Predictive CT findings of malignancy in ground-glass nodules on thin-section chest CT: the effects on radiologist performance

Abstract

Our purpose was to identify thin-section chest computed tomography (CT) findings of malignancy other than the presence of a solid portion within ground-glass nodules (GGNs) and to evaluate whether the radiologists` performance in determining malignancy can be enhanced with this information. The predictive CT findings of malignancy extracted from the CT findings of 80 GGNs (47 malignant, 33 benign) were a size of > 8 mm [odds ratio (OR), 10.930; P = 0.045] and a lobulated border (OR, 13.769; P = 0.016) for pure GGNs and a lobulated border (OR, 10.200; P = 0.024) for mixed GGNs. Four chest radiologists and five radiology residents participated in the observer performance study with CT of 130 GGNs (67 malignant, 63 benign). Receiver-operating characteristic (ROC) analysis was used to compare radiologists` performances before and after providing these predictive findings. For pure GGNs, mean areas under the curve (A(z)) of all readers without and with CT predictive information were significantly different (0.621 +/- 0.052 and 0.766 +/- 0.055, P < 0.05). For mixed GGNs, the A(z) values achieved without and with predictive information were not significantly different (0.727 +/- 0.064 and 0.764 +/- 0.056, P > 0.05). Information about lesion size and morphological characteristics can enhance radiologists` performance in determining malignancy of pure GGNs.