Objective Different clinical prediction models were constructed to provide an imaging basis for the diagnosis of multidrug-resistant pulmonary tuberculosis (MDR-PTB) by characterizing the lung imaging of MDR-PTB patients. Method Data on imaging characteristics of 110 MDR-PTB patients and 154 drug sensitive pulmonary tuberculosis drug-sensitive pulmonary tuberculosis (DS-PTB) patients admitted to the outpatient department of Shenzhen Center for Chronic Disease Control from January 2018 to December 2020, were retrospectively collected, in order to compare and analyze the differences in the lung imaging characteristics and cavity characteristics between MDR-PTB group and DS-PTB group. Different prediction models were constructed by logistic regression method and then further validated. We analyzed the baseline data, CT imaging characteristics between DS-PTB and MDR-PTB groups. The age, patient type, bud sign OR lobule-centered nodule, calcification, and involvement of the lung field were used to construct model1, the minimum internal diameter of lung cavities was used to construct model2, and the minimum internal diameter and bud sign OR lobule-centered nodule were used to construct model3. Result The incidence rate of pulmonary cavities was higher in the MDR-PTB group (64.5%), and both the number of thin-walled and thick-walled pulmonary cavities were higher than that of the DS-PTB group [(0.8±2.4) vs (0.1±0.3), (1.8±3.1) vs (0.3±0.7)], and the number of lobular segments affected by cavities was higher in the MDR-PTB group than that of the DS-PTB group [(1.0±1.1) vs (0.3±0.5)], with a predominance of the right upper and lower lobes, the antero-posterior and posterior apical segments of the left upper lobe, and the left lower lobe. All differences above were statistically different. In MDR-PTB group, tree bud sign or lobule-centered nodule and bronchiectasis were more likely to occur, and these signs were more likely to involve more lung segments and lung fields, with statistical significance. When factors of age, patient type, tree bud sign or lobule-centered nodule, calcification, and involvement of the lung fields were included in the prediction model, the model achieved a higher diagnostic performance with an area under the curve of 0.829 (95%CI 0.779-0.879). We selected cases with lung cavities of the DS-PTB and MDR-PTB groups and performed propensity matching and one-factor logistic regression analysis for them. The results showed that there were significant differences in the maximum outer diameter, the maximum inner diameter, the minimum inner diameter, the maximum wall thickness and the minimum wall thickness of lung cavities between the two groups. However, only one factor of minimum inner diameter of lung cavities was significantly different in the multivariate logistic regression analysis. The OR was 0.34(95%CI 0.17-0.68, P=0.002). Conclusion In this study, the three constructed models can provide a better performance on MDR-PTB diagnosis, especially when there is insufficient evidence of TB etiology and drug resistance test, providing a certain diagnostic basis.