In this paper,a method was developed for the detection of characteristic VOCs in exhaled breath of lung cancer patients by proton transfer reaction mass spectrometry(PTR-MS).An improved breath analysis system was used for 32 normal volunteers and 40 lung cancer patients.The data were statistically analyzed by Mann-Whitney test and logistic regression.The results showed that analytes VOC 33，VOC 39 and VOC 45 might be the breath markers for lung cancer patients,while VOC 45 was significantly different between patients with small cell lung cancer and non-small cell lung cancer.The area derived from the receiver operating curve(AUC),under the logistic regression model,reached to 0.878,with the sensitivity and specificity of 85.5% and 63.5%,respectively,while through the Fisher discriminant model,the area from AUC reached to 0.822,with sensitivity and specificity of 82.5% and 62.5%,respectively.The two models were both statistically significant for lung cancer prediction.