It is crucial to determine the extent to which the major physical degradations inherent in imaging diminish the detection and accuracy of activity estimation with tumor-avid imaging agents, and to ascertain how close state-of-the-art imaging and reconstruction strategies are to the upper limits of compensation. Compensation for degradations is costly in terms of processing time, added complexity in imaging and processing, and/or enhancement of other sources of degradation. Such an investigation would enable future research to focus on the major obstacles to accurate detection and quantitation where there is real potential for improvement. To model the clinical task, detection will be investigated using human-observer LROC studies with images in which the presence or absence of the sources of degradation can be controlled, and with known truth regarding tumor presence and location. The accuracy of estimation of tumor activity will also be evaluated with these images. A range of stimuli will be investigated to thoroughly probe the effects of the degradations and the utility of the compensation strategies. Ga-67 citrate is the tumor-avid imaging agent which will be used as a test-bed for these investigations. Four proposed human-observer LROC and quantitation estimation experiments will investigate: 1) noise correlation (filtered backprojection reconstruction versus maximum likelihood reconstruction versus Poisson noise in the slices) and attenuation correction (none versus uniform versus nonuniform attenuation correction); 2) spatial resolution (no compensation versus linear and iterative restoration versus inclusion of nonstationary resolution in reconstruction versus pseudo-electronic-collimation); 3) scatter and septal penetration (no compensation versus spectral-estimation methods versus spatial-domain compensation methods versus imaging with an idealized detector); and 4) all of these factors with simulated tumors added to actual clinical studies. The rankings of the human-observer studies will be compared to the predictions of the channelized Hotelling numerical-observer to determine if the two methods correlate sufficiently so that the numerical-observer would be useful as a screening method to select optimum strategies.
Showing the most recent 10 out of 19 publications