Practice-Related Changes in Neural Circuitry Supporting Eye Movements Investigated via Wavelet-Based Clustering Analysis

• Activity Number: 170
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Jinae Lee*+ and Cheolwoo Park and Benjamin Austin and Kara Dyckman and Qingyang Li and Jennifer E. McDowell and Nicole Lazar
• Companies: University of Georgia and University of Georgia and University of Wisconsin and University of Georgia and University of Georgia and University of Georgia and University of Georgia
• Keywords: fMRI ; temporal correlation ; clustering ; wavelet ; saccade tasks ; no trend test

Abstract

In fMRI studies clustering methods are used to detect similarities in the activation time series among voxels. It is assumed that the temporal pattern of activation is organized in a spatially coherent fashion such that clustering will extract the main temporal patterns and partition the dataset by grouping similarly behaved functions together. We propose a clustering procedure built in the wavelet domain to take temporal correlation into account. We also construct a no trend test based on wavelets to significantly reduce the high dimension of the data prior to clustering. In an actual clustering step, PCA K-means is applied and produces clustered maps that show the apparent structure of the activation of a brain. First, we evaluate the performance of our clustering method using the simulated data and compare it to other approaches. Second, we analyze the fMRI data acquired on two occasions while the 37 participants were engaged in saccade tasks (anti-saccade, pro-saccade, and fixation). We attempt to aggregate voxel time series into a small number of clusters and compare the clustered maps for the three practice groups and also between the two scan time points.