Breaking the threshold: developing multivariable models using computer-aided chest X-ray analysis for tuberculosis triage.

chest radiography deep learning prediction modelling tuberculosis

Journal

International journal of infectious diseases : IJID : official publication of the International Society for Infectious Diseases
ISSN: 1878-3511
Titre abrégé: Int J Infect Dis
Pays: Canada
ID NLM: 9610933

Informations de publication

Date de publication:
02 Sep 2024
Historique:
received: 31 03 2024
revised: 01 08 2024
accepted: 22 08 2024
medline: 5 9 2024
pubmed: 5 9 2024
entrez: 5 9 2024
Statut: aheadofprint

Résumé

Computer-aided detection (CAD) software packages quantify tuberculosis-compatible chest X-ray (CXR) abnormality as continuous scores. In practice, a threshold value is selected for binary CXR classification. We assessed the diagnostic accuracy of an alternative approach to applying CAD for tuberculosis triage: incorporating CAD scores in multivariable modelling. We pooled individual patient data from four studies. Separately for two commercial CAD, we used logistic regression to model microbiologically-confirmed tuberculosis. Models included CAD score, study site, age, sex, HIV status, and prior tuberculosis. We compared specificity at target sensitivities ≥90% between the multivariable model and the current threshold-based approach for CAD use. We included 4733/5640 (84%) participants with complete covariate data (median age 36 years; 45% female; 22% with prior tuberculosis; 22% people living with HIV). A total of 805 (17%) had tuberculosis. Multivariable models demonstrated excellent performance (areas under the receiver operating characteristic curve (95%CI): software A, 0.91 (0.90-0.93); software B, 0.92 (0.91-0.93)). Compared to threshold scores, multivariable models increased specificity (e.g. at 90% sensitivity, threshold vs model specificity (95%CI): software A, 71% (68%-74%) vs. 75% (74%-77%); software B, 69% (63%-75%) vs. 75% (74%-77%)). Using CAD scores in multivariable models outperformed the current practice of CAD-threshold-based CXR classification for tuberculosis diagnosis.

Sections du résumé

BACKGROUND BACKGROUND
Computer-aided detection (CAD) software packages quantify tuberculosis-compatible chest X-ray (CXR) abnormality as continuous scores. In practice, a threshold value is selected for binary CXR classification. We assessed the diagnostic accuracy of an alternative approach to applying CAD for tuberculosis triage: incorporating CAD scores in multivariable modelling.
METHODS METHODS
We pooled individual patient data from four studies. Separately for two commercial CAD, we used logistic regression to model microbiologically-confirmed tuberculosis. Models included CAD score, study site, age, sex, HIV status, and prior tuberculosis. We compared specificity at target sensitivities ≥90% between the multivariable model and the current threshold-based approach for CAD use.
RESULTS RESULTS
We included 4733/5640 (84%) participants with complete covariate data (median age 36 years; 45% female; 22% with prior tuberculosis; 22% people living with HIV). A total of 805 (17%) had tuberculosis. Multivariable models demonstrated excellent performance (areas under the receiver operating characteristic curve (95%CI): software A, 0.91 (0.90-0.93); software B, 0.92 (0.91-0.93)). Compared to threshold scores, multivariable models increased specificity (e.g. at 90% sensitivity, threshold vs model specificity (95%CI): software A, 71% (68%-74%) vs. 75% (74%-77%); software B, 69% (63%-75%) vs. 75% (74%-77%)).
CONCLUSIONS CONCLUSIONS
Using CAD scores in multivariable models outperformed the current practice of CAD-threshold-based CXR classification for tuberculosis diagnosis.

Identifiants

DOI: 10.1016/j.ijid.2024.107221 PMID: 39233047
pubmed: 39233047
pii: S1201-9712(24)00292-3
doi: 10.1016/j.ijid.2024.107221
pii:
doi:

Types de publication

Journal Article

Langues

eng

Sous-ensembles de citation

IM

Pagination

107221

Informations de copyright

Copyright © 2024. Published by Elsevier Ltd.

Auteurs

Coralie Geric (C)

McGill International TB Centre, Research Institute of the McGill University Health Centre, Montreal, Canada; Department of Medicine, McGill University, Montreal, Canada; Respiratory Epidemiology and Clinical Research Unit, Centre for Outcomes Research and Evaluation, Research Institute of the McGill University Health Centre, Montreal, QC, Canada.

Gamuchirai Tavaziva (G)

McGill International TB Centre, Research Institute of the McGill University Health Centre, Montreal, Canada; Respiratory Epidemiology and Clinical Research Unit, Centre for Outcomes Research and Evaluation, Research Institute of the McGill University Health Centre, Montreal, QC, Canada.

Marianne Breuninger (M)

Division of Infectious Diseases, Department I of Internal Medicine, University of Cologne, Cologne, Germany.

Keertan Dheda (K)

Centre for Lung Infection and Immunity Unit, Division of Pulmonology and UCT Lung Institute, University of Cape Town, Cape Town, South Africa; Faculty of Infectious and Tropical Diseases, Department of Infection Biology, London School of Hygiene and Tropical Medicine, London, United Kingdom.

Ali Esmail (A)

Centre for Lung Infection and Immunity Unit, Division of Pulmonology and UCT Lung Institute, University of Cape Town, Cape Town, South Africa.

Alex Scott (A)

Centre for Lung Infection and Immunity Unit, Division of Pulmonology and UCT Lung Institute, University of Cape Town, Cape Town, South Africa.

Mary Kagujje (M)

Tuberculosis Department, Centre for Infectious Disease Research in Zambia, Lusaka, Zambia.

Monde Muyoyeta (M)

Tuberculosis Department, Centre for Infectious Disease Research in Zambia, Lusaka, Zambia; Zambart, Lusaka, Zambia.

Klaus Reither (K)

Swiss Tropical and Public Health Institute, Allschwill, Switzerland; University of Basel, Basel, Switzerland.

Aamir J Khan (AJ)

IRD Global, Singapore.

Andrea Benedetti (A)

McGill International TB Centre, Research Institute of the McGill University Health Centre, Montreal, Canada; Department of Medicine, McGill University, Montreal, Canada; Department of Epidemiology, Biostatistics & Occupational Health, McGill University, Montreal, Canada.

Faiz Ahmad Khan (FA)

McGill International TB Centre, Research Institute of the McGill University Health Centre, Montreal, Canada; Department of Medicine, McGill University, Montreal, Canada; Respiratory Epidemiology and Clinical Research Unit, Centre for Outcomes Research and Evaluation, Research Institute of the McGill University Health Centre, Montreal, QC, Canada. Electronic address: faiz.ahmadkhan@mcgill.ca.

Classifications MeSH