Prior Knowledge of Target Direction and Intended Movement Selection Improves Indirect Reaching Movement Decoding

Robust neural decoding by kernel regression with Siamese representation learning

Objective. Brain-machine interfaces (BMIs) provide a direct pathway between the brain and external devices such as computer cursors and prosthetics, which have great potential in motor function restoration. One critical limitation of current BMI systems is the unstable performance, partly due to the variability of neural signals. Studies showed that neural activities exhibit trial-to-trial variability, and the preferred direction of neurons frequently changes under different conditions. Therefore, a fixed decoding function does not work well.Approach. To deal with the problems, we propose a novel kernel regression framework. The nonparametric kernel regression is used to fit diverse decoding functions by finding similar neural patterns to handle neural variations caused by varying tuning functions. Further, the representations of raw neural signals are learned by Siamese networks and constrained by kinematic parameters, which can alleviate neural variations caused by intrinsic noises and task-irrelevant information. The representations are jointly learned with the kernel regression framework in an end-to-end manner so that neural variations can be tackled effectively.Main results. Experiments on two datasets demonstrate that our approach outperforms most existing methods and significantly improves the robustness in challenging situations such as limited samples and missing channels.Significance. The proposed approach demonstrates robust performance with different conditions and provides a new and inspiring perspective toward robust BMI control.

An automated behavioral apparatus to combine parameterized reaching and grasping movements in 3D space

BACKGROUND

The neural principles underlying reaching and grasping movements have been studied extensively in primates for decades. However, few experimental apparatuses have been developed to enable a flexible combination of reaching and grasping in one task in three-dimensional (3D) space.

NEW METHOD

By combining a custom turning table with a 3D translational device, we have developed a highly flexible apparatus that enables the subject to reach multiple positions in 3D space, and grasp differently shaped objects with multiple grip types in each position. Meanwhile, hand trajectory and grip types can be recorded via optical motion tracking cameras and touch sensors, respectively.

RESULTS

We have used the apparatus to successfully train a macaque monkey to accomplish a visually-guided reach-to-grasp task, in which, six objects, fixed on the turning table, were grasped appropriately when they were transported to multiple positions in 3D space. A preliminary analysis of neural signals recorded in primary motor cortex, shows that plenty of neurons exhibit significant tuning to both target position and grip type.

COMPARISON WITH EXISTING METHOD(S)

Our apparatus realizes an arbitrary combination of parameterized reaching and grasping movements in a single task, which were usually separated or fixed in other systems. Meanwhile, the apparatus has high expansibility in terms of dynamic range, object shapes and applicable subjects.

CONCLUSIONS

The apparatus provides a valuable platform to study upper limb functions in behavioral and neurophysiological studies, and may facilitate simultaneous reconstruction of reaching and grasping movements in brain-machine interfaces (BMIs).