Delsys Prize 2006 Winning Proposal
Probability-Based Prediction Of Electromyographic Activity In Multiple Muscles
Dr. Andrew Fuglivand, Departments of Physiology and Neurobiology, College of Medicine, University of Arizona, Tuscon, AZ, USA.
A probability-based method is proposed to predict the patterns of electromyographic activity across multiple muscles during a wide range of movements. A reasonable correspondence between predicted and actual EMG signals is demonstrated with this method. Such an approach ultimately might provide a flexible means to control functional electrical stimulation and thereby expand the repertoire of motor functions available to paralyzed individuals.
Functional electrical stimulation (FES) involves artificial activation of muscles with implanted electrodes to restore motor function in paralyzed individuals. The range of motor behaviors that can be generated by FES, however, is limited to a small set of preprogrammed movements such as hand grasp and release. A broader range of movements has not been implemented because of the substantial difficulty associated with identifying the patterns of muscle stimulation needed to elicit specified movements. Most limb movements require intricate coordination among multiple muscles that act across several joints. Such complex mechanical systems do not readily lend themselves to deterministic solutions. Although electromyographic (EMG) signals recorded in able-bodied subjects can be used to identify patterns of muscle activity associated with a particular movement, this painstaking method yields control signals appropriate only for the motor task from which the EMG signals were originally recorded. In an attempt to overcome this limitation in controlling FES systems, we have developed a probabilistic method to predict patterns of EMG activity associated with, in theory, an unlimited set of movements involving multiple joints. The general approach involves two main stages. In the first stage, EMG and kinematic signals are recorded while a subject makes a wide variety of limb movements. These signals are then used as inputs to a computer algorithm to characterize the relationship between muscle activity and kinematics using a probabilistic method based on Bayes' theorem. In the second stage, the probabilistic relationship between muscle activity and kinematics identified in the first stage is used to predict muscle activity associated with a new set of desired movements in other subjects. We have used this method to estimate the patterns of muscle activity associated with a wide range of desired movements of a single finger (Seifert and Fuglevand, 2002). More recently, we found that it was possible to estimate accurately complex patterns of activity in 12 upper-arm muscles based on hand trajectory information only (Figure 1). In these experiments, the RMS error between predicted and actual EMG averaged < 10% of the peak EMG across all subjects, muscles, and movements. This reasonable correspondence suggests that such an approach could be used to predict patterns of muscle stimulation needed to produce a wide array of movements using FES systems given the desired trajectory of the hand such as might be obtained from brain-derived signals (e.g. Wessberg et al. 2000).
Figure 1: Example traces of actual (blue) and predicted (red) EMG signals for two muscles (teres major, top; anterior deltoid, bottom) associated with a complex hand trajectory.
Seifert HM , Fuglevand AJ (2002): Restoration of movement using functional electrical stimulation and Bayes' theorem. J Neurosci 22: 9465-9474.
Wessberg J, Stambaugh CR, Kralik JD, Beck PD, Laubach M, Chapin JK, Kim J, Biggs SJ, Srinivasan MA, Nicolelis MAL (2000): Real-time prediction of hand trajectory by ensembles of cortical neurons in primates. Nature 408: 361-365.