C. Kimme-smith et al., Computed radiography dual energy subtraction: Performance evaluation when detecting low-contrast lung nodules in an anthropomorphic phantom, J DIGIT IM, 12(1), 1999, pp. 29-33
A dedicated chest computed radiography (CR) system has an option of energy
subtraction (ES) acquisition. Two imaging plates, rather than one, are sepa
rated by a copper filter to give a high-energy and low-energy image. This s
tudy compares the diagnostic accuracy of conventional computed radiography
to that of ES obtained with two radiographic techniques. One soft tissue on
ly image was obtained at the conventional CR technique ((s) over bar = 254)
and the second was obtained at twice the radiation exposure ((s) over bar
= 131) to reduce noise. An anthropomorphic phantom with superimposed low-co
ntrast lung nodules was imaged 53 times for each radiographic technique. Fi
fteen images had no nodules; 38 images had a total of 90 nodules placed on
the phantom. Three chest radiologists read the three sets of images in a re
ceiver operating characteristic (ROC) study, Significant differences in Az
were only found between (1) the higher exposure energy subtracted images an
d the conventional dose energy subtracted images (P = .095, 90% confidence)
, and (2) the conventional CR and the energy subtracted image obtained at t
he same technique (P = .024, 98% confidence). As a result of this study, en
ergy subtracted images cannot be substituted for conventional CR images whe
n detecting low-contrast nodules, even when twice the exposure is used to o
btain them. Copyright (C) 1999 by W.B. Saunders Company.