| Abstract No.: | B-D2154 |
| Country: | Canada |
| | |
| Title: | PATHLETS IN UPPER LIMB MUSCLES? |
| | |
| Authors/Affiliations: | 1 Timothy Lillicrap*; 1 Andrew Pruszynski; 1 Stephen Scott;
1 Queen's University, Kingston, ON, Canada
|
| | |
| Content: | Objective: Many neurophysiological studies assume that individual primary motor cortical (M1) neurons encode time-invariant movement features such as hand velocity or joint torque. Recent studies have demonstrated that M1 neurons are better correlated with temporally extensive features of movement (Hatsopoulos et al., 2007). These correlations are taken as evidence that the activity of a neuron encodes/causes a given time-dependent trajectory, termed a "pathlet". The objective here was to determine whether upper limb muscles also correlate with temporally extensive movement features such as hand velocity.
Materials and Methods: Human subjects made planar reaches to visual targets which appeared at random locations in the workspace. Simultaneously, surface EMG was recorded from six upper limb muscles. For each muscle, we generated a set of candidate pathlet models which related hand velocity to the EMG of that muscle. The number of parameters in these models was held constant while their temporal windows (i.e. duration and offset) were varied. Models were then selected by examining their performance on the second half of the collected data, which was not used for fitting.
Results: Both model selection and fit were robust and suggest that upper limb muscles are best correlated with temporally extensive features of movement. Models with similar temporal windows were selected for all muscles: approximately 300 ms in duration and including both past and future kinematics. Furthermore, the model fits for each muscle exhibited spatial trajectories that were appropriate for the muscle's mechanical action. For example, the shoulder flexor deltoid anterior was always found to be most active for movements which were initially towards the body and then moved away from it. Both the temporal and spatial features were extremely reliable across subjects and recording sessions.
Conclusions: Our results demonstrate that muscle activity is correlated with hand velocity pathlets though it clearly does not encode hand velocity. Therefore, similar observations in M1 should be interpreted with caution; afterall, how could one discern whether M1 neurons explicitly encode hand velocity pathlets or whether they merely encode low-level time-invariant representations (e.g. muscle activity) that are filtered by musculoskeletal dynamics? |
| | |
| Back |
|
