Abstract:

In recent years, radiomics models, have effectively distinguished pulmonary tuberculosis from lung cancer, nontuberculous mycobacterial lung disease, community-acquired pneumonia, etc, through integrating clinical features with deep learning technologies. They have shown excellent performance in the differential diagnosis of pulmonary tuberculosis, significantly outperforming traditional imaging evaluations. In particular, they can provide a powerful non-invasive diagnostic tool for extrapulmonary tuberculosis (such as intestinal tuberculosis and lymph node tuberculosis) cases with difficult diagnosis and limited sample acquisition. The constructed multimodal fusion models not only demonstrate high accuracy in differentiating intestinal tuberculosis from Crohn’s disease and lymph node tuberculosis from lymphoma but also show significant potential in predicting drug-resistant tuberculosis and dynamically monitoring treatment responses, presenting broad prospects in the diagnosis and treatment response monitoring of tuberculosis. However, constrained by issues such as uneven dataset quality, limited model generalization ability, and insufficient clinical validation, radiomics models still face severe challenges in the diagnosis and treatment of tuberculosis. Through in-depth literature analysis, this paper systematically reviews the latest research progress and application value of artificial intelligence (AI)-driven computed tomography (CT) radiomics technology in the diagnosis and treatment of tuberculosis (including pulmonary and extrapulmonary tuberculosis). It focuses on the innovative analysis of multimodal fusion technologies and clinical implementation scenarios, aiming to provide references for guiding future research directions, further promoting its application and development in the diagnosis and treatment response monitoring of tuberculosis, and contributing to the precision medicine, prevention, and control of tuberculosis.

Key words: Artificial intelligence, Radiographic image interpretation, computer-assisted, Tuberculosis, Diagnosis, computer-assisted, Review literature as topic