Correlation study of different genotypes of Mycobacterium tuberculosis and the severity of pulmonary imaging in Hainan region

This study aimed to investigate the correlation between Mycobacterium tuberculosis (MTB) genotypes and the severity of pulmonary imaging, and to explore the application of artificial intelligence (AI) in assessing radiological severity in Hainan, China. In this retrospective study, MTB isolates from patients with active pulmonary tuberculosis were genotyped using spoligotyping and categorized into Beijing and non-Beijing genotype families. Drug susceptibility testing (DST) was performed. All patients underwent pretreatment chest computed tomography (CT), which was evaluated using a semi-quantitative CT severity score (range 0-30) and assessed for key imaging features. An AI model was developed to predict severe radiological involvement (CT score >15). Among 243 patients, the Beijing genotype was predominant (78.2%). Patients with Beijing genotype infection had a significantly higher mean CT severity score (14.05 vs. 10.02, P < 0.001) and a higher frequency of cavities (58.95% vs. 37.74%, P = 0.006) compared to those with non-Beijing genotypes. Patients with drug-resistant tuberculosis (DR-TB) also had a higher mean CT score than those with drug-susceptible TB (15.76 vs. 12.53, P < 0.001). Both Beijing genotype (adjusted odds ratio [aOR] = 2.45, P = 0.010) and drug resistance (aOR = 2.20, P = 0.017) were independent risk factors for severe radiological involvement. The AI model achieved an area under the curve (AUC) of 0.81 for predicting severe involvement on the test set. In Hainan, the Beijing genotype of MTB and drug resistance are correlated with more severe pulmonary radiological manifestations. The AI model demonstrated promising performance in predicting severe radiological involvement, suggesting its potential as a supportive tool for objective severity assessment.