A decoding filter for motor cortical (MI) neurons

A novel adaptation of an optimal filtering algorithm is used to reconstruct movement parameters from small populations of simultaneously recorded MI neurons in primates. The algorithm uses a doubly stochastic Poisson process driven by a Markov diffusion process to derive optimal and suboptimal filters to estimate arm trajectory during movement in a 2D plane. This algorithm significantly deviates from previous reconstruction techniques of arm trajectory in that it directly uses the action potential times rather than counts of the number of action potentials in small time intervals. Although tested in MI with cosine-tuned cells, this technique is readily adapted to stimulus reconstructions from other populations of cortical neurons with more general tuning curves