This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2014.2332353, IEEE Transactions on Neural Systems and Rehabilitation Engineering
Manuscript ID: TNSRE-2013-00314
1
Motor Imagery Learning Induced Changes in Functional Connectivity of the Default Mode Network Ruiyang Ge, Hang Zhang, Li Yao, and Zhiying Long Abstract—Numerous studies provide evidences that motor skill learning changes the activity of some brain regions during task as well as some resting networks during rest. However, it is still unclear how motor learning affects the resting-state default-mode network (DMN). Using functional magnetic resonance imaging (fMRI), this study investigated the alteration of the DMN after motor skill learning with mental imagery practice. 14 participants in the experimental group learned to imagine a sequential finger movement over a 2-week period while 12 control participants did not undergo motor imagery learning. For the experimental group, interregional connectivity, estimated by the graph theory method, between the medial temporal lobe, lateral temporal, and lateral parietal cortex within the DMN was increased after learning, whereas activity of the DMN network, estimated by the independent component analysis (ICA) method, remained stable. Moreover, the experimental group showed significant improvement in motor performance after learning and a negative correlation between the alteration of the execution rate and changes in activity in the lateral parietal cortex. These results indicate that the DMN could be sculpted by motor learning in a manner of altering interregional connectivity and may imply that the DMN plays a role in improving behavioral performance. Index Terms—functional Magnetic Resonance Imaging (fMRI), default mode network (DMN), motor learning, independent component analysis (ICA), graph theory.
Manuscript received November 21, 2013; revised April 21, 2014; accepted June 06, 2014. This work is supported by Key Program of National Natural Science Foundation of China (91320201), the National Natural Science Foundation of China (61271111), Fundamental Research Funds for the Central Universities (2012LYB04), the Funds for International Cooperation and Exchange of the National Natural Science Foundation of China (61210001) and the Scientific Research Foundation for the Returned Overseas Chinese Scholars, State Education Ministry. Z. Long is the corresponding author. R. Ge, and L. Yao are with the School of Information Science and Technology, Beijing Normal University, Beijing, 100875, China and the State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, 100875, China, and also with the Center for Collaboration and Innovation in Brain and Learning Sciences, Beijing Normal University, Beijing, 100875, China (e-mail: [email protected]; [email protected]). H. Zhang was with the State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, 100875, China. He is now with Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, 518055, China and the Center for Cognition and Brain Disorders and the Affiliated Hospital of Hangzhou Normal University, Hangzhou, 311121, China (e-mail: [email protected]). Z. Long is with the State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, 100875, China, and the Center for Collaboration and Innovation in Brain and Learning Sciences, Beijing Normal University, Beijing, 100875, China (e-mail: [email protected]).
I. INTRODUCTION
M
skill learning refers to the process by which movements are executed more smoothly and accurately with practice. It is thought to be a fundamental function of the complex central nervous system, play a fundamental role in our daily lives and be helpful to motor function rehabilitation [1]. Acquisition of motor skills as well as the improvement of subjects’ motor abilities could be accomplished through mental practice and physical practice [2][3]. Neuroimaging techniques have been widely utilized in characterizing the neural substrates mediating the incremental acquisition of skilled motor behaviors. It was demonstrated that motor learning could result in the structural plasticity in some brain regions, such as the supplementary motor areas, striatum, medial temporal lobe (MTL) and lateral parietal cortex (LPC) [4][5][6]. Using functional magnetic resonance imaging (fMRI) and positron emission tomography (PET) techniques, neuroimaging studies found motor learning contributed to functional changes in cortical regions including the primary motor cortex, dorsolateral prefrontal cortex, supplementary motor area, premotor cortex, posterior cingulate cortex (PCC) and LPC [2][7]. Some subcortical regions, such as striatum, cerebellum and MTL, also underwent functional alterations induced by motor skill learning [7][8]. Many previous brain imaging studies investigating motor skill learning revealed differentiated involvement of brain regions by comparing task states and control states [9]. However, it was pointed out that the brain is not idle at “rest” and spontaneous changes in regional neuronal firing occur even in a state of rest [10]. Consequently, task-independent measures of brain function may provide more insights into understanding the neural mechanism of motor learning. Resting-state fMRI is a popular tool to detect task-independent brain activities and spontaneous intrinsic neural activities generated by the brain [10]. Currently, resting-state measurements have been widely applied to clinical as well as non-clinical populations to explore the functional architecture of the human brain [11]. From an information theoretic view, human brain consists of multiple complex integrative networks in which information is continuously processed in these networks [12], and the interaction dynamics within these networks can be assessed through mutual functional influences. Several networks, including visual, auditory, motor, attention, self-referential, default-mode OTOR
1534-4320 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2014.2332353, IEEE Transactions on Neural Systems and Rehabilitation Engineering
Manuscript ID: TNSRE-2013-00314 network (DMN), have been identified in resting-state studies using time-course correlation approaches and independent component analysis (ICA) [11]. Recently, a few fMRI studies investigating the impact of motor skill learning on resting-state networks have found changes in fronto-parietal and cerebellar networks following visuo-motor tracking [13] and dynamic balancing training [14], changes in the motor network following motor sequence learning [15] and changes in the perceptual network following motor adaptation learning [16]. Although the above studies suggested that the motor and sensory loops involving cerebellum, striatum, motor cortex and somatosensory cortex during the resting-state can be changed by motor learning, it is unclear how the motor skill learning affect the DMN, a major contributor to the spontaneous neural activity that accounts for 60~80% of the brain’s energy consumption [17]. The DMN is an ensemble of the medial prefrontal cortex (MPFC), precuneus/posterior cingulate cortex (pC/PCC), LPC, lateral temporal cortex (LTC), and MTL [18][19]. From a functional point of view, DMN is a critical network of the resting human brain and plays a role in introspective function which linked to episodic memory, attention and working-memory [18] [20][21]. Activities within the DMN might reflect on-going off-line processing of information from prior experiences and learning [20][22] Functional imaging studies found that the DMN can be altered by maturing development [23], normal aging [24], cognitive impairment [25][26], meditation [27] as well as various training processes including working memory training [28], aerobic fitness training [29] and real-time neuro-feedback training [30]. It is noteworthy that the relevance of the DMN dysfunction has been emphasized in several neuropsychiatric diseases, such as Parkinson’s disease, autism and Alzheimer’s disease et al. [18]. Therefore, the inspection of the neural correlates underlying the improved motor performance by exploring the alteration of the DMN after the motor learning may be valuable to the motor rehabilitation training in clinical applications. Moreover, DMN has been demonstrated to be associated with mental imagery tasks. Two previous studies investigated modality-independent networks of mental imagery and found that both auditory and visual imagery yielded more DMN activity than the baseline condition [44][45]. In addition, it was reported that different imagery modalities, including mental imagery of shapes, sounds, touches, odours, flavours, self-perceived movements and internal sensations activated brain regions in the DMN [31]. These mental imagery studies suggested that the DMN acts as a supra-modal imagery network. Generally, motor learning can be carried out through mental imagery practice. Because DMN plays an important role in the mental imagery tasks, it is possible to alter DMN activity for motor learning through mental imagery practice. Currently, both graph theory analysis and ICA have been widely used to investigate brain networks [11][32]. Graph theory analysis method is specifically used to characterize the interactions among multiple brain regions and evaluate the information received by one particular brain region from other region/regions, thus it characterizes the brain connectivity from the perspective of topological organization [33]. The ICA
2 method considers brain voxels or regions as an integrated network and is powerful in extracting spatially distributed networks [11]. Therefore, combining the graph theoretical analysis and ICA method is capable of characterizing different properties of the brain networks. This study aimed to investigate the potential alterations of DMN induced by motor learning through mental imagery. By using ICA, the DMN was extracted from data and the alterations of activities in the DMN were examined. Furthermore, alterations of the functional connectivity between the putative regions of the DMN were investigated using the graph theory method. It has been demonstrated that the DMN is alterable through appropriate training and the motor learning can result in either structural or functional alterations of the MTL, PCC, and LPC that belong to the important nodes of the DMN. Moreover, DMN has been demonstrated to be a supra-modal imagery network. Thus, we hypothesized that alterations of the activity and functional connectivity of the DMN would be observed after motor imagery learning and some alterations of the DMN might be correlated with the improved behavior performance. II. METHODS A. Participants Twenty-six right-handed individuals participated with informed consent. The Institutional Review Board of the National Key Laboratory of Cognitive Neuroscience and Learning at Beijing Normal University approved our experimental protocols. To assess learning effects, participants were split into two groups: 14 individuals in the experimental group (7 men, mean age: 22 ± 2 years) with motor imagery learning and 12 in the control group (5 men, mean age: 24 ± 2 years) without motor imagery learning. None of the participants was a professional typist or musician and none reported any history of neurological and psychiatric problems. All participants passed Movement Imagery Questionnaire [34] and Vividness of Movement Imagery Questionnaires [35]. According to these questionnaires, the participants were requested to understand what kinetic imagery is, and to employ this imagery strategy during the experimental procedure. B. Experimental Procedures Briefly, the overall procedure of the experiment included (a) a familiarization stage, (b) a pre-learning fMRI scanning stage, (c) a 2-week motor imagery learning stage (experimental group) or a no-learning stage (control group), and (d) a post-learning fMRI scanning stage. In the familiarization stage, all participants were informed that the numbers 1-4 were to represent the digits of their right hands: index (1), middle (2), ring (3), and little (4) fingers. Next, they were instructed to tap a table using their right index finger in time with a metronome at 4 Hz for a 30-s epoch to learn the rhythm required in the following fMRI scanning. After that, they tapped the sequence 4-2-3-1-3-4-2 at 4 Hz for a 30-s epoch, and then imagined tapping the prescribed sequence at 4 Hz for a
1534-4320 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2014.2332353, IEEE Transactions on Neural Systems and Rehabilitation Engineering
Manuscript ID: TNSRE-2013-00314 30-s epoch. After finishing these exercises, the participants were prepared for the pre-learning fMRI scanning. There were 3 fMRI scanning runs in the pre-learning scanning stage. During the first resting-state run, all participants were instructed to remain still for 10 minutes, with their eyes closed. Thereafter, two task runs, including one each under motor execution and motor imagery, were completed. Each task run consisted of four 30-s epochs of either executing or imagining the motor sequence and alternated with five 30-s rest blocks. The two 4.5-min task runs (execution and imagery) were separated by a 5-min inter-run rest period. Assignment of scan orders was counterbalanced across subjects. In each task run, subjects were required to execute or imagine the sequence 4-2-3-1-3-4-2 with the right hand fingers at the same rate at which they practiced them outside the scanner when “PUSH” was displayed on the screen, and relaxed when “REST” was displayed on the screen. The sequential tapping was performed with a four-button response pad that was connected to a computer. Responses were recorded using the E-prime program (Psychology Software Tools, Pittsburgh, PA, USA). Immediately after the scanning, participants provided qualitative descriptions of performing the task. The contents of the qualitative description came from Movement Imagery Questionnaire [34] and included seven rating levels (1, Very Hard to feel; 2, Hard to feel; 3, Somewhat hard to feel; 4, Neutral; 5, Somewhat easy to feel; 6, Easy to feel; 7, Very easy to feel). No participants rated the scores lower than 5. In the two-week learning stage, a kinetic imagery task adapted from a sequential finger-tapping task [36] was utilized as a motor skill learning strategy. The experimental group practiced the task in 14 consecutive daily sessions. The participants in the experimental group were trained under the supervision of the experimenter, and their right hands were covered by a cardboard box to prevent visual feedback. The following instruction was provided to the participants in each learning session, “You will attend the motor imagery learning. Both the metronome-pacing and self-pacing learning were required to perform. During the 15-min metronome-pacing learning phase, you should imagine tapping 4-2-3-1-3-4-2 with your right hand fingers repeatedly as fast as the pace of the metronome. During the 15-min self-pacing learning, you should imagine tapping 4-2-3-1-3-4-2 with your right hand fingers repeatedly as fast as the pace controlled by yourself.” Each learning phase consisted of 15 30-s imagery practice blocks alternated with 15 30-s rest blocks. According to the mean tapping rate of the pre-learning fMRI scanning stage, participants were paced at 2 Hz in the first two practice sessions. This requirement was found to be important to ensure that participants could focus on establishing a representation of the sequence order [36]. From the third day onward, the frequency of pacing was increased to 4 Hz to encourage participants to improve the tapping rate. Moreover, an experimenter was required to continuously monitor the hand motion of the subjects to verify that the subjects only imaged during the whole imagery learning phase. Participants also provided the same qualitative description of performing after finishing each training session. During the entire learning period, no
3 participant rated the level of difficulty lower than 5. We further calculated the mean rating for each participant over 14 days and then checked the mean rating as well as the standard deviation over 14 participants; the results (14 participants, mean rating ± standard deviation: 5.9 ± 0.7) indicated that all participants performed the motor imagery properly. Participants in the control group did not attend any learning during the 14 days. Following the last learning session, all of the participants entered into the post-learning scanning stage. The procedure and instructions of post-learning scanning were identical to the pre-learning scanning. C. fMRI Data Acquisition Brain scans were performed at the MRI Center of Beijing Normal University using a 3.0-T Siemens whole-body MRI scanner. A single-shot T2-weighted gradient-echo, EPI sequence was used for functional imaging acquisition, with the following parameters: TR/TE/flip angle= 3000 ms/40 ms/90°, acquisition matrix = 64×64, field of view (FOV) = 240 mm, and slice thickness = 5 mm with no inter-slice gap. Thirty-two axial slices parallel to the AC-PC line were obtained in an interleaved order to cover the entire cerebrum and cerebellum. D. Behavioral Data Analysis Both the mean execution rate and mean execution error rate were calculated. To assess the training effect on behavioral performance, two separate ANOVAs were performed on the mean execution rate and mean execution error rate using State (pre-learning versus post-learning) as within-subjects factor and Group (experimental group versus control group) as between-subjects factor using SPSS (version 16.0, http://www.spss.com). E. Image Preprocessing Resting-state fMRI data were preprocessed using the Statistical Parametric Mapping software package (SPM8, http://www.fil.ion.ucl.ac.uk/spm). For each participant, the fMRI data were corrected for slice acquisition time and head motion. Functional images were then spatially normalized to a standard stereotactic space (EPI template provided by the Montreal Neurological Institute, MNI), resliced to a resolution of 3 × 3 × 4 mm, and spatially smoothed by a Gaussian kernel with a full width at half maximum of 8 mm. Finally, the images were detrended and temporally filtered by band-pass filter (0.01 Hz < f < 0.08 Hz) to remove linear trends and high-frequency noise using REST (http://www.restfmri.net/forum/REST). F. Independent Component Analysis (ICA) According to the spatial ICA model proposed by McKeown et al. [37], the fMRI data matrix can be modeled as the summation of independent components multiplied with the mixing matrix : . Here is a matrix consisting of fMRI raw data. T is the number of time points in the experiments and P is the number of voxels over the brain volumes. is the matrix consisting of spatially independent components.
1534-4320 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2014.2332353, IEEE Transactions on Neural Systems and Rehabilitation Engineering
Manuscript ID: TNSRE-2013-00314 Each row of the matrix,
, is a spatially independent source. K
Fig.1. Results of behavioral data. (a) Mean execution rate of the pre-learning and post-learning execution task for the two groups. (b) Mean execution error rate of the pre-learning and post-learning execution task for the two groups. Exp represents the experimental group and Con represents the control group. Error bars indicate standard error values. ** indicates p
Manuscript ID: TNSRE-2013-00314
1
Motor Imagery Learning Induced Changes in Functional Connectivity of the Default Mode Network Ruiyang Ge, Hang Zhang, Li Yao, and Zhiying Long Abstract—Numerous studies provide evidences that motor skill learning changes the activity of some brain regions during task as well as some resting networks during rest. However, it is still unclear how motor learning affects the resting-state default-mode network (DMN). Using functional magnetic resonance imaging (fMRI), this study investigated the alteration of the DMN after motor skill learning with mental imagery practice. 14 participants in the experimental group learned to imagine a sequential finger movement over a 2-week period while 12 control participants did not undergo motor imagery learning. For the experimental group, interregional connectivity, estimated by the graph theory method, between the medial temporal lobe, lateral temporal, and lateral parietal cortex within the DMN was increased after learning, whereas activity of the DMN network, estimated by the independent component analysis (ICA) method, remained stable. Moreover, the experimental group showed significant improvement in motor performance after learning and a negative correlation between the alteration of the execution rate and changes in activity in the lateral parietal cortex. These results indicate that the DMN could be sculpted by motor learning in a manner of altering interregional connectivity and may imply that the DMN plays a role in improving behavioral performance. Index Terms—functional Magnetic Resonance Imaging (fMRI), default mode network (DMN), motor learning, independent component analysis (ICA), graph theory.
Manuscript received November 21, 2013; revised April 21, 2014; accepted June 06, 2014. This work is supported by Key Program of National Natural Science Foundation of China (91320201), the National Natural Science Foundation of China (61271111), Fundamental Research Funds for the Central Universities (2012LYB04), the Funds for International Cooperation and Exchange of the National Natural Science Foundation of China (61210001) and the Scientific Research Foundation for the Returned Overseas Chinese Scholars, State Education Ministry. Z. Long is the corresponding author. R. Ge, and L. Yao are with the School of Information Science and Technology, Beijing Normal University, Beijing, 100875, China and the State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, 100875, China, and also with the Center for Collaboration and Innovation in Brain and Learning Sciences, Beijing Normal University, Beijing, 100875, China (e-mail: [email protected]; [email protected]). H. Zhang was with the State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, 100875, China. He is now with Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, 518055, China and the Center for Cognition and Brain Disorders and the Affiliated Hospital of Hangzhou Normal University, Hangzhou, 311121, China (e-mail: [email protected]). Z. Long is with the State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, 100875, China, and the Center for Collaboration and Innovation in Brain and Learning Sciences, Beijing Normal University, Beijing, 100875, China (e-mail: [email protected]).
I. INTRODUCTION
M
skill learning refers to the process by which movements are executed more smoothly and accurately with practice. It is thought to be a fundamental function of the complex central nervous system, play a fundamental role in our daily lives and be helpful to motor function rehabilitation [1]. Acquisition of motor skills as well as the improvement of subjects’ motor abilities could be accomplished through mental practice and physical practice [2][3]. Neuroimaging techniques have been widely utilized in characterizing the neural substrates mediating the incremental acquisition of skilled motor behaviors. It was demonstrated that motor learning could result in the structural plasticity in some brain regions, such as the supplementary motor areas, striatum, medial temporal lobe (MTL) and lateral parietal cortex (LPC) [4][5][6]. Using functional magnetic resonance imaging (fMRI) and positron emission tomography (PET) techniques, neuroimaging studies found motor learning contributed to functional changes in cortical regions including the primary motor cortex, dorsolateral prefrontal cortex, supplementary motor area, premotor cortex, posterior cingulate cortex (PCC) and LPC [2][7]. Some subcortical regions, such as striatum, cerebellum and MTL, also underwent functional alterations induced by motor skill learning [7][8]. Many previous brain imaging studies investigating motor skill learning revealed differentiated involvement of brain regions by comparing task states and control states [9]. However, it was pointed out that the brain is not idle at “rest” and spontaneous changes in regional neuronal firing occur even in a state of rest [10]. Consequently, task-independent measures of brain function may provide more insights into understanding the neural mechanism of motor learning. Resting-state fMRI is a popular tool to detect task-independent brain activities and spontaneous intrinsic neural activities generated by the brain [10]. Currently, resting-state measurements have been widely applied to clinical as well as non-clinical populations to explore the functional architecture of the human brain [11]. From an information theoretic view, human brain consists of multiple complex integrative networks in which information is continuously processed in these networks [12], and the interaction dynamics within these networks can be assessed through mutual functional influences. Several networks, including visual, auditory, motor, attention, self-referential, default-mode OTOR
1534-4320 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2014.2332353, IEEE Transactions on Neural Systems and Rehabilitation Engineering
Manuscript ID: TNSRE-2013-00314 network (DMN), have been identified in resting-state studies using time-course correlation approaches and independent component analysis (ICA) [11]. Recently, a few fMRI studies investigating the impact of motor skill learning on resting-state networks have found changes in fronto-parietal and cerebellar networks following visuo-motor tracking [13] and dynamic balancing training [14], changes in the motor network following motor sequence learning [15] and changes in the perceptual network following motor adaptation learning [16]. Although the above studies suggested that the motor and sensory loops involving cerebellum, striatum, motor cortex and somatosensory cortex during the resting-state can be changed by motor learning, it is unclear how the motor skill learning affect the DMN, a major contributor to the spontaneous neural activity that accounts for 60~80% of the brain’s energy consumption [17]. The DMN is an ensemble of the medial prefrontal cortex (MPFC), precuneus/posterior cingulate cortex (pC/PCC), LPC, lateral temporal cortex (LTC), and MTL [18][19]. From a functional point of view, DMN is a critical network of the resting human brain and plays a role in introspective function which linked to episodic memory, attention and working-memory [18] [20][21]. Activities within the DMN might reflect on-going off-line processing of information from prior experiences and learning [20][22] Functional imaging studies found that the DMN can be altered by maturing development [23], normal aging [24], cognitive impairment [25][26], meditation [27] as well as various training processes including working memory training [28], aerobic fitness training [29] and real-time neuro-feedback training [30]. It is noteworthy that the relevance of the DMN dysfunction has been emphasized in several neuropsychiatric diseases, such as Parkinson’s disease, autism and Alzheimer’s disease et al. [18]. Therefore, the inspection of the neural correlates underlying the improved motor performance by exploring the alteration of the DMN after the motor learning may be valuable to the motor rehabilitation training in clinical applications. Moreover, DMN has been demonstrated to be associated with mental imagery tasks. Two previous studies investigated modality-independent networks of mental imagery and found that both auditory and visual imagery yielded more DMN activity than the baseline condition [44][45]. In addition, it was reported that different imagery modalities, including mental imagery of shapes, sounds, touches, odours, flavours, self-perceived movements and internal sensations activated brain regions in the DMN [31]. These mental imagery studies suggested that the DMN acts as a supra-modal imagery network. Generally, motor learning can be carried out through mental imagery practice. Because DMN plays an important role in the mental imagery tasks, it is possible to alter DMN activity for motor learning through mental imagery practice. Currently, both graph theory analysis and ICA have been widely used to investigate brain networks [11][32]. Graph theory analysis method is specifically used to characterize the interactions among multiple brain regions and evaluate the information received by one particular brain region from other region/regions, thus it characterizes the brain connectivity from the perspective of topological organization [33]. The ICA
2 method considers brain voxels or regions as an integrated network and is powerful in extracting spatially distributed networks [11]. Therefore, combining the graph theoretical analysis and ICA method is capable of characterizing different properties of the brain networks. This study aimed to investigate the potential alterations of DMN induced by motor learning through mental imagery. By using ICA, the DMN was extracted from data and the alterations of activities in the DMN were examined. Furthermore, alterations of the functional connectivity between the putative regions of the DMN were investigated using the graph theory method. It has been demonstrated that the DMN is alterable through appropriate training and the motor learning can result in either structural or functional alterations of the MTL, PCC, and LPC that belong to the important nodes of the DMN. Moreover, DMN has been demonstrated to be a supra-modal imagery network. Thus, we hypothesized that alterations of the activity and functional connectivity of the DMN would be observed after motor imagery learning and some alterations of the DMN might be correlated with the improved behavior performance. II. METHODS A. Participants Twenty-six right-handed individuals participated with informed consent. The Institutional Review Board of the National Key Laboratory of Cognitive Neuroscience and Learning at Beijing Normal University approved our experimental protocols. To assess learning effects, participants were split into two groups: 14 individuals in the experimental group (7 men, mean age: 22 ± 2 years) with motor imagery learning and 12 in the control group (5 men, mean age: 24 ± 2 years) without motor imagery learning. None of the participants was a professional typist or musician and none reported any history of neurological and psychiatric problems. All participants passed Movement Imagery Questionnaire [34] and Vividness of Movement Imagery Questionnaires [35]. According to these questionnaires, the participants were requested to understand what kinetic imagery is, and to employ this imagery strategy during the experimental procedure. B. Experimental Procedures Briefly, the overall procedure of the experiment included (a) a familiarization stage, (b) a pre-learning fMRI scanning stage, (c) a 2-week motor imagery learning stage (experimental group) or a no-learning stage (control group), and (d) a post-learning fMRI scanning stage. In the familiarization stage, all participants were informed that the numbers 1-4 were to represent the digits of their right hands: index (1), middle (2), ring (3), and little (4) fingers. Next, they were instructed to tap a table using their right index finger in time with a metronome at 4 Hz for a 30-s epoch to learn the rhythm required in the following fMRI scanning. After that, they tapped the sequence 4-2-3-1-3-4-2 at 4 Hz for a 30-s epoch, and then imagined tapping the prescribed sequence at 4 Hz for a
1534-4320 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2014.2332353, IEEE Transactions on Neural Systems and Rehabilitation Engineering
Manuscript ID: TNSRE-2013-00314 30-s epoch. After finishing these exercises, the participants were prepared for the pre-learning fMRI scanning. There were 3 fMRI scanning runs in the pre-learning scanning stage. During the first resting-state run, all participants were instructed to remain still for 10 minutes, with their eyes closed. Thereafter, two task runs, including one each under motor execution and motor imagery, were completed. Each task run consisted of four 30-s epochs of either executing or imagining the motor sequence and alternated with five 30-s rest blocks. The two 4.5-min task runs (execution and imagery) were separated by a 5-min inter-run rest period. Assignment of scan orders was counterbalanced across subjects. In each task run, subjects were required to execute or imagine the sequence 4-2-3-1-3-4-2 with the right hand fingers at the same rate at which they practiced them outside the scanner when “PUSH” was displayed on the screen, and relaxed when “REST” was displayed on the screen. The sequential tapping was performed with a four-button response pad that was connected to a computer. Responses were recorded using the E-prime program (Psychology Software Tools, Pittsburgh, PA, USA). Immediately after the scanning, participants provided qualitative descriptions of performing the task. The contents of the qualitative description came from Movement Imagery Questionnaire [34] and included seven rating levels (1, Very Hard to feel; 2, Hard to feel; 3, Somewhat hard to feel; 4, Neutral; 5, Somewhat easy to feel; 6, Easy to feel; 7, Very easy to feel). No participants rated the scores lower than 5. In the two-week learning stage, a kinetic imagery task adapted from a sequential finger-tapping task [36] was utilized as a motor skill learning strategy. The experimental group practiced the task in 14 consecutive daily sessions. The participants in the experimental group were trained under the supervision of the experimenter, and their right hands were covered by a cardboard box to prevent visual feedback. The following instruction was provided to the participants in each learning session, “You will attend the motor imagery learning. Both the metronome-pacing and self-pacing learning were required to perform. During the 15-min metronome-pacing learning phase, you should imagine tapping 4-2-3-1-3-4-2 with your right hand fingers repeatedly as fast as the pace of the metronome. During the 15-min self-pacing learning, you should imagine tapping 4-2-3-1-3-4-2 with your right hand fingers repeatedly as fast as the pace controlled by yourself.” Each learning phase consisted of 15 30-s imagery practice blocks alternated with 15 30-s rest blocks. According to the mean tapping rate of the pre-learning fMRI scanning stage, participants were paced at 2 Hz in the first two practice sessions. This requirement was found to be important to ensure that participants could focus on establishing a representation of the sequence order [36]. From the third day onward, the frequency of pacing was increased to 4 Hz to encourage participants to improve the tapping rate. Moreover, an experimenter was required to continuously monitor the hand motion of the subjects to verify that the subjects only imaged during the whole imagery learning phase. Participants also provided the same qualitative description of performing after finishing each training session. During the entire learning period, no
3 participant rated the level of difficulty lower than 5. We further calculated the mean rating for each participant over 14 days and then checked the mean rating as well as the standard deviation over 14 participants; the results (14 participants, mean rating ± standard deviation: 5.9 ± 0.7) indicated that all participants performed the motor imagery properly. Participants in the control group did not attend any learning during the 14 days. Following the last learning session, all of the participants entered into the post-learning scanning stage. The procedure and instructions of post-learning scanning were identical to the pre-learning scanning. C. fMRI Data Acquisition Brain scans were performed at the MRI Center of Beijing Normal University using a 3.0-T Siemens whole-body MRI scanner. A single-shot T2-weighted gradient-echo, EPI sequence was used for functional imaging acquisition, with the following parameters: TR/TE/flip angle= 3000 ms/40 ms/90°, acquisition matrix = 64×64, field of view (FOV) = 240 mm, and slice thickness = 5 mm with no inter-slice gap. Thirty-two axial slices parallel to the AC-PC line were obtained in an interleaved order to cover the entire cerebrum and cerebellum. D. Behavioral Data Analysis Both the mean execution rate and mean execution error rate were calculated. To assess the training effect on behavioral performance, two separate ANOVAs were performed on the mean execution rate and mean execution error rate using State (pre-learning versus post-learning) as within-subjects factor and Group (experimental group versus control group) as between-subjects factor using SPSS (version 16.0, http://www.spss.com). E. Image Preprocessing Resting-state fMRI data were preprocessed using the Statistical Parametric Mapping software package (SPM8, http://www.fil.ion.ucl.ac.uk/spm). For each participant, the fMRI data were corrected for slice acquisition time and head motion. Functional images were then spatially normalized to a standard stereotactic space (EPI template provided by the Montreal Neurological Institute, MNI), resliced to a resolution of 3 × 3 × 4 mm, and spatially smoothed by a Gaussian kernel with a full width at half maximum of 8 mm. Finally, the images were detrended and temporally filtered by band-pass filter (0.01 Hz < f < 0.08 Hz) to remove linear trends and high-frequency noise using REST (http://www.restfmri.net/forum/REST). F. Independent Component Analysis (ICA) According to the spatial ICA model proposed by McKeown et al. [37], the fMRI data matrix can be modeled as the summation of independent components multiplied with the mixing matrix : . Here is a matrix consisting of fMRI raw data. T is the number of time points in the experiments and P is the number of voxels over the brain volumes. is the matrix consisting of spatially independent components.
1534-4320 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNSRE.2014.2332353, IEEE Transactions on Neural Systems and Rehabilitation Engineering
Manuscript ID: TNSRE-2013-00314 Each row of the matrix,
, is a spatially independent source. K
Fig.1. Results of behavioral data. (a) Mean execution rate of the pre-learning and post-learning execution task for the two groups. (b) Mean execution error rate of the pre-learning and post-learning execution task for the two groups. Exp represents the experimental group and Con represents the control group. Error bars indicate standard error values. ** indicates p
