Evaluation of chest X-ray with automated interpretation algorithms for mass tuberculosis screening in prisons: a cross-sectional study.

Automated interpretation Diagnostics Prisons Tuberculosis X-ray

Journal

Lancet regional health. Americas
ISSN: 2667-193X
Titre abrégé: Lancet Reg Health Am
Pays: England
ID NLM: 9918232503006676

Informations de publication

Date de publication:
Jan 2023
Historique:
received: 17 05 2022
revised: 28 09 2022
accepted: 18 10 2022
entrez: 13 2 2023
pubmed: 14 2 2023
medline: 14 2 2023
Statut: epublish

Résumé

The World Health Organization (WHO) recommends systematic tuberculosis (TB) screening in prisons. Evidence is lacking for accurate and scalable screening approaches in this setting. We aimed to assess the accuracy of artificial intelligence-based chest x-ray interpretation algorithms for TB screening in prisons. We performed prospective TB screening in three male prisons in Brazil from October 2017 to December 2019. We administered a standardized questionnaire, performed a chest x-ray in a mobile unit, and collected sputum for confirmatory testing using Xpert MTB/RIF and culture. We evaluated x-ray images using three algorithms (CAD4TB version 6, Lunit version 3.1.0.0 and qXR version 3) and compared their accuracy. We utilized multivariable logistic regression to assess the effect of demographic and clinical characteristics on algorithm accuracy. Finally, we investigated the relationship between abnormality scores and Xpert semi-quantitative results. Among 2075 incarcerated individuals, 259 (12.5%) had confirmed TB. All three algorithms performed similarly overall with area under the receiver operating characteristic curve (AUC) of 0.88-0.91. At 90% sensitivity, only LunitTB and qXR met the WHO Target Product Profile requirements for a triage test, with specificity of 84% and 74%, respectively. All algorithms had variable performance by age, prior TB, smoking, and presence of TB symptoms. LunitTB was the most robust to this heterogeneity but nonetheless failed to meet the TPP for individuals with previous TB. Abnormality scores of all three algorithms were significantly correlated with sputum bacillary load. Automated x-ray interpretation algorithms can be an effective triage tool for TB screening in prisons. However, their specificity is insufficient in individuals with previous TB. This study was supported by the US National Institutes of Health (grant numbers R01 AI130058 and R01 AI149620) and the State Secretary of Health of Mato Grosso do Sul.

Sections du résumé

Background UNASSIGNED
The World Health Organization (WHO) recommends systematic tuberculosis (TB) screening in prisons. Evidence is lacking for accurate and scalable screening approaches in this setting. We aimed to assess the accuracy of artificial intelligence-based chest x-ray interpretation algorithms for TB screening in prisons.
Methods UNASSIGNED
We performed prospective TB screening in three male prisons in Brazil from October 2017 to December 2019. We administered a standardized questionnaire, performed a chest x-ray in a mobile unit, and collected sputum for confirmatory testing using Xpert MTB/RIF and culture. We evaluated x-ray images using three algorithms (CAD4TB version 6, Lunit version 3.1.0.0 and qXR version 3) and compared their accuracy. We utilized multivariable logistic regression to assess the effect of demographic and clinical characteristics on algorithm accuracy. Finally, we investigated the relationship between abnormality scores and Xpert semi-quantitative results.
Findings UNASSIGNED
Among 2075 incarcerated individuals, 259 (12.5%) had confirmed TB. All three algorithms performed similarly overall with area under the receiver operating characteristic curve (AUC) of 0.88-0.91. At 90% sensitivity, only LunitTB and qXR met the WHO Target Product Profile requirements for a triage test, with specificity of 84% and 74%, respectively. All algorithms had variable performance by age, prior TB, smoking, and presence of TB symptoms. LunitTB was the most robust to this heterogeneity but nonetheless failed to meet the TPP for individuals with previous TB. Abnormality scores of all three algorithms were significantly correlated with sputum bacillary load.
Interpretation UNASSIGNED
Automated x-ray interpretation algorithms can be an effective triage tool for TB screening in prisons. However, their specificity is insufficient in individuals with previous TB.
Funding UNASSIGNED
This study was supported by the US National Institutes of Health (grant numbers R01 AI130058 and R01 AI149620) and the State Secretary of Health of Mato Grosso do Sul.

Identifiants

pubmed: 36776567
doi: 10.1016/j.lana.2022.100388
pii: S2667-193X(22)00205-8
pmc: PMC9904090
doi:

Types de publication

Journal Article

Langues

eng

Pagination

100388

Subventions

Organisme : NIAID NIH HHS
ID : R01 AI130058
Pays : United States
Organisme : NIAID NIH HHS
ID : R01 AI149620
Pays : United States

Informations de copyright

© 2022 The Author(s).

Déclaration de conflit d'intérêts

The authors declare no conflict of interest.

Références

J Biomed Inform. 2009 Apr;42(2):377-81
pubmed: 18929686
S Afr Med J. 2016 Dec 01;106(12):1263-1269
pubmed: 27917775
Clin Infect Dis. 2022 Apr 28;74(8):1390-1400
pubmed: 34286831
Eur Respir J. 2013 Aug;42(2):480-94
pubmed: 23222871
Lancet. 2021 Apr 24;397(10284):1591-1596
pubmed: 33838724
Public Health Rep (1896). 1961 Jan;76:19-24
pubmed: 13694948
Lancet Digit Health. 2020 Nov;2(11):e573-e581
pubmed: 33328086
BMJ Open. 2021 Aug 24;11(8):e045289
pubmed: 34429305
Thorax. 2004 Apr;59(4):286-90
pubmed: 15047946
J Biomed Inform. 2019 Jul;95:103208
pubmed: 31078660
Clin Infect Dis. 2022 Jul 6;74(12):2115-2121
pubmed: 34718459
Sci Rep. 2020 Mar 26;10(1):5492
pubmed: 32218458
Lancet Respir Med. 2020 Apr;8(4):368-382
pubmed: 32066534
Lancet. 1999 Feb 6;353(9151):444-9
pubmed: 9989714
IEEE Trans Med Imaging. 2014 Feb;33(2):233-45
pubmed: 24108713
Lancet Public Health. 2021 May;6(5):e300-e308
pubmed: 33765455
PLoS One. 2015 Oct 14;10(10):e0139487
pubmed: 26466312
J Rheumatol Suppl. 2014 May;91:32-40
pubmed: 24788998
BMC Bioinformatics. 2011 Mar 17;12:77
pubmed: 21414208
Sci Rep. 2018 Mar 26;8(1):5201
pubmed: 29581435
Clin Infect Dis. 2021 Mar 1;72(5):771-777
pubmed: 32064514
Lancet Digit Health. 2021 Sep;3(9):e543-e554
pubmed: 34446265
Ann Am Thorac Soc. 2022 Aug;19(8):1313-1319
pubmed: 34914539
Sci Rep. 2020 Jan 14;10(1):210
pubmed: 31937802
Clin Infect Dis. 2021 Aug 2;73(3):e830-e841
pubmed: 32936877

Auteurs

Thiego Ramon Soares (TR)

Faculty of Health Sciences of Federal University of Grande Dourados, Dourados, MS, Brazil.

Roberto Dias de Oliveira (RD)

Faculty of Health Sciences of Federal University of Grande Dourados, Dourados, MS, Brazil.
Nursing School, State University of Mato Grosso do Sul, Dourados, MS, Brazil.

Yiran E Liu (YE)

Division of Infectious Diseases and Geographic Medicine, Stanford University School of Medicine, Stanford, CA, United States of America.

Andrea da Silva Santos (ADS)

Faculty of Health Sciences of Federal University of Grande Dourados, Dourados, MS, Brazil.

Paulo Cesar Pereira Dos Santos (PCPD)

Faculty of Health Sciences of Federal University of Grande Dourados, Dourados, MS, Brazil.

Luma Ravena Soares Monte (LRS)

Nursing School, State University of Mato Grosso do Sul, Dourados, MS, Brazil.

Lissandra Maia de Oliveira (LM)

Oswaldo Cruz Foundation, Campo Grande, MS, Brazil.

Chang Min Park (CM)

Department of Radiology, Seoul National University College of Medicine, Seoul, Korea.
Department of Radiology, Seoul National University Hospital, Seoul, Korea.

Eui Jin Hwang (EJ)

Department of Radiology, Seoul National University College of Medicine, Seoul, Korea.
Department of Radiology, Seoul National University Hospital, Seoul, Korea.

Jason R Andrews (JR)

Division of Infectious Diseases and Geographic Medicine, Stanford University School of Medicine, Stanford, CA, United States of America.

Julio Croda (J)

Oswaldo Cruz Foundation, Campo Grande, MS, Brazil.
Department of Epidemiology of Microbial Diseases, Yale University School of Public Health, New Haven, CT, United States of America.
School of Medicine, Federal University of Mato Grosso do Sul, Campo Grande, MS, Brazil.

Classifications MeSH