Thresholding a t-map does not require additional information. However, its hard to get SPM to threshold a t-map computed outside of the SPM.mat structure.
In terms of thresholding the t-map:
(1) using voxel p-value and cluster extent requires no additional information [although you'd want to say how you came up with the cluster extent (e.g. 3dClustSim);
(2) using voxel-wise FDR correction requires no additional information; and
(3) using any RFT correction (tFDR, cluster FDR, cluster FWE) requires knowing the FWHM that is computed from the residuals or residual image or another source.
With that being said, computing the t-map in SPM is the best approach and can be done by setting the variance of the groups to be equal. The unequal variance option will cause the model to fail.