Observer performance with computer-generated images of 201Tl-Cl myocardial perfusion.

Publisher Website
Google Scholar
Abstract
The effect of simple image processing on the interpretation of 201Tl-Cl myocardial perfusion images was evaluated by ROC [receiver operating characteristics] analysis. Polaroid images of the cathode ray tube of an Anger camera and computer-processed transparent images recorded in color and shades of gray were examined by multiple observers. A total of 198 observer responses was accumulated for each of 5 image formats. The observer responses were compared with coronary angiography diagnosis. Better performance was obtained for all computer-generated images, except in the region of low false positive values. The responses for color-scale formats were generally superior to those from the gray-scale formats. No significant improvement in performance resulted from the use of background subtraction with a rescaling of the 16-level gray scale to the residual information. When 9-point smoothing was applied to the color-scale display format, there was a poorer observer response.