Decoding of muscle activity from the sensorimotor cortex in freely behaving monkeys

Abstract

Remarkable advances have recently been made in the development of Brain-Machine Interface (BMI) technologies for restoring or enhancing motor function. However, the application of these technologies may be limited to patients in static conditions, as these developments have been largely based on studies of animals (e.g., non-human primates) in constrained movement conditions. The ultimate goal of BMI technology is to enable individuals to move their bodies naturally or control external devices without physical constraints. Here, we demonstrate accurate decoding of muscle activity from electrocorticogram (ECoG) signals in unrestrained, freely behaving monkeys. We recorded ECoG signals from the sensorimotor cortex as well as electromyogram signals from multiple muscles in the upper arm while monkeys performed two types of movements with no physical restraints, as follows: forced forelimb movement (lever-pull task) and natural whole-body movement (free movement within the cage). As in previous reports using restrained monkeys, we confirmed that muscle activity during forced forelimb movement was accurately predicted from simultaneously recorded ECoG data. More importantly, we demonstrated that accurate prediction of muscle activity from ECoG data was possible in monkeys performing natural whole-body movement. We found that high-gamma activity in the primary motor cortex primarily contributed to the prediction of muscle activity during natural whole-body movement as well as forced forelimb movement. In contrast, the contribution of high-gamma activity in the premotor and primary somatosensory cortices was significantly larger during natural whole-body movement. Thus, activity in a larger area of the sensorimotor cortex was needed to predict muscle activity during natural whole-body movement. Furthermore, decoding models obtained from forced forelimb movement could not be generalized to natural whole-body movement, which suggests that decoders should be built individually and according to different behavior types. These results contribute to the future application of BMI systems in unrestrained individuals.

Introduction

Significant progress has been made in the development of Brain-Machine Interfaces (BMIs) over the past two decades (Lebedev and Nicolelis, 2017; Moxon and Foffani, 2015). This progress relies heavily on the development of new methodology in the field of neural decoding, in which neuronal activities are translated into signals that can be used to control external devices such as limb prostheses or computers (Lebedev et al., 2008). Currently, movement-related information such as the kinematic parameters of motion, forces, and muscle activity can be accurately decoded from the neuronal activities of behaving animals (Georgopoulos and Carpenter, 2015; Schwartz, 2016). However, previous neural decoding studies using non-human primates have generally been conducted under unnatural conditions, in which most body parts are physically restrained, and only a small subset of movements, such as arm reaching and hand grasping, are studied. Ultimately, the goal of BMI technology is implementation in a non-restrained condition to enable severely disabled individuals to move naturally and freely.

Animal studies have indicated that neuronal coding of limb movement in unconstrained conditions is different from that in constrained conditions. For example, the tuning properties of single neurons in the primary motor cortex (M1) have been reported to vary among movements performed in different working spaces (Aflalo and Graziano, 2006; Caminiti et al., 1990; Griffin et al., 2015) or between movements performed using focal muscles vs. the whole body (Jackson et al., 2007). Thus, it is not clear whether insights obtained from recordings conducted under constrained conditions can be directly applied to the decoding of movement parameters from cortical neuronal activity recorded during more natural movements.

Recent advances in the development of a head-mounted wireless device have demonstrated the possibility of recording neuronal activity in freely-moving animals (Borton et al., 2013; Chestek et al., 2009; Fernandez-Leon et al., 2015; Mavoori et al., 2005; Schwarz et al., 2014). For example, several studies have wirelessly recorded neuronal action potentials from the sensorimotor cortex of rhesus macaques using multi-electrode arrays during arm reaching towards a reward or walking on a treadmill (Capogrosso et al., 2016; Foster et al., 2012, 2014; Rajangam et al., 2016; Schwarz et al., 2014; Yin et al., 2014). In these studies, the modes of motor behavior (e.g., walking or reaching), as well as the specific epoch in each motor behavior (e.g. the flexor or extensor phase during locomotion), were decoded from an ensemble of spiking activities. However, the researchers only examined a small subset of behaviors using specific tasks. Since a hallmark of non-human primates is the rich variety of motor behavior that forms their daily movements (Jaman and Huffman, 2008), neural coding of their natural behaviors can be more informative by recording neuronal signals during free movements in their home cage.

In the present study, we recorded electrocorticograms (ECoGs) from the sensorimotor cortex and electromyograms (EMGs) from forelimb muscles in common marmosets. We used a wireless recording system to record activity as the marmosets moved freely in a cage. The common marmoset is a New World primate species that is considered to be useful as a research subject in the field of neuroscience (Okano et al., 2012; Prins et al., 2014; Walker et al., 2017). The marmoset has several advantages in terms of recording cortical activity during free movement, as follows: 1) as- in humans and rhesus macaques, the cortical sheet in the marmoset is divided into functionally distinct cortical regions (Rosa and Tweedale, 2005); 2) the structure of the marmoset cortex is lissencephalic, which is advantageous for obtaining neuronal signals from the entire cortex using a two-dimensional grid electrode (Komatsu et al., 2017; Newman et al., 2009; Tia et al., 2017); 3) marmosets are much smaller than rhesus macaques, and are thus easier to handle (Prins et al., 2017); 4) the marmoset shows a wide variety of movements in three-dimensional space, such as walking, jumping, and grasping a pole located in the cage (Wang et al., 2014). Thus, we consider the common marmoset to be optimal for our research.

Here, we demonstrated that accurate prediction of EMG from ECoG is possible during natural whole-body movement of marmosets freely moving in a large space, as well as during forced forelimb movement. While high-gamma activity in the M1 contributed most strongly to the prediction of muscle activity in both movement types, a larger area of the sensorimotor cortex contributed to decoding for natural whole-body movement than for forced forelimb movement. These results provide insights that are relevant to future applications of BMI systems in unrestrained individuals.