Unsupervised Spatiotemporal Analysis of FMRI Data For Measuring Relative Timings of Brain Responses

Abstract

Functional magnetic resonance imaging (fMRI) is a non-invasive imaging technique that has emerged as a powerful tool to identify the brain regions involved in cognitive processes. FMRI offers spatial and temporal resolutions adequate to measure the location, amplitude and timing of brain activity. FMRI data are commonly analyzed voxel-by-voxel using linear regression models (statistical parametric mapping). This requires information about stimulus timing and assumptions about the shape and timing of the hemodynamic response. This approach may be too restrictive to capture the broad range of possible brain activation patterns in space and time and across subjects. This dissertation presents a multivariate data-driven approach using self-organizing maps that overcome the aforementioned limitations. A self-organizing map is a topology-preserving artificial neural network model that transforms high-dimensional data into a low-dimensional map of output nodes using unsupervised learning. This dissertation proposes novel graph-based visualizations of self-organizing maps for extracting fine spatiotemporal patterns of brain activities from fMRI data to measure relative timings of brain responses.

This approach was employed to identify voxels responding to the task and detect differences as small as 28 ms in the timings of brain responses in visual cortex. It outperformed other common techniques for voxel selection including independent component analysis, voxelwise univariate linear regression analysis and a separate localizer scan. This was verified by observing a statistically strong linear relationship between induced and measured timing differences. The approach was also used to correctly identify and classify task-related brain areas in an fMRI reaction time experiment involving a visuo-manual response task. In summary, the graph-based visualizations of self-organizing maps help in advanced visualization of cluster boundaries in fMRI data, thereby enabling the separation of regions with small differences in the timings of their brain responses and helping to measure relative timings of brain responses.