Gao, Y.
Bienenstock, E.
Black, M.
Shoham, S.
Serruya, M.
Donoghue, J.
Statistical learning and probabilistic inference methods were used to (i) investigate the nature of encoding in motor cortex, (ii) characterize probabilistic relationships between arm kinematics (hand position or velocity) and activity of a simultaneously recorded neural population, and (iii) optimally reconstruct (decode) hand trajectory from population activity. Data was obtained from simultaneous multiple electrode recordings of 25 neurons in the arm area of primary motor cortex (MI) while subjects manually tracked a smoothly and randomly moving visual target (Paninski et al., this volume). Statistical learning methods were used to derive optimal Bayesian estimates of the conditional probability of firing for each cell given the kinematic variables. Non-parametric models of conditional firing were learned using regularization (smoothing) techniques with cross-validation and suggest that the cells encode information about the position and velocity of the hand. Decoding involves the inference of the kinematics from the firing rate of the cells. The posterior probability distribution of the kinematics given the spike data was represented with discrete samples. Predictions of hand parameters were estimated in 50msec intervals using a Bayesian estimation method called particle filtering. Experiments with real and synthetic data suggested that this approach provides probabilistically sound estimates of kinematics and allows the probabilistic combination of information from multiple neurons, the use of priors, and the rigorous evaluation of models and results.
Supported by: NIH #R01 NS25074 & #N01-NS-9-2322
Related Publications
Gao, Y., Black, M. J., Bienenstock, E., Shoham, S., Donoghue, J., Probabilistic inference of arm motion from neural activity in motor cortex, Advances in Neural Information Processing Systems 14, The MIT Press, 2002. (postscript), (pdf).
Gao, Y., Bienenstock, E., Black, M., Shoham, S., Serruya, M., Donoghue, J., Encoding/decoding of arm kinematics from simultaneously recorded MI neurons, Society for Neuroscience Abst., Vol. 27, Program No. 572.14 2001.
Home