Back in 2012, Churchland and his colleagues proposed that “rotational dynamics”, uncovered through linear transformations of multidimensional neuronal data, represent a fundamental type of neuronal population processing in a variety of organisms, from the isolated leech central nervous system to the primate motor cortex. Here, we evaluated this claim using Churchland’s own data and simple simulations of neuronal responses. We observed that rotational patterns occurred in neuronal populations when (1) there was a temporal sequence in peak firing rates exhibited by individual neurons, and (2) this sequence remained consistent across different experimental conditions. Provided that such a temporal order of peak firing rates existed, rotational patterns could be easily obtained using a rather arbitrary computer simulation of neural activity; modeling of any realistic properties of motor cortical responses was not needed. Additionally, arbitrary traces, such as Lissajous curves, could be easily obtained from Churchland’s data with multiple linear regression. While these observations suggest that temporal sequences of neuronal responses could be visualized as rotations with various methods, we express doubt about Churchland et al.’s bold assessment that such rotations are related to “an unexpected yet surprisingly simple structure in the population response”, which “explains many of the confusing features of individual neural responses”. Instead, we argue that their approach provides little, if any, insight on the underlying neuronal mechanisms employed by neuronal ensembles to encode motor behaviors in any species.

Although many real-time neural decoding algorithms have been proposed for brain-machine interface (BMI) applications over the years, an optimal, consensual approach remains elusive. Recent advances in deep learning algorithms provide new opportunities for improving the design of BMI decoders, including the use of recurrent artificial neural networks to decode neuronal ensemble activity in real time. Here, we developed a long-short term memory (LSTM) decoder for extracting movement kinematics from the activity of large ( N = 134–402) populations of neurons, sampled simultaneously from multiple cortical areas, in rhesus monkeys performing motor tasks. Recorded regions included primary motor, dorsal premotor, supplementary motor, and primary somatosensory cortical areas. The LSTM's capacity to retain information for extended periods of time enabled accurate decoding for tasks that required both movements and periods of immobility. Our LSTM algorithm significantly outperformed the state-of-the-art unscented Kalman filter when applied to three tasks: center-out arm reaching, bimanual reaching, and bipedal walking on a treadmill. Notably, LSTM units exhibited a variety of well-known physiological features of cortical neuronal activity, such as directional tuning and neuronal dynamics across task epochs. LSTM modeled several key physiological attributes of cortical circuits involved in motor tasks. These findings suggest that LSTM-based approaches could yield a better algorithm strategy for neuroprostheses that employ BMIs to restore movement in severely disabled patients.

No abstract