Duchenne muscular dystrophy (DMD) is a neuromuscular disease that affects males in the pediatric age group. Currently, there is no painless, cost-effective prognostic method available to monitor DMD progression. The main hypothesis of this study was that the biochemical composition of extracellular vesicles (EVs) isolated from the urine of DMD patients can be distinctly differentiated from that of healthy controls using surface-enhanced Raman Spectroscopy (SERS) combined with machine learning models. This differentiation is expected to provide a noninvasive, rapid, and accurate diagnostic tool for the early detection, staging, and monitoring of DMD by identifying the molecular signatures captured by SERS and leveraging the analytical power of machine learning algorithms. We collected fasting morning urine samples from 52 DMD patients and 17 healthy controls and isolated EVs using a Total Exosome Isolation kit. The SERS substrates are prepared using silver nanoparticles, which were employed to capture the molecular fingerprints of the EVs with uniformity and reproducibility, achieving relative standard deviation values of 7.3% and 8.9%. We observed alterations in phenylalanine and α-helical proteins in patients with DMD compared to controls. These spectral data were analyzed using PCA, Support Vector Machines, and k-Nearest Neighbor (KNN) algorithms to identify distinct patterns and stage DMD based on biochemical composition. Our integrated approach demonstrated 60% sensitivity and 100% specificity in distinguishing DMD patients from healthy controls, highlighting the potential of SERS and KNN for noninvasive, accurate, and rapid diagnosis of DMD. This method offers a promising avenue for early detection and personalized treatment strategies, ultimately improving patient outcomes and quality of life.