Grading stenotic lesions

This experiment measured human observers' ability to grade stenotic lesions and compared their performance with densitometric calculations on the same noisy images. Five observers graded 192 vessels containing stenotic lesions of 60%, 70%, 80%, or 90% randomly rendered on film transparencies. The precision of the observers' gradings was determined by means of a generalized receiver operating characteristic procedure and was comparable to the precision of the densitometric calculation. The observers' performance decreased only slight as the display window as increased by a factor of 64, a result quite unlike the large display window effects found for observes' detection of disk features on the same noise backgrounds. The vessel remained visible over the range of display windows. These results demonstrate the usefulness of measuring the observers' grading performance as a metric of total system performance