CAD4TB vs. Timika Score: Diagnostic Performance for TB Screening in Indonesia

Indonesia is the focus of this cross-sectional retrospective study evaluating the diagnostic performance of an AI tool (CAD4TB) for tuberculosis screening. The dataset includes analysis of 3,254 chest X-ray images from patients (2018–2020), including 600 with microbiological confirmation. The study, authored by Reyhan Eddy Yunus and last updated in 2026, compares CAD4TB's area under the curve and specificity to the expert-assessed Timika score.

Use Cases

• Benchmarking AI diagnostic performance against traditional clinical scoring systems based on the comparison with the Timika score.
• Evaluating the operational feasibility of automated TB screening tools in resource-limited settings based on the study's focus on Indonesia.
• Analyzing the relationship between diagnostic sensitivity and specificity for chest X-ray-based AI based on the reported 73.91% sensitivity and 71.43% specificity for CAD4TB.

Strengths

• Includes 3,254 patient records with chest X-ray images from a defined period (2018–2020).
• Provides a direct comparison of an AI tool (CAD4TB) with an established clinical score (Timika) using metrics like AUC and specificity.
• Contains a subset of 600 cases with microbiological confirmation, including 46 smear-positive PTB cases, for validation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the underlying analysis data is unknown, which may limit suitability assessment.
• The single-centre, retrospective design may limit generalizability to other populations or settings.

Provenance

Source

Reyhan Eddy Yunus via figshare

Collection Method

Cross-sectional retrospective single-centre study analyzing patient records and chest X-ray images.

Time Range

2018–2020

Freshness

Last updated 2026-05-29 06:13:21; freshness should be verified.

Geography

Indonesia (specific centre not named)