Using machine learning to enhance the capabilities of functional magnetic resonance imaging
Dr. Stephen LaConte
Within both the machine learning and cognitive neuroscience communities, there has been a growing interest in multi-voxel pattern analysis applied to functional magnetic resonance imaging (fMRI) data. This type of analysis allows the investigator to decode brain states associated with the time the images were acquired (that is, determine what the volunteer was “doing” – e.g. receiving sensory input, effecting motor output, or otherwise internally focusing on a prescribed task or thought). In part, this interest has been fostered by a growing number of fundamental methodological studies of predictive modeling approaches, combined with an increasing awareness that such analyses can make profound contributions to how we interpret mental representations. This talk focuses on the author’s work in applying supervised learning methods to directly impact fMRI technology with the aim of improving data acquisition and analysis. Specifically, we will describe i) the development of data-driven validation techniques for evaluating and optimizing the experimental parameters of image acquisition and analysis, iii) the application of multivariate regression to achieve image-based eye tracking, and iii) the implementation of a real-time fMRI biofeedback system based on brain state prediction.