Single-trial anticipatory beta activity predicts the maintenance of efficient motor plans
Poster No:
2071
Submission Type:
Abstract Submission
Authors:
Martina Bracco1, Varsha Vesudevan1, Quentin Welniarz1, Mihoby Razafinimanana2, Sabine Meunier1, Antoni Valero-Cabré1, Traian Popa3, Denis Schwartz4, Cécile Gallea1
Institutions:
1Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière, Paris, France, 2Sorbonne Université, CNRS, Inserm, IBPS, Neurosciences Paris Seine, CeZaMe Lab, Paris, France, 3Lausanne University Hospital (CHUV), Department of Clinical Neurosciences, Lausanne, Switzerland, 4CERMEP‐Imagerie du Vivant, MEG Departement, Lyon, Ile-de-France
First Author:
Martina Bracco
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Co-Author(s):
Varsha Vesudevan
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Quentin Welniarz
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Mihoby Razafinimanana
Sorbonne Université, CNRS, Inserm, IBPS, Neurosciences Paris Seine, CeZaMe Lab
Paris, France
Sorbonne Université, CNRS, Inserm, IBPS, Neurosciences Paris Seine, CeZaMe Lab
Paris, France
Sabine Meunier
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Antoni Valero-Cabré
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Traian Popa
Lausanne University Hospital (CHUV), Department of Clinical Neurosciences
Lausanne, Switzerland
Lausanne University Hospital (CHUV), Department of Clinical Neurosciences
Lausanne, Switzerland
Cécile Gallea
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Sorbonne Université, Paris Brain Institute, Inserm, CNRS, APHP, Hôpital de la Pitié Salpêtrière
Paris, France
Introduction:
Originally considered an 'idling rhythm' associated with a mere lack of movement, pre-movement beta oscillations may instead carry critical information related to anticipation of forthcoming actions. They may reflect error prediction based on past movement outcomes and retention of efficient motor plans (Engel & Fries, 2010; Torrecillos et al., 2015). We tested this idea by recording participants' oscillatory activity with magnetoencephalography (MEG) while manipulating trial-to-trial movement-execution errors in a non-ballistic, target-reaching task.
Methods:
Seventeen healthy volunteers (mean age 46±13) performed 12 blocks of 50 consecutive trials of a reaching task with their right (dominant) hand, under normal (0°) or rotated visual feedback (25° or -30° rotation). Normal visual feedback is associated with small trajectory errors and high predictability of movement outcomes. Rotated feedback induces substantial errors, gradually decreasing as consecutive trials lead to a predictable mismatch between the expected and the actual feedback (Shadmehr et al., 2010). Participants' brain activity was recorded via a 306-sensors MEG system (Megin Triux System).
We explored whether pre-movement beta activity was influenced by (1) the error context (normal or rotated visual feedback), and (2) the error history across trials within the same block. For error history, we divided MEG data into Early and Late stages, as the association between predicted and actual outcomes is gradually refined across repetitions. We assessed statistical differences across conditions using a 2-by-2 non-parametric cluster-based statistics analysis ((Maris & Oostenveld, 2007); Error context: Rotation vs No Rotation; Error History: Early vs. Late stage) across all 204 gradiometers, time points before movement onset (-2.5 to 0 s), and frequencies within the beta band (14-26 Hz). We used the Dynamical Imaging of Coherent Sources (DICS) approach to pinpoint possible significant sources of beta activity (Gross et al., 2001).
Importantly, we then investigated if single-trial beta activity predicted the trial-by-trial motor outcomes (Shin et al., 2017; Fig. 1a). Specifically, we used an ARMAX (AutoRegressive, MovingAverage, Exogenous) model to predict the evolution of error amplitude across 50 trials, using beta activity as an exogenous predictor. We fixed the ARMAX architecture based on the coefficients obtained from the population for all rotation conditions. We then tested the model's predictability through 'leave-one-out' cross-validation and Monte Carlo simulations (1000 repetitions; Fig. 1b).
We explored whether pre-movement beta activity was influenced by (1) the error context (normal or rotated visual feedback), and (2) the error history across trials within the same block. For error history, we divided MEG data into Early and Late stages, as the association between predicted and actual outcomes is gradually refined across repetitions. We assessed statistical differences across conditions using a 2-by-2 non-parametric cluster-based statistics analysis ((Maris & Oostenveld, 2007); Error context: Rotation vs No Rotation; Error History: Early vs. Late stage) across all 204 gradiometers, time points before movement onset (-2.5 to 0 s), and frequencies within the beta band (14-26 Hz). We used the Dynamical Imaging of Coherent Sources (DICS) approach to pinpoint possible significant sources of beta activity (Gross et al., 2001).
Importantly, we then investigated if single-trial beta activity predicted the trial-by-trial motor outcomes (Shin et al., 2017; Fig. 1a). Specifically, we used an ARMAX (AutoRegressive, MovingAverage, Exogenous) model to predict the evolution of error amplitude across 50 trials, using beta activity as an exogenous predictor. We fixed the ARMAX architecture based on the coefficients obtained from the population for all rotation conditions. We then tested the model's predictability through 'leave-one-out' cross-validation and Monte Carlo simulations (1000 repetitions; Fig. 1b).
Results:
We found an increase of pre-movement beta activity in the Late compared to the Early stage of the task (Main effect of Error History, p<0.001; Fig. 2). This increase was lower in the Rotation than in the No Rotation condition (Interaction: Error Context x Error History; p=0.014; Fig. 2). Source estimates revealed a network including the parietal cortices bilaterally, the left prefrontal cortex, and the right cerebellum (Fig. 2).
Furthermore, the ARMAX (1,0,1) model showed that pre-movement beta amplitude could predict movement outcomes based on trial-by-trial errors across all 17 participants and rotations (normalized prediction mean square error – NPMSE for 0°: 0.31±0.6; 25°: 0.17±0.22; -30°: 0.26±0.13). The cross-validation revealed stronger beta contribution in blocks with higher predictability of motor outcomes (0° vs 25° rotation, p<0.001; 25° vs 30° rotation, p<0.001; Fig. 1c).
Furthermore, the ARMAX (1,0,1) model showed that pre-movement beta amplitude could predict movement outcomes based on trial-by-trial errors across all 17 participants and rotations (normalized prediction mean square error – NPMSE for 0°: 0.31±0.6; 25°: 0.17±0.22; -30°: 0.26±0.13). The cross-validation revealed stronger beta contribution in blocks with higher predictability of motor outcomes (0° vs 25° rotation, p<0.001; 25° vs 30° rotation, p<0.001; Fig. 1c).
Conclusions:
Our findings support the role of anticipatory beta activity in maintaining efficient motor plans, particularly in contexts with stable and low motor-execution errors (Engel & Fries, 2010; Torrecillos et al., 2015). These results provide novel insights into (1) the dynamics of this process, with changes of beta power predicting trial-by-trial motor execution outcomes and (2) the functional neuro-anatomical network supporting this anticipatory process that involves the cerebellum.
Learning and Memory:
Learning and Memory Other
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI)
EEG/MEG Modeling and Analysis
Motor Behavior:
Motor Planning and Execution 1
Visuo-Motor Functions 2
Keywords:
Cerebellum
MEG
Modeling
Motor
1|2Indicates the priority used for review
Provide references using author date format
Gross, J. (2001). Dynamic imaging of coherent sources: Studying neural interactions in the human brain. PNAS, 98(2), 694–699.
Maris, E. (2007). Nonparametric statistical testing of EEG- and MEG-data. Journal of Neuroscience Methods, 164(1), 177–190.
Shadmehr, R. (2010). Error Correction, Sensory Prediction, and Adaptation in Motor Control, 33(1), 89-108.
Shin, H. (2017). The rate of transient beta frequency events predicts behavior across tasks and species. ELife, 6.
Torrecillos, F. (2015). Behavioral/Cognitive Distinct Modulations in Sensorimotor Postmovement and Foreperiod-Band Activities Related to Error Salience Processing and Sensorimotor Adaptation, 35 (37) 12753-12765.