<p>The aim of this study is to develop and validate a radiomics model based on high-resolution computed tomography (HRCT) for distinguishing tracheobronchial tuberculosis (TBTB) patients from pulmonary tuberculosis (PTB) patients and healthy controls and to evaluate its diagnostic performance against specialist radiologists. In this multicenter, retrospective study, high-resolution computed tomography (HRCT) data from 479 patients were included (389 from Hospital X, 90 from Hospital Y). Two classification tasks were defined: Group A (tuberculosis vs. healthy) and Group B (TBTB vs. PTB). The data were divided into training, internal test, and external test sets. Whole lungs were automatically segmented via 3D-Slicer to define regions of interest (ROIs), with manual correction by two radiologists. A total of 1770 radiomic features were extracted. After feature stability assessment (intraclass correlation coefficient (ICC) &gt; 0.8) and selection via least absolute shrinkage and selection operator (LASSO) regression, a logistic regression classifier was built. Model performance was evaluated via the receiver operating characteristic (ROC) curve (AUC), accuracy, precision, recall, and F1 score and was compared with that of two blinded radiologists. The logistic regression model demonstrated superior performance. For Group A, the external test set achieved an AUC of 0.867 and an accuracy of 0.822. For Group B, the external test set achieved an AUC of 0.854 and an accuracy of 0.831. In contrast, on the external test set, the accuracies of the two radiologists for identifying TBTB were 0.678 and 0.746, respectively. We developed a robust, externally validated radiomic model that noninvasively and accurately distinguishes TBTB from routine high-resolution computed tomography (HRCT). Its performance surpasses that of specialist radiologists, highlighting its potential as a clinical decision support tool to prompt bronchoscopy and improve patient outcomes.</p>