Lung cancer has a high mortality rate, but early diagnosis can contribute to a favorable prognosis. Liquid biopsy, which restores and identifies tumor-related biomarkers in body fluids, has great potential for early diagnosis. The exome, found in the blood of nanoscale cellular vesicles, has been proposed as a promising biomarker for liquid biopsy. Here, we demonstrate an accurate diagnosis in the early stages of lung use using enhanced Raman spectroscopy (SERS) on a deep-based study of exosomes. Our approach was to study the characteristics of cell exotomies through in-depth study and to determine the similarity of human plasma exotomies without studying human insufficient data. The in-depth training model was trained with SERS signals that received exotomies obtained from normal and lung cancer cell lines and could be classified with 95% accuracy. In 43 patients, including patients with type I and II cancer, the in-depth study model predicted that 90.7% of patients with plasma exoskeletons had more similarities to lung cancer cell exoskeletons than with healthy controls. Such similarities were proportional to the progression of the cancer. It is noteworthy that the model predicts lung cancer with an area under the curve (AUC) of 0.912, for patients with whole cohort and stage I, with an AUC of 0.910. These results suggest great potential for a combination of exosomal analysis and deep learning as a method for liquid biopsy in the early stages of lung cancer.