the most accurate placement I've been able to obtain is by transforming the grid with the inverted matrix produced from linearly registering MRI to CT, however it is far from exact (see attached). any other ideas would be much appreciated.