User-definable bull's-eye database analysis.

Abstract

Several quantitative bull's-eye database programs have been developed and employed successfully, but generally they restrict the user to limited types of quantitative analysis. We developed a type of bull's-eye analysis which facilitates user-defined processing, and then explored the effects of various types of processing on the comparisons of patient information with that of reference databases. Male and female bull's-eye database were generated from 32 normal patients using unweighted 2D prefiltering, ramp backprojection, unweighted 3D postfiltering, and peak value circumferential plotting (base method). The data from each patient were then reprocessed and compared to the databases by means of three different approaches: (1) using the base method, (2) using average as opposed to peak value profiles, and (3) using a resolution recovery prefilter instead of a smoothing prefilter. Significant differences in the number of apparently abnormal regions were found between the three methods. In other words, the type of single-photon emission tomography (SPET) processing affected the accuracy of comparisons between patient and database information. Because even sophisticated analysis can now be performed on personal computers, we conclude that, rather than a preprocessed data file, clinical "normal reference" information should consist of original SPET data (in a standard format, e.g., Interfile) from a series of documented normal patients. Each user could then generate reference bull's-eye database by applying his or her own clinical processing procedures to the data.