Analysis of motor cortical control and adaptation in a brain-machine interface setting

Most of current modeling efforts in brain-machine interface (BMI) have concentrated on generating regression models for the prediction of movement parameters, such as hand trajectory, velocity and control force, from neural activities collected during physical movements under stereotypical lab conditions. This study, however, focuses on a different BMI paradigm, where abstract supervisory control commands such as "Go left," "Go right," "Stop," were extracted from neurons in the motor and premotor areas of the rat's brain. The control output was thus formulated as a solution of a classification problem.; A nonlinear support vector machine (SVM) model was developed and had enabled a real-time closed-loop control with superior performance compared with Bayes classifier and the self-organizing map. The selection of neurons and other parameters used in the model was performed using both model independent criteria and model dependent methods.; In addition, it was found that the principal component feature vectors revealed the weight of importance of individual neurons and windows of time in the decision making process. One of the first principal components had much higher discrimination capability than others. A two stage approach using principal components with a Bayes classifier achieved classification accuracy comparable to that obtained by an SVM.; Furthermore, evidence was found that the animal changed his behavior and neural activity during the use of the interface from the hand-control phase to the brain-control phase. The analysis showed that the animal adapted a subset of its neural activities to make his decisions more distinct in the SVM decision space from neural activities, which subsequently led to improved brain-control task performance. Two independent approaches, an SVM model sensitivity analysis and a model-free mutual information analysis, pointed to the same subset of neurons that were responsible for such changes.; In summary, the contribution of the dissertation includes: the design of the decision algorithm for a novel BMI using SVM; the discovery of the role of principal components in neural feature space; and an in-depth analysis of a possible neural computation mechanism based on the rat's neural firing activities in relation to the rat's behavioral adaptation in using the BMI.