The goal of the proposed research is to develop a combination of methods which will improve both tumor detection and the accuracy of biodistribution studies when monoclonal antibodies are used to image the abdomen with SPECT. To correct for the imaging of scattered photons, we propose to refine the dual photopeak window (DPW) method. DPW uses a regression relation between the scatter fraction and the ratio of counts between two, non-overlapping windows positioned over the photopeak to estimate the counts due to scatter at each pixel. We propose to achieve, by using Monte Carlo methods, an investigation of the underlying assumptions of this method, optimization of window placement, and design of the low-pass filter used to control noise in the estimated scatter distribution. To correct for attenuation in a uniform medium, and decrease the variation in spatial resolution with distance from the collimator, we propose the extension from 2D to 3D of the simultaneous correction algorithm (SCA). With use of DPW and SCA, the 3D modulation transfer function (MTF) will be approximately stationary. Thus, the third method we propose to investigate will be stationary, 2D pre- and 3D post-reconstruction restoration filters. These filters will balance the suppression of image noise against the restoration of spatial resolution according to the criterion used in designing the filter. The performance of these filters, according to the following physical measures of image quality, shall be determined: 1) FWHM, FWTM, and 3D MTF; 2) contrast recovery of """"""""cold"""""""" lesions; 3) contrast recovery of """"""""hot"""""""" lesions; 4) fractional standard deviation in voxel count; and 5) mean square error compared to known truth. A Monte Carlo simulation of SPECT abdominal imaging will be developed and used to test the performance of the methods for """"""""hot"""""""" and """"""""cold"""""""" tumor detection in the liver as judged using the Hotelling trace model observer. An ROC study of tumor detection using a standard reconstruction method, and three combinations of the method selected based on the results of the Hotelling trace observer, will be conducted to quantitate the difference in the detection of tumors which occurs with use of these methods by human observers. The methods will also be assessed in terms of their ability to quantitate activity accurately in phantoms.
Showing the most recent 10 out of 19 publications
