Lung carcinoma remains the leading cause of cancer death worldwide. The tactic to change this unfortunate rate may be a timely and rapid diagnostic, which may in many cases improve patient prognosis.In our study, we focus on the comparison of two novel methods of rapid lung carcinoma diagnostics, label-free in vivo and ex vivo Raman spectroscopy of the epithelial tissue, and assess their feasibility in clinical practice. As these techniques are sensitive not only to the basic molecular composition of the analyzed sample but also to the secondary structure of large biomolecules, such as tissue proteins, they represent suitable candidate methods for epithelial cancer diagnostics.During routine bronchoscopy, we collected 78 in vivo Raman spectra of normal and cancerous lung tissue and 37 samples of endobronchial pathologies, which were subsequently analyzed ex vivo. Using machine learning techniques, namely principal component analysis (PCA) and support vector machines (SVM), we were able to reach 87.2% (95% CI, 79.8–94.6%) and 100.0% (95% CI, 92.1–100.0%) of diagnostic accuracy for in vivo and ex vivo setup, respectively. Although the ex vivo approach provided superior results, the rapidity of in vivo Raman spectroscopy might become unmatchable in the acceleration of the diagnostic process.