Design And Implementation Of A Multichannel Low Power Wireless Bidirectional Brain Machine Interface

All neural information systems (NIS) rely on sensing neural activity to supply commands and control signals for computers, machines and a variety of prosthetic devices. Brain Machine Interfaces (BMI) emerged in recent years as nonconventional direct channels of communication between the human brain and physical devices. BMI found application in control of assistive devices providing handicapped patients with communication and movement skills, monitoring progress of or controlling the onset of certain brain disorders, monitoring the affective state of a patient during and after a rehabilitation session, and rehabilitation of the brain to regain motor functionality.The principle design spaces for an implantable BBMI system has been proposed in this paper. Due to the invasive nature of the implantable system, detailed design requirements has been discussed for each component.A common platform that provides the user with the flexibility to choose any hardware and software BMI components enables collaborative and comparative study opportunities in various research fields. This dissertation presents an open architecture wireless BMI system that is capable of recording 32-channel EEG, ECoG or intracortical data, or as an alternative 8-channel of EEG or EMG data with additional motion data. The system presented in this work offers a low energy communication interface and component based software framework. The system, which is reconfigurable at a minimum integration effort, has been tested with different configurations and the result is comparable with other medical grade physiological signal recording systems.With component-based architecture, the BMI system can be extended to a bidirectional BMI (BBMI) system by adding stimulation module, featured 4-channel differential, or unipolar high voltage stimulation outputs for cortical stimulation or spinal cord stimulation. Invasive systems achieve a high signal-to-noise ratio (SNR) by eliminating the volume conduction problems caused by tissue and bone. An implantable BBMI system using intracortical electrodes provides excellent detection of a broad range of frequency oscillatory activities through the placement of a sensor in direct contact with cortex. The presented system is designed to monitor brain sensorimotor rhythms and present current stimuli with a configurable duration, frequency and amplitude in real time to the brain based on the brain activity report. The battery is charged via a novel ultrasonic wireless power delivery module developed for efficient delivery of power into a deep-implanted system. The system was successfully tested through bench tests and in vivo tests with both time domain analysis and frequency-time domain analysis on a behaving primate to record the local field potential (LFP) oscillation and stimulate the target area at the same time.Physiological signal processing is the most important factor considering the neurophysiologic background behind the signal acquired. The muscle signal recorded from the wearable signal acquisition system developed in this dissertation was compared with commercial product and the achieved promising result. We also combined muscle signals recorded from different muscle groups to classify seven forearm and hand motions. Deep learning algorithms also play significant role when analyzing physiological signals. A deep belief networks has been proposed and successfully applied to muscle data to differentiate different muscle fatigue levels.EEG data analysis has been done to detect concentration related a-rhythm presents. Besides traditional feature extraction methods used in concentration detection, persistent homology was first applied to extract the periodicity in the time series data and used as a supplemental feature for classification.The performance of the systems implemented in this dissertation has been evaluated with different configurations for power analysis. A comparison among this work and related works presented in recent years proved promising result and potential in promoting neuroplasticity.