Spatiotemporal independent component analysis combine general linear model applied to fMRI for eliminating neural noise

Abstract

Functional magnetic resonance imaging (fMRI) has recently become an effective means to explore the mechanism of functional rehabilitation in stroke patients. Neural noise is an inevitable structural noise, and is an important factor caused individual differences in fMRI data, therefore, eliminating the neural noise is being regarded as one of the task that cannot be ignored. In this paper, a new algorithm combines spatiotemporal independent component analysis and general linear model (GLM) is proposed to eliminate the effect caused by excess neural activity. This new algorithm simultaneously maximizes the independence over time and space in fMRI data for establishing the spatiotemporal balance. The new technique was applied to extract the active regions of ankle dorsiflexion during fMRI scanning process. Compared to results of GLM, the results of new combined algorithm is more reasonable with an 8 % improvement in correlation coefficient. It confirmed that this new algorithm is effective in eliminating system noise and neural disturbance.

Appendix

StICA algorithm process is as follows

• (1) Initialize the unmixing matrix W _S, orthogonalization W _S . $W_{T} = (W_{S}^{\text{T}} )^{ - 1} \varLambda^{ - 1}$

• (2) Calculate ${\mathbf{y}}_{S} = {\mathbf{W}}_{{\mathbf{S}}} {\mathbf{X}}_{{\mathbf{S}}}$, ${\mathbf{y}}_{T} = {\mathbf{W}}_{T} {\mathbf{X}}_{{\mathbf{T}}}$

• (3) Use Infomax algorithm and select the sigmoid as a nonlinear function g (·)

  $$\begin{gathered} g({\mathbf{y}}_{S} ) = (1 + e^{{ - {\mathbf{y}}_{S} }} )^{ - 1} \hfill \\ g({\mathbf{y}}_{T} ) = (1 + e^{{ - {\mathbf{y}}_{T} }} )^{ - 1} \hfill \\ \end{gathered}$$

• (4) Calculate regulation increment

  $$\varDelta {\mathbf{W}}_{{\mathbf{S}}} = \mu \left\{ \begin{gathered} \alpha \left[ {{\mathbf{I}} + (1 - 2g({\mathbf{y}}_{{\mathbf{S}}} )){\mathbf{y}}_{{\mathbf{S}}}^{\text{T}} } \right] + \hfill \\ (1 - \alpha )\left[ {{\mathbf{I}} + \left( {1 - 2g({\mathbf{y}}_{{\mathbf{T}}} )} \right){\mathbf{y}}_{{\mathbf{T}}}^{\text{T}} } \right] \cdot \left( {\varLambda^{ - 2} } \right) \hfill \\ \end{gathered} \right\}{\mathbf{W}}_{{\mathbf{S}}}$$

• (5) Update weight value ${\mathbf{W}}_{S}^{ + } = {\mathbf{W}}_{{\mathbf{S}}} + \varDelta {\mathbf{W}}_{{\mathbf{S}}}$ ${\mathbf{W}}_{{\mathbf{S}}}^{ + }$, orthogonalization .${\mathbf{W}}_{{\mathbf{T}}}^{ + } = {\mathbf{W}}_{S}^{ + } \varLambda^{ - 1}$, orthogonalization ${\mathbf{W}}_{{\mathbf{T}}}^{ + }$

• (6) Judge convergence or not