Delay Correlation Subspace Decomposition Algorithm and Its Application in fMRI

Abstract

Abstract-This paper reports a new delay subspace decomposition (DSD) algorithm. Instead of using the canonical zero-delay correlation matrix, the new DSD algorithm introduces a delay into the correlation matrix of the subspace decomposition to suppress noises in the data. The algorithm is applied to functional magnetic resonance imaging (fMRI) to detect the regions of focal activities in the brain. The efficiency is evaluated by comparing with independent component analysis and principal component analysis method of fMRI.Index Terms-Functional magnetic resonance imaging (fMRI), independent component analysis (ICA), noise suppression, principal component analysis (PCA), subspace decomposition.

EEG-fMRI study of the interictal epileptic activity in patients with partial epilepsy

“…According to the null hypothesis, the threshold is reached when there is a significant difference at the p b 0.05 level. In fMRI data processing, all t-statistics for threshold should exceed this obvious level (pb 0.05), that is, it should be p b 0.001, with the corresponding t-threshold value being t N 3.1 and highest threshold being the maximum mean global activity [17]. For removal of noise, the number of activated voxels (the true activated voxels evoked by the task) must be determined and differentiated from the number of other voxels evoked by noise.…”

EEG‐fMRI study of the ictal and interictal epileptic activity in patients with eyelid myoclonia with absences

“…The lowest threshold (t > 3.1, p < 0.001) according to the Hull hypnosis theory, must satisfy the statistical difference of approximately p < 0.05. Generally, during fMRI data processing, thresholds should be specified to exceed the statistical significance level of p < 0.05, thus it was specified as p < 0.001, whose corresponding t threshold value was t > 3.1, and the highest threshold that resulted in mean global maximum activity was used (Chen et al, 2005). To remove the influence of noise, in general, a number of continuously activated voxles were specified as the true activation voxels by an evoked task, otherwise, they were not specified as true activation voxels, but considered as being evoked by noise.…”

Partially Adaptive STAP Algorithm Approaches to Functional MRI

“…Unlike the methods in [6][7], it is not a subspace approach. Rather, it is a Fourier-based joint beamformer-frequency filter.…”