Skip to main content
Fig. 1 | Insights into Imaging

Fig. 1

From: A deep learning model using chest X-ray for identifying TB and NTM-LD patients: a cross-sectional study

Fig. 1

Flowchart of dataset establishment from patients with presumptive mycobacterial lung diseases is presented. In brief, a total of 2987 and 1887 patients with tuberculosis (TB), non-tuberculous mycobacteria lung disease (NTM-LD) or suspicious of mycobacterial lung disease whose sputum cultures were negative for mycobacteria (Imitator) were identified in the internal and external cohort, respectively. After excluding patients with anteroposterior chest X-ray (CXR) or with visible medical devices on CXR, 1314 and 971 patients were enrolled in the internal and external cohort. Then, we randomly and equally collected 300 patients for each TB/NTM-LD/Imitator group in the internal cohort, and 200 patients for each TB/NTM-LD/Imitator group in the external cohort to ensure our model could fairly learn from each diagnosis. Finally, 900 patients in the internal cohort were randomly assigned to one of the three datasets: training, internal validation and internal test

Back to article page