application) This study will use five figures-of-merit to compare four reconstruction methods for single photon emission computed tomography (SPECT). The SPECT application which will be modeled is the detection of ovarian cancer using a monoclonal antibody for such cancer (OC-125) labeled with In-111. The results of this study should also be applicable to other clinical applications which image the distribution of In-111 labeled pharmaceuticals in the abdomen using SPECT. The four reconstruction methods to be compared all account for the effects of 1/r blurring and attenuation (1/r blurring occurs if straight backprojection is used to reconstruct tomographic slices). The effects of the spatially dependent blurring, which is characteristic of scintigraphy, are accounted for, to differing degrees, by the four methods. One method does no correction. A second does only a 2-dimensional correction for the 3-dimensional (3D) blur. The third and fourth methods both do a full 3D correction for the blurring, an approach which has significantly improved image quality in previous studies. However, the computational loads of the last two methods differ greatly. The fourth method, maximum likelihood estimation using the expectation maximization (EM) algorithm, requires considerable computational effort. Apparently, this study will be the first to apply the EM algorithm to SPECT data in a fully 3D sense. The spatial distribution of the In-111 labeled OC-125 antibody in the abdomen will be modeled. The distance dependence of the collimator and camera blur, the depth dependence of the scatter blur, and the septal penetration associated with imaging In-111 will be measured. Simulated and experimental SPECT acquisition sets will be obtained. These will be reconstructed and five figures-of-merit will be computed for each reconstruction method. One of the figures-of-merit has been shown to correlate well with the ability of trained observers to detect a known """"""""lesion"""""""" in noisy images. The other figures-of-merit have been used in previous evaluations of reconstruction and restoration methods. The noise and biases in the reconstructions will also be characterized.
