Performance of Surface-enhanced Raman Spectroscopy in Biopsy Tissue for Rapid Diagnosis of Tracheobronchial Tuberculosis: A Prospective Study

<bold>Background:</bold> It is essential to find rapid and accurate diagnostic tools for tracheobronchial tuberculosis (TBTB). Surface-enhanced Raman spectroscopy (SERS) is a technique that provides information about the structure and composition ratio of biomolecules in tissue, having enormous potential on clinical diagnosis for TBTB. This study aimed to develop and validate a diagnostic model based on spectral characteristics of SERS in bronchial biopsies for identifying TBTB. <bold>Methods:</bold> The training set included patients with TBTB (n=18), airway malignant diseases (n=20), and normal mucosal tissue (n=20). Using a machine learning support vector machine algorithm, a diagnostic SERS model for TBTB was developed in the training set. Then a cohort of 121 patients with suspected TBTB were enrolled as an independent validation set. <bold>Results:</bold> SERS model had a diagnostic yield of 93.39% (113/121), a specificity of 100% (56/56), and produced an AUROC of 0.938. The sensitivity of SERS model was 87.69% (57/65) in patients with TBTB, significantly higher than that of sputum smear (38.46%, 25/65, p<0.01), bronchial brush smear (40.00%, 26/65, p<0.01), and the Bactec MGIT 960 Culture of bronchoalveolar lavage fluid (67.69%, 44/65 p<0.01). SERS correctly diagnosed 85.00% (34/40) of patients with a negative Mycobacterium sputum smear, which was not significantly different from diagnostic accuracy (92.00%, 23/25, p=0.471) in patients with positive sputum smear. The SERS model demonstrated consistent sensitivity across different TBTB subtypes. <bold>Conclusion:</bold> The SERS model using bronchial tissue provides a rapid and practical method of accurately identifying TBTB.