Characterizing the correlation between motor cortical neural firing and grasping kinematics

The hand has evolved to allow specialized interactions with our surroundings that define much of what makes us human. Comprised of numerous joints allowing 23 separate degrees-of-freedom (DoF) (joint motions) of movement, the hand and wrist are exceedingly complex. In order to better understand the constraints and principles underlying the neural control of the hand, we have carried out a series of neurophysiological experiments with monkeys performing a variety of reaching and grasping tasks. This work uses linear regression and low dimensional analysis to probe the neural representation of hand kinematics.We find that the kinematics of the three wrist DoFs (flexion, abduction and rotation) are rashly independent from hand-shape DoFs, and are considered separately. With respect to the wrist DoF, we show that the firing patterns of individual motor cortical cells are more linearly related to joint position than joint angular velocity. Using tuning functions from multivariate linear regressions, the firing rates from a population of cells accurately predicted three DoF of wrist orientation. We used principal components analysis to simplify the complex kinematics of the hand. Although the majority of the variability in hand kinematics can be explained with a small number (~7) of characteristic hand shapes (synergies), we find that these synergies do not capture the majority of neural variability. Both higher-order and lower-order synergies are well represented in the neural data. Although the kinematic synergies do not fully characterize neural firing, they can be utilized to simplify hand shape decoding. Using an optimal linear estimator, we predicted the average wrist and hand shape from the firing rates of 327 motor cortical cells with an accuracy as high as 92%. Individual motor cortical neurons are not well correlated with single joint variables; rather, they correlate with a number of joints in a complex way. This work provides evidence that hand movements are likely controlled through an intricate network of motor systems, of which motor cortical neurons contribute by making fine adjustments to a basic substrate. Further understanding of the control system will be gained by establishing a model that captures both the hand kinematic and neural variability