To evaluate the diagnostic performance of fast scanning tomosynthesis in comparison with that of chest radiography for the detection of pulmonary nodules, using multidetector-row computed tomography (MDCT) as the reference, and to assess the association of the true-positive fraction (TPF) with the size, CT attenuation value, and location of the nodules.
Materials and Methods:
The institutional review board approved this study, and written informed consent was obtained from all patients. Fifty-seven patients with and 59 without pulmonary nodules underwent chest MDCT, fast scanning tomosynthesis, and radiography. The images of tomosynthesis and radiography were randomly read by 3 blinded radiologists; MDCT served as the reference standard. Free-response receiver-operating characteristic (FROC) and receiver-operating characteristic (ROC) analyses, Cochran-Armitage trend or Fisher exact test, a conditional logistic regression model, and McNemar test were used.
Both FROC and ROC analyses revealed significantly better performance (P < 0.01) of fast scanning tomosynthesis than radiography for the detection of pulmonary nodules. For fast scanning tomosynthesis, the average TPF and false-positive rate as determined by FROC analysis were 0.80 and 0.10, respectively. For both fast scanning tomosynthesis and radiography, the average TPF increased with increasing nodule size and CT attenuation values, and was lower for subpleural nodules (all P < 0.01).
The diagnostic performance of fast scanning tomosynthesis for the detection of pulmonary nodules was significantly superior to that of radiography. The TPF was affected by the size, CT attenuation value, and location of the nodule, in both fast scanning tomosynthesis and radiography.