You could probably change the parameters in the art_batch.m so that every time it will pick the parameters and it will automatically apply them in all the spm.mat you chose and save the outliers_xxx.mat files which you can use in the subsequent analysis. Essentially you can specify 1st level model without estimating it, run art_batch, come back, plug back outliers_xxx.mat and then estimate. But that is for when you have spm.mat for typically smoothed data.
For using unsmoothed data, it might be more complicated than that. You might need to either put all the unsmoothed in the spm.mat just so art_batch can pick up on them or feed the wrfxxx.img list directly to art.m itself. I have not tried these. But I think the latter involves in changing parameters in art.m itself or build a .cfg file.