Spatial transformations of positron emission tomographic data for aligning images or transforming to standard anatomical space are usually performed with reconstructed images. However, they can also be performed during the reconstruction process, thereby interpolating the raw data fewer times. We investigated the performance of spatial transformations during reconstruction, implemented it in a standard 3D reconstruction algorithm, and tested it on phantom and patient H215O activation studies for the application of aligning both transmission and emission scans. Performing the transformations during reconstruction was shown to be equivalent to performing the transformations with reconstructed images for this particular application.