Extracellular vesicles (EVs) are emerging as crucial biomarkers in cancer diagnostics and therapeutics with their heterogeneity presenting both challenges and opportunities in prostate cancer research. However, existing methods for isolating and characterizing EV subtypes have been limited by inefficient separation and inadequate proteomic analysis. Here we show an optimized centrifugal microfluidic device, Exodisc, that efficiently isolates large quantities of EV subtypes from particle-enriched medium, enabling comprehensive proteomic analysis of small (EV-S, 20-200 nm) and large (EV-L, >200 nm) EVs. Using this device, we successfully separated EV-S and EV-L from prostate cancer cell lines LNCaP and PC3. Mass spectrometry-based proteomics revealed that EV proteins reflect parental cell characteristics more than EV size, with EV-L demonstrating increased expression of PSMA-correlated proteins. Our optimized protocol addresses challenges in EV isolation and characterization, providing a more effective method for studying cellular and molecular mechanisms of specific EV subtypes. This study extends the potential use of EVs as a liquid biopsy for cancer theranostics, paving the way for more precise isolation of EV subtypes and potentially leading to improved biomarker discovery and the development of personalized treatments.