Extracellular vesicles (EVs) are heavily implicated in diverse pathological processes. Due to their small size, distinct biogenesis, and heterogeneous marker expression, isolation and detection of single EV subpopulations are difficult. Here, we develop a λ-DNA- and aptamer-mediated approach allowing for simultaneous size-selective separation and surface protein analysis of individual EVs. Using a machine learning algorithm to EV signature based on their size and marker expression, we demonstrate that the isolated microvesicles are more efficient than exosomes and apoptotic bodies in discriminating breast cell lines and Stage II breast cancer patients with varied immunohistochemical expression of HER2. Our method provides an important tool to assess the EV heterogeneity at the single EV level with potential value in clinical diagnostics.