Objective: The aim of this study was to identify novel diagnostic biomarkers of diabetic vascular dementia (DVD) and unraveling the underlying mechanisms using mass spectrometry (MS). Methods: Label-Free LC-MS/MS proteomics analysis was applied to urine samples from four groups including 14 vascular dementia CC(VD), 22 type 2 diabetes mellitus (T2DM), 12 DVD, 21 Normal Control (NC). Searching the MS data by Proteome Discoverer software, protein abundances were analyzed qualitatively and quantitatively and compared between these groups. Combining bioinformatics analysis using GO, pathway crosstalk analysis using KEGG, PPI network analysis using STRING and literature searching, the differentially expressed proteins (DEPs) of DVD can be comprehensively judged and were further quantified by receiver operating characteristic (ROC) curve methods. Results: The proteomic findings showed quantitative changes in DVD was compared to NC, T2DM, and VD groups; among 7,527 identified urine proteins, 1222, 1152, and 1180 of the proteins displayed quantitative changes unique to DVD vs NC, T2DM, and VD, respectively, including 481 overlapped common DEPs. Then 9 unique proteins and 2 composite markers (CM) were confirmed by a ROC curve method. Conclusion: This study provided a novel insight into the potential pathogenesis of DVD and elucidated a method for the earlier detection.