Abstract:

Objective    To study the role of receiver operating characteristic(ROC) analysis in comprehensive diagnosis to differentiate malignant from benign solitary pulmonary nodules (SPNs) with multi-slice spiral computed tomography(MSCT).  Methods    177 SPNs were examined with MSCT and pathologically confirmed by surgery and/or needle biopsy. All MSCT images were assessed and graded by three senior radiologists on the basis of malignant likelihood of morphological and density characteristics. According to the maximum diameter (D), ROC curves were constructed and optimal thresholds were determined for diagnosis of malignancy in SPNS of D≤3.0cm, 2.0cm＜D≤3.0cm and D≤2.0cm. ROC analysis was used to evaluate the efficiency of MSCT in differentiating malignant from benign SPNs. The area under the ROC curve (Az) was used as an index for diagnostic efficiency with MSCT. Az values were statistically analyzed by Z-test. Results    Az values of ROC curves were 0.879±0.030 in SPNs of D≤3.0cm, 0.899±0.039 in SPNs of 2.0cm＜D≤3.0cm and 0.867±0.044 in SPNs of D≤2.0cm. The differences were statistically significant according to the Ztest (P＜0.05). The three optimal thresholds for diagnosis of malignancy were 3.0, 3.5 and 2.5. Conclusion     Application of ROC analysis is of obvious value in the differential diagnosis of SPNs with comprehensive imaging characteristics obtained by MSCT. It is favorable to combine the diagnosis of SPNs by MSCT with the malignant probability and to improve the treatment of SPNs in clinical practice.

Key words: Coin lesion, pulmonary; Tomography, X-ray computed； ROC curve