An improvement of independent component analysis with projection method applied to multi-task fMRI data

Authors:
Zhiying Long;Rui Li;Mingqi Hui;Zhen Jin;Li Yao
Affiliations:
State Key Lab of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing 100875, China;State Key Lab of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing 100875, China;State Key Lab of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing 100875, China;Laboratory of Magnetic Resonance Imaging, Beijing 306 Hospital, Beijing 100101, China;State Key Lab of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing 100875, China and School of Information Science, Beijing Normal University, Beijing 100875, China
Venue:
Computers in Biology and Medicine
Year:
2013

Citing 2
Cited 0

An information-maximization approach to blind separation and blind deconvolution

Neural Computation
Evaluating the effective connectivity of resting state networks using conditional Granger causality

Biological Cybernetics

Quantified Score

Hi-index	0.00

Visualization

Abstract

Independent Component Analysis with projection (ICAp) method proposed by Long et al. Hum. Brain Mapp. 30 (2009) 417-431, can solve the interaction among task-related components of multi-task functional magnetic resonance imaging (fMRI) data. However, the departure of the ideal homodynamic response function (HRF) for projection from the true HRF may worse the ICAp results. In order to improve the performance of ICAp, the deconvolved ICAp (DICAp) method is proposed. Both the simulated and real fMRI experiments demonstrate that DICAp can separate more accurate time course corresponding to each task-related components and is more powerful to detect regions activated by each task only than ICAp.