We were automatically using Medcon to recreate a directory structure and it also served to renumber the images to something that linux could read properly, ie, 001, 002... 010, rather than 1,2... 10 etc.

However, following Rich's tip abou the butchered DICOM header I've retaken images from the scanner that haven't been run though Medcon and run them through Tom Ross's Perl renamer script. This does the same but leaves the headers untouched. It doesn't reorder the anatomical data but as these images are not mosaics, they can be read simply into to3d having been run through Medcon.

I would just like to thank everyone for their excellent advice on this topic. I hope to see you soon for fMRI 7 next month.

Cheers