Abnormal lipid metabolism is a common feature in most solid tumors, and occurs in early stages of the tumor progression. As benign breast tumor is different from malignant tumor of breast cancer, it is particularly important to take benign breast tumor into consideration when investigating cancer biomarkers. In this study, by using a normal-phase/reversed-phase two-dimensional liquid chromatography-mass spectrometry (NP/RP 2D LC-MS) method, we conducted comprehensive lipid profiling in human plasma obtained from six benign breast tumor patients and five breast cancer patients, as well as nine healthy controls. As a result, 512 lipid species were successfully identified. Principal component analysis allowed clear separation of the three groups. Quantitative analysis revealed that many lipid contents were similar in benign and malignant breast tumors compared with controls, and these were proposed as potential breast tumor biomarkers other than breast cancer biomarkers. Two phosphatidylinositol (PI) species, including PI (16:0/16:1) and PI (18:0/20:4), could differentiate between benign and malignant breast tumors, as well as breast cancer patients and healthy controls, indicating that they could be utilized as potential breast cancer biomarkers. In addition, PI (16:0/18:1), phosphatidylglycerol (36:3), and glucosylceramide (d18:1/15:1) were demonstrated to be potential biomarkers to evaluate the level of malignancy of breast tumor. Taken together, our results indicate the usefulness of lipid profiling in the discrimination between patients with breast cancer and non-carcinoma lesions, which might provide assistance in clinical diagnosis.