Recent diagnostic procedure advances have considerably improved early lung cancer detection. However, the invasive, unpleasant, and inconvenient nature of current diagnostic procedures limits their application. There is a great need for novel noninvasive biomarkers for early lung cancer diagnosis. In the present study, we aimed to determine whether microRNA (miRNA) blood signatures are suitable for early detection of lung cancer. Using quantitative reverse transcriptase PCR analysis, we first selected and identified three aberrant plasma expression miRNAs (miR-21, miR-145, and miR-155) in a training set of 62 patients and 60 healthy smokers to define a panel that had high diagnostic efficiency for lung cancer. Then, we validated the detective ability of this miRNA panel in a testing set of 34 malignant tumor patients, 30 patients with benign pulmonary nodules and 32 healthy smokers. In the training set, miR-21 and miR-155 showed higher plasma expression levels, whereas miR-145 showed a lower expression level in patients with malignant cancer, compared with healthy controls (P≤0.001). The three miRNAs used in combination produced the area under receiver operating characteristic curve at 0.847, which helped distinguish lung cancer from healthy smokers with 69.4% sensitivity and 78.3% specificity. A logistic regression model with the best prediction was constructed on the basis of miR-21, miR-145, and miR-155. Validation of the miRNA panel in the testing set confirmed their diagnostic value, which yields a significant improvement over any single one. Plasma miR-21, miR-145, and miR-155 have strong potential as novel noninvasive biomarkers for early detection of lung cancer.