Sort of. The s-form will be used from the NIFTI if that's how the dataset was created versus DICOM or some other format and if align_epi_anat.py is "deobliquing" at the same time. By default, align_epi_anat.py matches the obliquity of the anat to the EPI dataset. The dataset attribute IJK_TO_DICOM_REAL holds the RAI equivalent of the NIFTI s-form (which is in LPI). If you use the -deoblique off option, that deobliquing step is skipped, and the closest cardinal version of that transform is used to identify the coordinates of the dataset.