Edited by: Linda Douw, VU University Medical Center, Netherlands
Reviewed by: Fahad Sultan, University Tübingen, Germany; Behnam Molavi, University of British Columbia, Canada
*Correspondence: Andrew J. Butler, Department of Physical Therapy, Byrdine F. Lewis School of Nursing and Health Professions, Georgia State University, Urban Life Building, Suite 819, Atlanta, GA 30303, USA e-mail:
This article was submitted to the journal Frontiers in Systems Neuroscience.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Coherent network oscillations (<0.1 Hz) linking distributed brain regions are commonly observed in the brain during both rest and task conditions. What oscillatory network exists and how network oscillations change in connectivity strength, frequency and direction when going from rest to explicit task are topics of recent inquiry. Here, we study network oscillations within the sensorimotor regions of able-bodied individuals using hemodynamic activity as measured by functional near-infrared spectroscopy (fNIRS). Using spectral interdependency methods, we examined how the supplementary motor area (SMA), the left premotor cortex (LPMC) and the left primary motor cortex (LM1) are bound as a network during extended resting state (RS) and between-tasks resting state (btRS), and how the activity of the network changes as participants execute left, right, and bilateral hand (LH, RH, and BH) finger movements. We found: (i) power, coherence and Granger causality (GC) spectra had significant peaks within the frequency band (0.01–0.04 Hz) during RS whereas the peaks shifted to a bit higher frequency range (0.04–0.08 Hz) during btRS and finger movement tasks, (ii) there was significant bidirectional connectivity between all the nodes during RS and unidirectional connectivity from the LM1 to SMA and LM1 to LPMC during btRS, and (iii) the connections from SMA to LM1 and from LPMC to LM1 were significantly modulated in LH, RH, and BH finger movements relative to btRS. The unidirectional connectivity from SMA to LM1 just before the actual task changed to the bidirectional connectivity during LH and BH finger movement. The uni-directionality could be associated with movement suppression and the bi-directionality with preparation, sensorimotor update and controlled execution. These results underscore that fNIRS is an effective tool for monitoring spectral signatures of brain activity, which may serve as an important precursor before monitoring the recovery progress following brain injury.
Based on converging electrophysiological and neuroimaging data, the brain is known to be a self-organizing dynamical system consisting of anatomically distinct and efficiently connected brain regions supporting inherent electrical, chemical, hemodynamic, and metabolic processes (Buzsaki,
fNIRS technology uses specific wavelengths of light in the near-infrared range between 700 and 1000 nm, irradiated through the scalp, to enable the non-invasive measurement of changes in the relative ratios of deoxygenated hemoglobin (deoxy-Hb) and oxygenated hemoglobin (oxy-Hb) following neuronal activity in the brain (Arno and Britton,
Previous studies have focused on brain functional and effective connectivity during resting state (RS), motor imagery (MI), and motor execution (ME) tasks using fMRI and EEG. Most of these studies compare the connection strengths and directionality between MI and ME tasks and demonstrate modulation during task execution (Solodkin et al.,
Reports of network activity in unilateral and bilateral hand (BH) movements have found that during right hand movements, left SMA and left premotor cortex (LPMC) promote activity in left primary motor cortex (LM1) positively whereas networks are modulated negatively toward right M1 (Grefkes et al.,
In the current study, we used near-infrared spectroscopy (NIRS), a relatively new technique as compared to other neuroimaging modalities, to investigate the directed functional connectivity using GC approach during rest as well as during ME. NIRS can only be used to measure activities on cortical surfaces whereas fMRI can be used to measure activations throughout the whole brain. However, NIRS is a non-invasive, safe, cost effective and more flexible and portable technique with reasonable spatial and excellent temporal resolution compared to fMRI. NIRS also allows monitoring children as well as patients who are psychologically unfit to be studied under traditional neuroimaging methods.
Considering the above-mentioned advantages of NIRS over fMRI, our present study explores and compares the cortical network dynamics during RS, between task resting state (btRS) and ME tasks using the parametric GC approach (Granger,
Twenty-seven able-bodied adult volunteers participated in the study (8 males; 19 females; age 22–63, mean = 31.8 ± 12.8 years). People were excluded from the study if they: (a) had any medical conditions that could interfere with ability to complete questionnaires and visual-motor tasks, (b) had any unstable medical conditions, (c) were unable to attend both testing sessions, or (d) were taking any medications which could influence motor or visual ability. All subjects provided written informed consent and procedures were reviewed and approved by the local Institutional Review Board (IRB).
During recordings, participants were seated comfortably in an adjustable chair facing a computer screen approximately 24 inches away. Data was collected in two 1-h sessions separated by 7 days using the Hitachi ETG-4000 52-channel NIRS system (Hitachi Medical Co., Tokyo Japan). The absorption of near infrared light at two wavelengths (695 and 830 nm) was measured with a sampling rate of 10 Hz. Changes in reduced (deoxy) hemoglobin (HbR), and oxy-Hb (HbO) concentrations at each time point from each channel were computed using the modified Beer-Lambert law (Cope et al.,
During testing, cap position was adjusted to make sure that the hair bundles were not blocking sensors and detectors, a strong evoked activity occurred over sensorimotor regions for ME and adequate signals were obtained from all the channels (Figure
The position of the cap and ROIs were determined by calculating
During each fNIRS recording session, participants underwent a 7 min RS followed by a 9 min 30 s motor task. Each participant read a standardized set of task instructions before each recording to ensure normalization of procedures within and between sessions. The RS occurred prior to the motor task for all participants at each session in order to control for possible confounding effects of task performance (Fransson,
Three ROIs were defined anatomically as follows: the M1 as the area extending from the anterior bank of the central sulcus to the anterior edge of the precentral gyrus (Dassonville et al.,
Each ROI was assigned a position within the international 10–20-electrode system, which corresponded to its anatomical location based on previous research and neuroanatomical atlases (Jasper,
For channel selection, the three closest channels to the international 10–20 system locations were calculated mathematically for each participant and selected to represent each ROI. The channels closest to ROIs were the same for every subject for M1 (channels 4, 8, and 9), SMA (channels 2, 6, and 11), and PM (channels 13, 17, and 18) (Figure
Raw fNIRS data were linearly detrended, band-pass filtered between 0.01 and 0.1 Hz. The detrending and filtering removed slow trends and other physiological noise such as respiration and cardiac activities.
In the resting condition, participants visualized a fixation cross and the recordings form the baseline (extended RS). The second condition was the task condition, which had fNIRS data for: (i) resting period prior to each task period, i.e., btRS and (ii) during task periods, i.e., during ME. During ME, time series of whole length (9 min 30 s) from each node and run was broken into segments each of length 30 sec. These segments were grouped together separately for different conditions: btRS, RH, LH, and BH finger tapping. Each movement condition was repeated three times and rest condition occurred for a total of 10 times. These repetitions were treated as trials for the spectral analysis. For each condition, concentration changes for oxy-Hb, deoxy-HB, and total-Hb were calculated. We used only oxy-Hb signal changes in current analysis since previous studies showed no significant correlation between changes in deoxy-Hb and ability to imagine movements measured by MIQ-R (Kober et al.,
Spectral interdependency measures are a means of statistically quantifying the inter-relationship between a pair of oscillatory processes; say 1 and 2, as a function of frequency of oscillation. In practice, there are three measures that characterize the spectral interdependency between a pair of processes: total interdependence (
The measures of spectral interdependency are derived from a spectral density matrix (S), which is constructed from the time series of oscillatory systems by using optimal autoregressive (AR) modeling in the parametric method.
Diagonal elements of the spectral density matrix (S) represented node activity in terms of spectral power (P), whereas the coherence function
The coherence function is a well-accepted measure to characterize frequency-specific interdependence between multiple time series from multisite recordings such as multi-electrode electrophysiological recordings, electro/magneto encephalography (E/MEG) and functional magnetic resonance imaging (fMRI). It ranges from 0 (no interdependence) to 1 (maximum interdependence).
Directional influences between processes 1 and 2 are given by (Geweke,
Here, ∑ (noise covariance matrix), H (transfer function matrix),
GC values were integrated over the frequency range from 0.01 Hz (
Significant connections for each condition (RS, btRS, RH, LH and BH finger tapping) were found using permutation test at
Power, coherence and GC spectra for all the nodes (SMA, LM1, and LPMC), which were found to be involved during RS, btRS and task execution, were computed. Figures
During the RS, for all the three nodes, the peaks for power were in the frequency band 0.01–0.04 Hz (Figure
Directionality of causal flow among the three nodes, SMA, LM1, and LPMC was computed during RS. A bidirectional causal flow was observed among all the three nodes (Figures
Similarly, directionality of causal flow among all the three nodes was computed and compared among btRS condition, RH, LH, and BH finger movement.
Considering RS condition (condition 1) as reference 1 (Figure
For reference 2 vs. LH, we found all the connections were significantly different. There was modulation of 75% and 46% from SMA to LM1 and LM1 to SMA respectively. Further, there was modulation of 68% and 61% from LPMC to SMA and LPMC to LM1 respectively (Figure
For reference 2 vs. BH, all the connections were significantly different except from LPMC to LM1, which was, 39% modulated. There was 90 and 88% modulation from LM1 to SMA and SMA to LM1 respectively and 78 and 72% modulation from LPMC to SMA and SMA to LPMC respectively (Figure
For reference 2 vs. RH, there was only one connection SMA to LM1, which was significantly different and was 42% modulated. There was 50 and 26% modulation for significant connection from LPMC to LM1 and LM1 to LPMC respectively (Figure
We have observed differences in oscillatory motor network activity within the sensorimotor regions of able-bodied individuals during rest and finger movement using the hemodynamic activity as measured by fNIRS. During RS, there were significant node and network oscillations in the frequency band 0.01–0.04 Hz. These GC results obtained from the parametric approach showed that there were significant bidirectional connections among LM1, LPMC, and SMA during RS, from LM1 to SMA and LPMC and between SMA and LPMC during btRS. There were significant modulations between connections during ME task in comparison to btRS, especially, from SMA to LM1 during RH finger movement and bidirectional significant modulations between LM1 and SMA during LH and BH finger movements. We found significant positive modulations from LPMC to LM1 under all the conditions whereas LPMC to SMA during LH and bidirectional positive modulations between these during BH finger movement.
Our findings confirm and extend previous findings that showed significant network activity changes in going from RS to movement (Jiang et al.,
Dominant components of LFO below 0.1 Hz is found using Fourier analysis of temporal components of oxy-hemoglobin signal in an NIRS study (Tong and Frederick,
Comparing the connection strengths and directionality between RS and btRS conditions, there are significant bidirectional connections between SMA and LPMC although weaker in case of btRS. Furthermore, there is unidirectional significant causal flow from LM1 to SMA during btRS in comparison to significant bidirectional causal flow between them in RS. Although, participants are not directly instructed to imagine motor task in between two ME tasks, our results are consistent with MI conditions where participants are usually instructed to imagine some hand or finger movements (Kasess et al.,
The connection from LPMC to LM1 was found to be significantly stronger in LH finger tapping in comparison to RS. This connection is also significantly stronger in all three conditions (LH, BH, and RH) when compared to btRS. During execution of task, whether unimanual or bimanual, both SMA and LPMC were shown to have a significant influence on LM1. The extent of modulation appears to depend on the condition: LH, RH, or bimanual hand finger tapping movements. Our results are in accord with previous studies. Both SMA and LPMC are known to be involved in movement selection and execution of movements, especially PMC which is thought to be involved in the execution of triggered movements and the transformation of external stimuli to motor planning (Lutz et al.,
We compared the connection strengths and directionality between SMA and LM1 for btRS and ME. We observed that this connection is suppressed during preparation to execute the task. Hence if planning, preparation and then execution of motor task are considered in the same domain, we find a closed-loop circuit between both of these areas where there is feed-forward influence from LM1 to SMA during preparation and feed-backward influence from SMA to LM1 during actual execution of task.
Previous studies compare the effective connectivity between MI and ME using other approaches like DCM and SEM and find similar results suggesting positive influence of task resulting in strong activation of M1 and a suppressive influence during MI but strong enough to keep M1 active to execute the task (Solodkin et al.,
In this study, we used the bivariate version of the Geweke's spectral decomposition-based GC (Geweke,
In conclusion, results of the present work show (i) power, coherence and GC spectra had peaks within the frequency band (0.01–0.04 Hz) during RS whereas the peaks shifted to a higher frequency range (0.04–0.08 Hz) during btRS and finger movement tasks, (ii) there was significant bidirectional connectivity between all the nodes during RS and unidirectional connectivity from the LM1 to SMA and LM1 to LPMC during btRS, and (iii) the connections from SMA to LM1 and from LPMC to LM1 were significantly modulated in LH, RH, and BH finger movements relative to btRS. These results are consistent with the other studies, which used fMRI and EEG techniques. This provides us confidence that NIRS can be effectively used for monitoring slow hemodynamic fluctuations and underlying brain functional connectivity during rest and task. These data serve as a foundation for studies to follow comparing the characteristics of motor and RS networks between healthy subjects and people with neurologic insult such as stroke. Analysis of brain network in patients with lesions may lead to an effective approach to determine the functional and structural damage to cortical network connections, which may better inform us as we develop clinical recovery pathway for these clients.
Mukesh Dhamala, Andrew J. Butler, Daniel Drake, and Sahil Bajaj wrote manuscript. Sahil Bajaj, Daniel Drake and Mukesh Dhamala analyzed data. Mukesh Dhamala, Andrew J. Butler, Daniel Drake, and Sahil Bajaj provided concept/idea/research design and project management.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
A US NSF CAREER Award (BCS 0955037) supported Mukesh Dhamala. A U.S. Department of Veterans Affairs Award (5I21RX000561) supported Daniel Drake and Andrew J. Butler. The authors thank J. Rajendra for his assistance in data collection and pre-processing.