The role of Artificial Intelligence as a community screening tool for Pulmonary Tuberculosis in Tanzania

🔗2024

🔗Journal/Publication: Research Square

🔗Read it in full version: https://doi.org/10.21203/rs.3.rs-4990059/v1

Abstract 

Background

Tanzania has adopted the use of AI in TB screening, largely known as CAD4TB, with GeneXpert remaining a major confirmatory test. East setting using the technology is supposed to setup its own threshold score. But the country has not yet established the score, and no study has been done to understand the efficiency of the technology. This study proposes a threshold and analysis on the efficiency of using CAD4TB in Tanzania.

Methods

This study collected data from four mobile clinic vans, moving around the country for screening of TB. The vans are installed with digital Xray machines, CAD4TB and GeneXpert machines. We compared the score of human readers, CAD4TB against the ground truth of GeneXpert.

Results

From the study of 11923 individuals, out of the 818 suggestive cases as per radiologists reading, only 15 cases were confirmed positive by GeneXpert, that is 6.7% of all cases compared to 99% correlation of CAD4TB. Apart from this correlation of CAD4TB and GeneXpert, we also learn that most of the cases were directly subjected into GeneXpert without the use of CAD4TB although CAD4TB was installed in the vans. The analysis was based on the suggested CAD4TB score by this study.

Conclusion

This study suggests the potential use of CAD4TB in screening of TB in Tanzania and a potential threshold to use. Further studies need to be done for analysis in clinical settings.