Summary of Results:
Back
Detection Rate: AI screening identified 11.2% of cases as presumptive TB, similar to human-reader screening (11.3%).
GeneXpert Testing: AI-screened patients proceeded to GeneXpert testing at a higher rate (83.4% vs. 56.7% for human-readers).
Confirmed TB Cases: AI approach diagnosed 3,841 TB cases (3.8% of screened individuals), whereas the human-reader approach diagnosed 4,616 cases (2.4%).
Dropout Rate: AI screening had a significantly lower dropout rate (16.6%) compared to human readers (43.1%).
Number Needed to Screen (NNS): AI screening required screening 26.3 people to find one TB case, compared to 41.5 for human-reader screening.
Conclusion:
The AI-powered chest X-ray screening demonstrated clear advantages over human-reader screening, including higher diagnostic yield (3.8% vs. 2.4%), lower dropout rates (16.6% vs. 43.1%), and greater efficiency in identifying TB cases (NNS of 26.3 vs. 41.5). These findings suggest that AI can accelerate TB case detection, optimize healthcare resource allocation, and improve public health outcomes, particularly in high-burden regions like the Philippines. However, further cost-effectiveness studies are recommended to support large-scale implementation.
