Elsevier

NeuroImage

Volume 40, Issue 2, 1 April 2008, Pages 828-837
NeuroImage

The suppressive influence of SMA on M1 in motor imagery revealed by fMRI and dynamic causal modeling

https://doi.org/10.1016/j.neuroimage.2007.11.040Get rights and content

Abstract

Although motor imagery is widely used for motor learning in rehabilitation and sports training, the underlying mechanisms are still poorly understood. Based on fMRI data sets acquired with very high temporal resolution (300 ms) under motor execution and imagery conditions, we utilized Dynamic Causal Modeling (DCM) to determine effective connectivity measures between supplementary motor area (SMA) and primary motor cortex (M1). A set of 28 models was tested in a Bayesian framework and the by-far best-performing model revealed a strong suppressive influence of the motor imagery condition on the forward connection between SMA and M1. Our results clearly indicate that the lack of activation in M1 during motor imagery is caused by suppression from the SMA. These results highlight the importance of the SMA not only for the preparation and execution of intended movements, but also for suppressing movements that are represented in the motor system but not to be performed.

Introduction

Motor imagery is of significant clinical importance as there is strong evidence that it is beneficial in motor learning (Yaguez et al., 1998, Yue and Cole, 1992) and is therefore commonly used in sports training (Atienza et al., 1998, Brouziyne and Molinaro, 2005) as well as motor rehabilitation following stroke (Sharma et al., 2006). Accordingly, neuronal representations of motor execution (ME) and motor imagery (MI) have been studied extensively using brain imaging methods including functional magnetic resonance imaging (fMRI), positron emission tomography (PET) and electrophysiological measurements including magnetoencephalography (MEG) and subdural recordings (for a short review, see, e.g. Dechent et al., 2004; Lotze et al., 1999).

Overall, it has been shown that ME and MI activate similar (Jeannerod, 1994) or at least partially overlapping networks (Gerardin et al., 2000). Some areas of the brain have consistently been shown to be involved in both tasks, most notably the pre-motor cortex (PMC) and the supplementary motor area (SMA), whereas data on the primary motor cortex (M1) during motor imagery are still controversial. Some fMRI studies have shown consistent activation of M1 during MI (Lotze et al., 1999, Porro et al., 2000, Porro et al., 1996) while others have shown either a transient activation of M1 at the beginning of the task (Dechent et al., 2004) or no participation of M1 at all (Binkofski et al., 2000, Hanakawa et al., 2003).

Since the majority of studies of MI have focused on the extent of brain activation, information on motor system interaction is still limited. A promising approach for characterizing neuronal networks is based on “effective connectivity” between network components, defined as the influence of one neural system over another (Friston et al., 1995). Assessment of effective connectivity measures provides the unique opportunity to determine if and how activity in premotor and supplementary motor areas influences activity in primary motor cortex during ME and MI. Connectivities between brain regions are estimated using statistical models based on anatomically motivated assumptions related to the basic structure of the network (Penny et al., 2004b, Ramnani et al., 2004). Effective connectivity methods are thus strongly hypothesis driven. Several methods to analyze effective connectivity in fMRI have been proposed, most prominently Structural Equation Modeling (SEM) and, more recently, Dynamic Causal Modeling (DCM). Originating in econometrics, the first applications of SEM to neuroimaging data have been published in the early 1990s (MacIntosh and Gonzalez-Lima, 1991, McIntosh and Gonzalez-Lima, 1994). With regard to SEM, the BOLD response of a region at a given time is considered a weighted sum of the BOLD response of other regions at the same time. Hence, a source region induces an instantaneous correlation in a target region with the strength of the correlation determined by the weighting factor.

Contrary to SEM, the more recent approach of DCM distinguishes between activity at the neuronal and activation at the BOLD level (Friston et al., 2003). Most importantly, interactions between brain regions are limited to the neuronal level and each region generates a BOLD signal depending only on its own activity. The BOLD signal is generated from the neuronal activity using a biophysical model for the hemodynamic response (Friston et al., 2000) based on the Balloon model (Buxton and Frank, 1997, Buxton et al., 1998).

In DCM, connectivity values represent time scales at which regions interact, with higher values indicating shorter interaction times and hence stronger connections. This is in stark contrast to SEM where connection implies instantaneous correlation between two regions. In addition to interactions between regions, DCM allows for external stimuli to influence regions, either directly or by modifying the connections between regions. Hence, DCM does not require a partitioning of the acquired data according to the context (e.g. motor execution and imagery) as is the case with standard SEM. Different conditions are reflected by modified inter-regional connections caused by some external factor.

DCM has been shown to yield results comparable to SEM when applied to the same set of data (Penny et al., 2004b). Additionally, the more sophisticated DCM has already been successfully applied to a number of different neuronal networks, e.g. face perception (Fairhall and Ishai, 2007) or functionality of callosal connections (Stephan et al., 2005), and was thus selected for the study presented here.

So far, effective connectivity methods have rarely been applied to the human motor system. A single study estimated networks for motor execution, visual and kinesthetic motor imagery using SEM (Solodkin et al., 2004). During kinesthetic MI, subjects “mentally simulate the movements associated with a kinesthetic feeling of the movement” (Solodkin et al., 2004), whereas in visual MI, a visual representation of the movement is formed. Visual MI is therefore sometimes referred to as a third-person process (Dechent et al., 2004). Conventional analysis revealed largely overlapping areas of activation, especially for ME and kinesthetic MI (Solodkin et al., 2004). However, the connectivity strengths within the two networks were different. Among other changes, the connection from SMA to M1 acted enhancing during ME but strongly suppressive during MI.

Here, we specifically aimed to investigate the interactions in the human motor system during ME and kinesthetic MI using very high temporal resolution fMRI to appropriately capture the dynamics of the hemodynamic response. Of particular importance was the question whether SMA-M1 connection strength is suppressed during motor imagery as might be expected from recent studies in patients with SMA lesions following stroke (Sumner et al., 2007) and consistent with the previous study of Solodkin et al. (2004). This study extends on the previous study of Solodkin et al. (2004) in two important ways. Firstly, using DCM methods, we are able to examine effective connectivity within the motor system using an event-related fMRI paradigm to separate preparation and execution phases of the motor task. By using high temporal resolution fMRI, we examined the precise time course of the hemodynamic response during the preparation and readiness phase and during the execution or imagery of the finger sequence movement, consistent with our previous studies (Cunnington et al., 2002, Cunnington et al., 2003). Secondly, we incorporated full feedback loops into the DCM models to examine whether motor execution or imagery involves any feedback from M1 to the SMA, as well as the expected input from the SMA to M1.

As data quality and sensitivity are key issues in modeling, fMRI data subjects were scanned at 3 T. A set of 28 different models connecting SMA and M1 with stimulus-related inputs was constructed and tested for the best performing model based on a method using Bayes factors (Penny et al., 2004a).

Section snippets

Subjects

Eight young healthy subjects (four male, mean age 26 years) without known history of neurological, psychiatric or movement disorder participated in this study and gave written informed consent prior to the experiment. All were right-handed according to the Edinburgh inventory (Oldfield, 1971) and the study was approved by the local ethics committee.

Experimental paradigm

Measurements were performed on a 3 T Medspec scanner (Bruker Biospin, Germany) using gradient-recalled EPI. Four axial slices with a thickness of

Results

Mean ROI volumes were 10.1 ± 3.8 and 11.3 ± 2.4 cm3 for M1 and SMA ROIs, respectively. MNI coordinates of mean M1 and SMA ROI barycenters were − 36, − 28, 63 and 0, − 5, 60, respectively. SMA-proper quota across subjects was 75 ± 12% of SMA ROI.

Averaging M1 and SMA time courses across all trials and subjects showed clear differences in M1 during MI and ME (Fig. 4). While M1 was strongly active during ME, it showed only subtle activation during MI. In contrast, SMA was comparably active during both ME

Discussion

We successfully applied DCM to data on imagined and executed actions and were able to show that the connection from SMA to M1 suppresses M1 activity during MI. Previous studies reported that strong activation of the primary motor cortex during ME was significantly decreased during motor imagery (e.g. Lotze et al., 1999; Porro et al., 2000, 1996). Our results expand on these findings and demonstrate that this reported decrease in M1 activity is caused by the SMA actively suppressing M1 activity

Acknowledgments

This study was financially supported by the Austrian Science Fund (FWFP-16669-B02) and the Austrian National Bank (OeNB P10943 and P11903)

References (48)

  • K.F. Muakkassa et al.

    Frontal lobe inputs to primate motor cortex: evidence for four somatotopically organized ‘premotor’ areas

    Brain Res.

    (1979)
  • P. Nachev et al.

    The role of the pre-supplementary motor area in the control of action

    NeuroImage

    (2007)
  • R. Oldfield

    The assessment and analysis of handedness: the Edinburgh inventory

    Neuropsychologia

    (1971)
  • D.N. Pandya et al.

    Intra- and interhemispheric projections of the precentral, premotor and arcuate areas in the rhesus monkey

    Brain Res.

    (1971)
  • W. Penny et al.

    Comparing dynamic causal models

    NeuroImage

    (2004)
  • W. Penny et al.

    Modelling functional integration: a comparison of structural equation and dynamic causal models

    NeuroImage

    (2004)
  • N. Ramnani et al.

    New approaches for exploring anatomical and functional connectivity in the human brain

    Biol. Psychiatry

    (2004)
  • P. Sumner et al.

    Human medial frontal cortex mediates unconscious inhibition of voluntary action

    Neuron

    (2007)
  • L. Yaguez et al.

    A mental route to motor learning: improving trajectorial kinematics through imagery training

    Behav. Brain Res.

    (1998)
  • H. Akaike

    Information theory and an extension of the maximum likelihood principle

  • G.E. Alexander et al.

    Parallel organization of functionally segregated circuits linking basal ganglia and cortex

    Annu. Rev. Neurosci.

    (1986)
  • F. Atienza et al.

    Video modeling and imaging training on performance of tennis service of 9- to 12-year-old children

    Percept. Mot. Skills

    (1998)
  • F. Binkofski et al.

    Broca's region subserves imagery of motion: a combined cytoarchitectonic and fMRI study

    Hum. Brain Mapp.

    (2000)
  • M. Brouziyne et al.

    Mental imagery combined with physical practice of approach shots for golf beginners

    Percept. Mot. Skills

    (2005)
  • Cited by (216)

    View all citing articles on Scopus
    1

    Contributed equally.

    View full text