Currently, the custom datastore reads frames from a (tiff/nwb) dataset with functions that read only a single frame at a time. By implementating the read function in such a way that it reads multiple frames at once (for example, all flanking frames and center frame at once). It is likely this will significantly reduce the total time spend on reading.