Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 37 of the 38 references it cites were found in a registry or an index; 29 of the 50 sources its claims cite were checked for support.

Computer-aided detection thresholds for digital chest radiography interpretation in tuberculosis diagnostic algorithms.

Open access

Abstract

Objectives: Use of computer-aided detection (CAD) software is recommended to improve tuberculosis screening and triage, but threshold determination is challenging if reference testing has not been performed in all individuals. We aimed to determine such thresholds through secondary analysis of the 2019 Lesotho national tuberculosis prevalence survey. Methods: Symptom screening and chest radiographs were performed in participants aged ≥15 years; those symptomatic or with abnormal chest radiographs provided samples for Xpert MTB/RIF and culture testing. Chest radiographs were processed using CAD4TB version 7. We used six methodological approaches to deal with participants who did not have bacteriological test results to estimate pulmonary tuberculosis prevalence and assess diagnostic accuracy. Results: Among 17 070 participants, 5214 (31%) had their tuberculosis status determined; 142 had tuberculosis. Prevalence estimates varied between methodological approaches (0.83-2.72%). Using multiple imputation to estimate tuberculosis status for those eligible but not tested, and assuming those not eligible for testing were negative, a CAD4TBv7 threshold of 13 had a sensitivity of 89.7% (95% CI 84.6-94.8) and a specificity of 74.2% (73.6-74.9), close to World Health Organization (WHO) target product profile criteria. Assuming all those not tested were negative produced similar results. Conclusions: This is the first study to evaluate CAD4TB in a community screening context employing a range of approaches to account for unknown tuberculosis status. The assumption that those not tested are negative - regardless of testing eligibility status - was robust. As threshold determination must be context specific, our analytically straightforward approach should be adopted to leverage prevalence surveys for CAD threshold determination in other settings with a comparable proportion of eligible but not tested participants.

Overview

CoChat Summary

## TL;DR

Using secondary analysis of the 2019 Lesotho national tuberculosis prevalence survey, the paper estimated a CAD4TBv7 threshold of 13 for chest radiograph screening that achieved about 90% sensitivity and 74% specificity under its preferred assumptions. The main message is that threshold calibration is context-specific, and that treating untested participants as tuberculosis-negative gave robust results in this setting.

## Key findings

  • Among 17,070 participants analyzed, 5,214 (31%) had tuberculosis status determined, and 142 had tuberculosis.
  • Estimated tuberculosis prevalence varied by method from 0.83% to 2.72%; the authors’ “best” estimate was 0.97%.
  • Under the NN/NN approach, the CAD4TBv7 threshold of 13 gave sensitivity of 90.1% (95% CI 84.0–94.1) and specificity of 74.2% (95% CI 73.6–74.9); under NN/MI, sensitivity was 89.7% (95% CI 84.6–94.8) with the same specificity.
  • Subgroup analysis found lower specificity among men, symptomatic participants, and those with a history of tuberculosis; sensitivity was worse for women than men, with differences of around 30% at some thresholds.

## Methods

  • Secondary analysis of the 2019 Lesotho national tuberculosis prevalence survey; adults aged ≥15 years were screened with symptoms and chest radiography, and symptomatic or radiographically abnormal participants provided sputum for Xpert MTB/RIF Ultra and culture.
  • Chest radiographs were processed with CAD4TB versions 6 and 7, and the study focused on CAD4TBv7 threshold determination.
  • Six approaches handled missing bacteriological status, including complete case analysis, treating untested as negative, multiple imputation, and Bayesian latent class analysis; sensitivity, specificity, ROC curves, and AUC were estimated across thresholds.

## Limitations

  • The paper states that its tuberculosis case definition differed from the national survey’s definition, which risked including false positives; it also notes that Ultra and culture are imperfect reference standards, especially in community screening.
  • Diagnostic accuracy for chest radiography comparisons is said to be cautious to interpret because field readers were instructed to over-read, and abnormal chest radiographs were an inclusion criterion for bacteriological testing, creating selection bias.
  • The text makes evident that many participants had missing chest radiographs and that 19% of invited individuals refused participation; the authors note these issues could affect representativeness, though they expect little impact on threshold determination.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1183/23120541.00508-2023

Cites

37 of 37 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

FIGURE 1 Inclusion of participants and tuberculosis determination. # Determination of symptomatic based on cough, fever, night sweats and body weight loss. Chest radiography based on field reading, with abnormal indicating lung fields suggestive of tuberculosis; ¶ Determined as having tuberculosis for this secondary analysis, if Ultra (excluding trace) or culture positive; not having tuberculosis if both negative; otherwise unknown tuberculosis status. CXR: chest radiograph; TB: tuberculosis.; Ultra: Xpert MTB/RIF Ultra.
FIGURE 1 Inclusion of participants and tuberculosis determination. # Determination of symptomatic based on cough, fever, night sweats and body weight loss. Chest radiography based on field reading, with abnormal indicating lung fields suggestive of tuberculosis; ¶ Determined as having tuberculosis for this secondary analysis, if Ultra (excluding trace) or culture positive; not having tuberculosis if both negative; otherwise unknown tuberculosis status. CXR: chest radiograph; TB: tuberculosis.; Ultra: Xpert MTB/RIF Ultra.CC BY-NC · PubMed Central
FIGURE 2 Diagnostic accuracy of CAD4TBv7 under different methodological approaches. a) Receiver operating characteristic curves for CAD4TBv7. Horizontal and vertical dashed lines illustrate WHO minimum target product profile criteria of >90% sensitivity and >70% specificity, respectively. b) Sensitivity (orange) and specificity (blue) plotted against CAD4TBv7 scores, under the six methodological approaches (CC/CC, NN/NN, NN/CC, MI/MI, NN/MI, LCA; see table 1) with 95% CIs. Horizontal dashed lines illustrate WHO minimum target product profile criteria of >90% sensitivity and >70% specificity. Vertical dashed lines are at the CAD4TBv7 threshold of 13. LCA analyses were performed for CAD4TBv7 scores ≤50 only, so the AUC is approximate and the graphs are truncated. AUC: area under the curve; CAD4TBv7: CAD4TB version 7; CC: complete case; LCA: latent class analysis; MI: multiple imputation; NN: not tested=negative.
FIGURE 2 Diagnostic accuracy of CAD4TBv7 under different methodological approaches. a) Receiver operating characteristic curves for CAD4TBv7. Horizontal and vertical dashed lines illustrate WHO minimum target product profile criteria of >90% sensitivity and >70% specificity, respectively. b) Sensitivity (orange) and specificity (blue) plotted against CAD4TBv7 scores, under the six methodological approaches (CC/CC, NN/NN, NN/CC, MI/MI, NN/MI, LCA; see table 1) with 95% CIs. Horizontal dashed lines illustrate WHO minimum target product profile criteria of >90% sensitivity and >70% specificity. Vertical dashed lines are at the CAD4TBv7 threshold of 13. LCA analyses were performed for CAD4TBv7 scores ≤50 only, so the AUC is approximate and the graphs are truncated. AUC: area under the curve; CAD4TBv7: CAD4TB version 7; CC: complete case; LCA: latent class analysis; MI: multiple imputation; NN: not tested=negative.CC BY-NC · PubMed Central
FIGURE 3 Diagnostic accuracy of CAD4TBv7 by subgroups, under NN/NN approach (see table 1). a) Sensitivity and specificity. b) Differences in sensitivity and specificity. Differences are for women versus men (i.e. values above 0 indicate better diagnostic accuracy for women), symptoms versus no symptoms, HIV positive versus HIV negative, and history of tuberculosis versus no history of tuberculosis. Vertical dashed lines are at the CAD4TBv7 threshold of 13. Squares and horizontal lines show the overall median and interquartile range, respectively, of the CAD4TBv7 scores, for those with tuberculosis (orange) and those without tuberculosis (blue), to illustrate where the data lie and hence aid interpretation of the sensitivity and specificity results. CAD4TBv7: CAD4TB version 7; NN/NN: not tested=negative
FIGURE 3 Diagnostic accuracy of CAD4TBv7 by subgroups, under NN/NN approach (see table 1). a) Sensitivity and specificity. b) Differences in sensitivity and specificity. Differences are for women versus men (i.e. values above 0 indicate better diagnostic accuracy for women), symptoms versus no symptoms, HIV positive versus HIV negative, and history of tuberculosis versus no history of tuberculosis. Vertical dashed lines are at the CAD4TBv7 threshold of 13. Squares and horizontal lines show the overall median and interquartile range, respectively, of the CAD4TBv7 scores, for those with tuberculosis (orange) and those without tuberculosis (blue), to illustrate where the data lie and hence aid interpretation of the sensitivity and specificity results. CAD4TBv7: CAD4TB version 7; NN/NN: not tested=negativeCC BY-NC · PubMed Central

Verification

Claims the paper makes

16 Valid · 0 Not valid · 5 Not checked · 61 Own results and interpretations

33 more claims · 5 with their cited sources checked

Create a free account to see all 82 claims and to verify the 9 references still open.

See all claims — free account
  • Probably not a substantial claim. 19
  • It may state two separate claims as one. 25
  • The quote only partly states this claim. 3

Create a free account to see these claims.

Citations

Cites

Show all 38 references

Cited by

See all 13

Metadata

Authors
Vanobberghen, F., Keter, A. K., Jacobs, B. K. M., Glass, T. R., Lynen, L., Law, I., Murphy, K., van Ginneken, B., Ayakaka, I., van Heerden, A., Maama, L., & Reither, K.
Affiliations
  • Swiss Tropical and Public Health Institute (CH)
  • Human Sciences Research Council (ZA)
  • Instituut voor Tropische Geneeskunde (BE)
  • World Health Organization (CH)
  • Radboud University Medical Center (NL)
  • SolidarMed (CH)
  • Ministry of Health (LS)
Corresponding author
Vanobberghen, Fiona
Venue
ERJ Open Research
Published
2023-12-07
Volume
10
Issue
1
Pages
00508-2023
ISSN
2312-0541 (ISSN-L)
OpenAlex
W4389431098
Cited by
13
References
38
Citations by year
2026: 3 · 2025: 8 · 2024: 2
FWCI
2.68
Type
journal article
Language
en
Open access
Locations
  • ERJ Open Research (publishedVersion, cc-by-nc)
  • PubMed Central (publishedVersion, cc-by-nc)
  • PubMed (publishedVersion)
  • Radboud Repository (Radboud University) (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
European and Developing Countries Clinical Trials Partnership (RIA2018D-2498)
Topics
  • Health Sciences › Medicine › Infectious Diseases › Tuberculosis Research and Epidemiology
  • Physical Sciences › Computer Science › Artificial Intelligence › AI in cancer detection
  • Health Sciences › Medicine › Pulmonary and Respiratory Medicine › Lung Cancer Diagnosis and Treatment
Keywords
computer-aided detection, CAD4TB, tuberculosis screening, chest radiography, diagnostic thresholds, diagnostic accuracy, multiple imputation, Xpert MTB/RIF, culture testing, Lesotho

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Computer-aided detection thresholds for digital chest radiography interpretation in tuberculosis diagnostic algorithms. https://app.cochat.ai/papers/10.1183/23120541.00508-2023
The paper
Vanobberghen, F., Keter, A. K., Jacobs, B. K. M., Glass, T. R., Lynen, L., Law, I., Murphy, K., van Ginneken, B., Ayakaka, I., van Heerden, A., Maama, L., & Reither, K. (2024). Computer-aided detection thresholds for digital chest radiography interpretation in tuberculosis diagnostic algorithms. ERJ Open Research. https://doi.org/10.1183/23120541.00508-2023

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.