Meeting Banner
Abstract #3245

Distributed Pattern of BOLD fMRI: Neuronal Activity or Hemodynamic Artifact

Sun Mi Park1, Francesca Branzoli2, Misun Kim1, Hyerin Lim1, Matthias J.P. van Osch2, Itamar Ronen2, Dae-Shik Kim1

1Department of Electrical Engineering, Korea Advanced Institute of Science and Technology, Daejeon, Korea; 2C. J. Gorter Center for High-field MRI, Department of Radiology, Leiden University Medical Center, Leiden, Netherlands

Although Multi-voxel pattern analysis (MVPA) has been widely studied to classify recognizing object categories, the signal characteristic of distributed patterns on the human ventral stream remains elusive. Ultra-high field MRI enables imaging at higher spatial resolution and improved localization, thereby potentially leading to improved discrimination between different neuronal patterns. Our results showed that the distributed patterns of 7T BOLD fMRI were better to distinguish each object category than 3T despite equal spatial resolution to provide fair comparison. It has been hypothesized that improving SNR and signal localization at 7T BOLD fMRI resulted in improved discrimination in MVPA methods.

Abstract PDF Presentation Video

Keywords

achieve acquisition activation activity although anatomical asked automatic available began best beta better beyond block blocks blue body bold bottom categories category cats caused classify closer coding considered consist consisting constitute control correlations cortex coverage database degrees detection determine determining discrimination discriminative distinct distributed eight electrical elusive enables ended equal even explain face faces fair favorable female field fusiform generated goal gradient gray healthy house houses hypothesized identify improved included indicates institute interval intervals larger located locations mainly male measured medical mental might mind model modeled modeling neural neuronal nonsense normalized object objects onto overlapping park pattern patterns potentially processing radiology reading regressors remains remove representations represents resolution respectively responding response rest runs scanners scanning science sciences scissors scrambled selected selective separated series sets shoes side significantly spatial stimuli stimulus strength strengths subject task technology temporal thereby thick tics tool transverse trends ultra understanding useful ventral views volunteers whereas whether whole