Medical student Matthias Renker, from the Medical University of South Carolina in Charleston, won a trainee research prize from the RSNA for this poster study, which reviewed triple-rule-out CT scans in 163 consecutive patients (mean age, 53 ± 14 years) from the university's acute chest pain center. The research team ran the CAD and 3D volumetry package on the images, which were then reviewed by two expert radiologists as well as a third observer who called divergent findings.
Ground-truth interpretation showed 51 incidental lung nodules 4 mm or larger among the datasets. CAD had 73% sensitivity, 44% accuracy, and a 76% negative predictive value for finding lung nodules. In 20 patients, CAD missed nodules described in the radiology report. But in its favor, CAD found nodules in three patients that had been missed during the initial interpretation. CAD has the potential to reduce the miss rate for incidental lung lesions, the researchers concluded.