The diagnostic potential of brain SPECT imaging is limited by its low spati
al resolution. We propose an iterative reconstruction technique that is bas
ed on the conjugate gradient (CG) algorithm and offers the possibility of i
mproving the spatial resolution of the SPECT activity map of a patient, pro
vided that an MR scan of the same patient is available.
The SPECT data are reconstructed twice. The first reconstruction is to perm
it an accurate coregistration of the MR image to the SPECT one. Then, this
coregistered MR image is used as the starting point for the CG algorithm, w
hich solves a "shifted problem." While the iterations are carried out, the
MR signal of low spatial frequency is replaced by the SPECT activity. Howev
er, the high spatial frequency anatomical features contained in the MR imag
e are weakly affected by the SPECT projector-back-projector operator. They
remain in the iterates and increase their diagnostic potential. If an appro
priate number of iterations is performed, quantitation is not significantly
affected.