Abstract: Proposes an effective analysis method to decrease the interference effect of non-brain signals on the functional magnetic resonance imaging (fMRI) at low field, by which the activated area of the brain can be located precisely. The method combines the image segment method based on a deformable surface model with the voxel-based morphometry (VBM). The extracted brain tissue is taken as an explicit mask. Then, an analysis mask is constructed to correct the estimated value of the parameter β in general linear model (GLM), thus making the statistical inference just available to the brain tissue. The practical application revealed that the method proposed can effectively eliminate the pseudo-activated area generated by the changes of the signals from background, skull and scalp and improve the reliability and precision of fMRI at low field.