The study found that diabetes was associated with more severe radiographic abnormalities in TB, as detected by AI-based computer-aided detection (CAD) software. Participants with diabetes had a significantly higher median CAD abnormality score (85.3 vs. 74.1, P = 0.002) compared to those without diabetes. Additionally, 43% of diabetic TB patients had cavitary disease outside the upper lung zones, compared to 25% in non-diabetic TB patients (P = 0.01). These findings indicate that AI-CXR screening can effectively quantify TB severity and suggest that diabetes contributes to more extensive and atypical TB presentations, emphasizing the need for targeted screening and management strategies in high-risk populations.
