Neural activity is spatially clustered in motor and dorsal premotor cortex

Abstract

Motor (M1) and dorsal premotor (PMd) cortex are central to arm and hand control in primates. Motor outputs in both areas confer somatotopically organized arm and hand zones. Here, we investigate the spatial mapping between those motor zones and movement-related neural activity to gain insight about functional organization in M1/PMd. Two macaques reached and grasped while cortical activity was measured with intrinsic signal optical imaging. Activity maps were quantified in relation to microstimulation motor maps from the same hemispheres. Each activity map was comprised of many patches and overlapped surprisingly small portions of the motor map. Functional differences between the patches were inferred from their activity time courses and location within the motor map. We propose that M1/PMd contain subzones that are preferentially tuned for specific actions. Thus, the spatial dimension of neural activity in frontal motor areas is an important organizing principle of the neural code for movement control.

A Spiking Neural Network with Continuous Local Learning for Robust Online Brain Machine Interface

“…For this study, we focus on the task of predicting 2D velocity of an arm controlling a computer cursor. This task was selected because it is one of the most widely studied BMI tasks [10,38] with several public data sets containing neural and kinematic data [30,31]. Cursor control has the practical application of enabling BMI users to navigate a computer, granting them greater autonomy.…”

“…Evaluation Data: The SNN was trained and tested on two different data sets containing intracortical neural recordings and corresponding kinematic data from non-human primates performing reaching tasks [30,31,42,43]. The first data set, taken from the Collaborative Research in Computational Neuroscience (CRNCS) dataset repository, contained recordings from the premotor cortex of a monkey as it moved a cursor towards an indicated 2 cm by 2 cm target on a 20 cm by 20 cm computer screen.…”

“…There are two main aims of this work: first, to show that the continuous learning SNN decoder can achieve comparable accuracy to existing methods in an offline decoding setting with reduced memory consumption; and second, to show that the continuous learning SNN is capable of quickly adapting to sudden changes in neural behavior and can also learn entirely online with no prior training. To address the first aim, we train the continuous learning algorithm on two different datasets consisting of neural and kinematic data from nonhuman primates performing reaching tasks [30,31]. To address the second aim, we train the continous learning SNN in a simulated closed loop environment and evaluate its performance after disruptions to the neural inputs.…”

Motor somatotopy impacts imagery strategy success in human intracortical brain-computer interfaces

“…Studies in non-human primates [25][26][27][28] and humans 3,29 have shown that activity recorded from a single electrode can be related to movements of multiple body parts, particularly across the arm and hand during coordinated actions. When using cortical stimulation in conjunction with intrinsic signal optical imaging, one group found that movement representations within the upper limb region were highly distributed and overlapping, but with local regions exhibiting more dominant responses to certain body parts 30 . Further, one recent study using intracortical recordings in humans measured significant neural activity associated with speaking in an area of the precentral gyrus that was thought to be specific to hand actions, significantly challenging the idea of a rigid somatotopic organization.…”