the long-term 15 min 1 Hz rTMS in after 15 sessions, rTMS strengthened the beta and gamma band EEG-EMG connectivity in stroke group whereas no such phenomenon was found in control group (Liu et al., 2024). Theta burst stimulation (TBS) is a special rTMS. By applying cTBS over left primary motor cortex, both CMC and cortico-cortical coherence increased, it may be attributed to improving the cortical excitability of the non-dominant hemisphere, since cTBS increased the concentration of inhibitory neurons such as gamma aminobutyric acid (GABA) in the stimulated cortex to induce cerebral cortex inhibition (Xu et al., 2024).

4.5. Ultrasound stimulation

Recently, transcranial ultrasound stimulation (TUS) showed effectiveness in oscillations neuromodulation. Compared to other non-invasive neuromodulation methods, TUS has higher spatial resolution and deeper penetration. The mechanisms of how TUS modulate neural activities are not clear, there are three main mechanisms, cavitation, temperature change and mechanical deformation (Tan, n.d.).

TUS has showed the oscillatory activities modulation in both healthy and stroke subjects. When applied TUS focally to S1 can attenuate the amplitude of somatosensory evoked potentials (SEPs) and gamma band oscillations induced by median nerve stimulation (Legon et al., 2014). The pulsed acoustic pressure waves dampen excitation or increase local interneuron firing and perhaps modulate the activity of fast-spiking interneurons and resulted in the oscillatory activities changes (Legon et al., 2014). Besides, the effects of S1 TUS relied on the phase state of SEP, suggested tFUS stimulation is uniquely able to modulate the phase activity of the beta frequency band of SEP components dependent upon spatial positioning (Mueller et al., 2014). In motor cortex, TUS caused effective suppression of motor corticospinal excitability and reduced response times during a visuomotor task. Systematically varying sonication parameters (showed that TUS suppresses MEPs more effectively at longer sonication durations (in a dose-dependent manner) and shorter duty cycles (Fomenko et al., 2020).

For cortiomuscular oscillations, under different ultrasound stimulation pulse parameters, specific enhancement of neural-muscular coupling oscillation strength between the mouse cortex and tail muscles can be achieved by TUS. More neurons are recruited during modulation with increasing stimulation duration (Xie et al., 2022). These modulations are related to ultrasound parameters and rhythm, while increase of intensity could induce the action potential of the neurons and then produce the motor response. Low-intensity pulsed ultrasound stimulation in motor cortex showed increasement in the LFP and tail muscle EMG phase synchronization in broad frequency band, suggest the LIPUS was capable of modulating descending motor outputs and provide the potential to influence motor function (Wang et al., 2019).

5. Limitation of non-invasive neuromodulation strategies

The manifestation of motor impairments caused by stroke injuries may appear similar, but there are significant differences in the cortico-muscular control patterns and recovery mechanisms. Previous research has proposed that the neural mechanisms of motor recovery after stroke are associated with the extent of injury. In patients with severe damage, motor recovery is related to the integrity of the corticospinal tract but not the corpus callosum, relying mainly on compensation from the unaffected hemisphere. In patients with mild damage, motor recovery is related to the integrity of the corpus callosum but not the corticospinal tract, relying primarily on the reconstruction of the perilesional areas of the affected hemisphere (Stewart et al., 2017). Studies have shown significant differences in connectivity and networks in the beta frequency band between cortical-subcortical and subcortical strokes during the subacute phase, with the relationship between low-frequency oscillations and behavioral assessments being opposite in the two types of strokes (Fanciullacci et al., 2021). Therefore, it is difficult to achieve effective functional recovery through generic neuromodulation techniques. In order to improve the efficiency of rehabilitation, personalized rehabilitation strategies need to be developed based on individual's residual neuro-muscular connectivity after stroke.

Currently, neuromodulation strategy conduct modulation by external information, neglecting the internal state. Previous research has utilized neurofeedback of EEG rhythms during task execution to enable individuals to modulate their activity and restore it to a "normal" state (Remsik et al., 2016). For example, reduced cortical activity and slowed brain rhythms in the affected hemisphere are major characteristics following stroke and are associated with motor impairments (Boyd et al., 2017; Kim and Winstein, 2017). In various motor rehabilitation studies, patients are instructed to upregulate the activity in specific frequency bands (such as alpha or beta) of the damaged primary motor cortex to improve upper limb motor function (Boyd et al., 2017; Mottaz et al., 2015). Besides, these neuromodulation strategies have shown the state-dependent effects. Therefore, it is necessary to assess the CMC during motor execution and develop individualized neuromodulation strategies based on the individual's cortico-muscular pattern to promote correct and precise targeted modulation.

6. Conclusions

Voluntary movement execution relies on the coordination between the central nervous system and peripheral muscles. Oscillatory activities enable efficient communication and information processing across brain regions and muscles. CMC has been observed in during movements with specific coherence patterns, while, stroke-induced motor deficits always correlate with abnormal CMC patterns. Yet how to modulate these patterns to facilitate particular movements is still poorly understood. Although neuromodulation techniques show promise in influencing CMC and potentially restoring motor function, more research is necessary to determine their effectiveness in modulating motor-related oscillations. Addressing these knowledge gaps is essential for advancing both theoretical insights and clinical approaches to motor recovery.

CRediT authorship contribution statement

Yuchen Xu: Writing – original draft. Shaomin Zhang: Writing – review & editing. Minmin Wang: Conceptualization. Mohamad Sawan: Writing – review & editing, Supervision, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

No data was used for the research described in the article.