JMIR Publications

ABSTRACT

Background:

In resource-limited areas, severe shortages of radiologist contribute to high rates of missed pulmonary tuberculosis (PTB) cases when relying solely on conventional chest X-ray (CXR). Although artificial intelligence (AI)-powered computer-aided detection (CAD) has proven effective in PTB diagnosis, its real-world performance remained underexplored.

Objective:

To evaluate the real-world diagnostic yield of CAD technology as a triage tool for detecting PTB in primary healthcare facilities in high-burden areas.

Methods:

We conducted a retrospective paired-design diagnostic yield study using CXR images collected from 7 county- and 32 township-level healthcare facilities in Yichang city between 2022 and 2024. All images were retrospectively reprocessed with CAD software (JF CXR-1 v2), and the original reports interpreted by radiologists from the time of patient admission were extracted. CAD and radiologist performances were compared using two primary evaluation indicators: diagnostic yield among diagnosed cases (DYD) and positive predictive value (PPV). Subgroup analysis (by region, age, sex, healthcare facility tier, and patient category) and sensitivity analysis were conducted to assess the robustness of the results.

Results:

Among 93,319 enrolled study patients (including 273 bacteriologically confirmed PTB cases), CAD demonstrated a substantially higher DYD (83.88%, 229/273) than radiologists (25.64%, 70/273), though with much lower PPV (1.70% vs. 10.31%). This high-sensitivity performance achieved an 85.52% reduction (only 13,515 instead of 93,319 CXRs) in radiologist workload via selective review of CAD-positive images, without missing any radiologist-identified PTB cases. Furthermore, among CAD-positive images, probability scores >0.75 was a key threshold for identifying high-risk PTB patients, prioritizing for radiologist review. Subgroup analysis further revealed that CAD outperformed radiologists in identifying PTB cases across all scenarios, despite some heterogeneity. CAD performance was significantly better in township-level medical facilities (DYD: 86.72%; PPV: 2.00%) than in county-level hospitals (DYD: 62.50%; PPV: 0.65%).

Conclusions:

CAD technology demonstrates a valuable capability for detecting PTB in primary healthcare facilities. Combined with a tiered "AI pre-screening with selective human review" strategy, this approach effectively alleviates workload of radiologist in resource-constrained regions, offering a scalable solution for tuberculosis prevention and control.