Comparative Evaluation of Digital and Analog Chest Radiographs to Identify Tuberculosis using Deep Learning Model

Results Summary:

qXR AUC (digital CXRs): 0.928, Sensitivity: 0.841, Specificity: 0.806.

Analog vs. Digital AUC Differences:

iPhone 8: 2.55%, Samsung S8: 5.10%, iPhone XS: 1.91%.

qXR showed consistent accuracy across digital and analog CXRs.

Conclusion:

qXR reliably detects TB in both digital and smartphone-captured analog CXRs, making it a valuable triage tool for resource-limited settings. Minimal performance differences across smartphone models support its robustness for TB screening, even where digital radiography is unavailable. Further validation is needed for diverse environments.