Limitations of visual assessment of redistribution in thallium images.

Potential limitations of visual assessment of redistribution in thallium (TI) images were studied and results were compared with computer assessment of redistribution. A four-section phantom filled with TI was imaged (300K counts, 128 X 128 matrix) with appropriate background activity and scatter material. Activity in a "defect" section (DS) was varied from 20% to 100% of reference sections (RS). After interpolative background correction, pseudo "initial" and "late" image pairs (N = 35) were photographed on polaroid film and read by three "blinded" observers using an 0-2, 1/2 step, scale (0 = absent and 2 = normal activity). Scan defects were detected by all readers when DS activity was less than or equal to 59% of RS activity. No reader detected a defect when DS activity was greater than or equal to 67% of RS activity. All "initial" defects were detected by computer analysis. Visual assessment of "initial" DS:RS activity ratio did not correlate well with DS:RS activity ratio of the phantom. In contrast, computer assessment of "initial" DS:RS activity ratio correlated well with phantom DS:RS activity ratio (r = 0.96, p less than .0001). Although 22 of 27 scan pairs with partial (N = 26) or full (N = 1) redistribution were correctly identified as showing redistribution by at least two of three observers, the extent of redistribution was not estimated well by visual analysis. Thus, visual assessment of absolute change ("initial"-to-"late") in DS:RS activity ratio showed considerable scatter in relations to actual changes in DS:RS activity ratio of the phantom.(ABSTRACT TRUNCATED AT 250 WORDS)