Anatomically constrained functional image reconstruction

Anatomical information can be used to improve the sensitivity of functional images. The authors have developed an anatomical pixellation method of incorporating registered MR or CT information in brain SPECT image reconstruction. This produces a functional image constrained by an anatomical segmentation. The method has been tested with simulated SPECT images containing a region of increased uptake (hot spot). Using least squares, the simulated image is fitted to the uniform regions defined by MR segments. Hot spots are identified by refining the segmentation based upon assessment of the unprocessed and processed SPECT images. This approach may offer a practical alternative to the more computationally demanding maximum likelihood expectation maximization (ML-EM) Bayesian approach which is the most commonly reported method of incorporating anatomy.