Development Of A Wrist Decoding Syste Based On Monkey Primary Motor Cortex

Brain-machine interfaces (BMIs) can provides a direct communication and control channel between the brain and external devices. Its aim is to translate neural activities of the brain into specific instructions that can be carried out by external devices. Nowadays, non-human primate model has been the main subject in BMI research, In these studies, the key point is how to get neural signals and movement signals with high quality in a long time. This paper presented exploratory research on neural ensemble recording from motor cortex and wrist movement signal recording techniques based on monkey model. And an invasive BMI system had been developed, in which neural signals from primary motor cortex could be accurately decoded into wrist movement parameters. These included:Firstly, a wrist movement signal recording system was designed. This system could record movement parameters while monkey used his wrist to do rotary actions of supination and pronationturn or to do abduction and flexion movement. This system was consisted of several subsystems, including lever system, headpost system, water reward system, personal computer system, MCU control system and camera monitor system. A simple four direction center-out paradigm was designed for monkey wrist training. After training step by step, monkey could learn to finish this task with his wrist. And the wrist motor parameters were recorded.Secondly, some important neural ensemble recording techniques were explored in this paper. We developed two kinds of headpost and head holder systems. The performance of the two headpost systems was compared and analyzed. It showed that the eight-leg headpost and spheric head holder were better for monkey head fixation. Two kinds of Utah arrays with different connector were used to record neural signals. Also, the characteristics of the connector were analyzed. A fixed pedestal for ICS-96connector was designed. The Utah array could be implanted in the primary motor cortex. We could record neuronal spike activities from motor cortex for a long time.Thirdly, the neural signals from M1cortex and the wrist movement signals were recorded synchronously. The correlation between neuronal firing rates and wrist movement direction was analyzed. The results showed that the neuronal firing rates changed significantly during moving onset. And the neuronal firing mode was different among four directions. These differences were capable of discriminating direction pairs. The LE algorithm was used to visualize the variety pattern of the neural signals in a3-D space. The visualization results showed that the neural signals had regular variety patterns. And the neural trajectory could match the movement trajectory very well.At last, the neuronal firing rates were used to decoded monkey wrist moving parameters, such as direction, position and velocity. The factors with respect to decoding were analyzed systematically. The four directions could be predicted with high accuracy using support vector machine (SVM) and k-Nearest Neighbor algorithm respectively. The four directions could be classified with96%accuracy using SVM algorithm. The wrist position and velocity were predicted by Kalman Filter and general regression neural network (GRNN) respectively. The correlation coefficients (CCs) of trajectory and velocity prediction were above0.8for both horizontal and vertical directions with GRNN algorithm. The best decoded CCs of position and velocity for horizontal and vertical directions were0.9170±0.045,0.8872±0.0778,0.8254±0.0798,0.8376±0.0915, respectively.In a word, the neural signals recorded from M1cortex could be used to predict monkey wrist moving parameters. It was indicated that this monkey wrist decoding system was a successful BMI system. The system could also be used to research the encoding and decoding principle of the motor cortex and to understand the biological mechanism of brain control movement.