Computer-aided detection (CAD) of TB
|Read our new technology landscape report on Digital chest radiography and computer-aided detection (CAD) solutions for TB diagnostics.|
What is this project?
This project is designed to understand the potential of computer-aided detection (CAD) technologies to improve TB diagnosis using chest X-rays. High-quality evidence on the performance of available CAD technologies for TB is needed to inform policy guidance on their use.
Why are we working on it?
Chest X-ray is an important tool for triage (a process to identify risk and select the most appropriate care pathway when a person presents with symptoms) and proactive screening for pulmonary TB in adults, children and people living with HIV. Alone, it does not lead to a confirmative diagnosis of TB, but it is a highly sensitive tool that can pick up early forms of TB, including in people without symptoms. Accurate, fast information at this first step can potentially reduce the number of tests and costs associated with confirmatory testing.
One of the barriers for large-scale implementation of chest X-ray is the need for trained human readers, who can be scarce in many high TB burden countries. CAD-based technologies rely on artificial intelligence (AI) systems that learn from large amounts of data. To learn whether a chest X-ray could be showing signs of TB, hundreds of thousands of images with and without TB are read by the software. A new chest X-ray is then compared with that body of “experience”, and attributed a score; above a certain threshold, this score indicates that the person should be examined further. Using CAD for this interpretation of chest X-ray images either as a replacement for human readers or as a first triaging step, enables more people to be screened for TB.
What does it involve?
From 2017, we have been collating a first-of-its-kind global archive of chest X-ray images, including those of people from different geographical regions, with different risk factors, and with TB or other lung pathologies. This archive contains images from both triage and screening situations. We continue to expand the archive in collaboration with several partners who are contributing to the archive with chest X-ray images and data from their studies, and we are also using it to conduct independent evaluations of the performance of available CAD technologies to identify pulmonary TB.
To inform TB researchers and implementers about the CAD technologies that are available today, together with the Stop TB Partnership, we have developed the ai4hlth resource centre, which currently includes 11 CAD tools: nine already on the market with two more in validation.
We are also studying the utility of CAD for other use cases, including its ability to detect early forms of TB (subclinical TB) or for diseases other than TB.
What do we expect to achieve?
Evaluating additional CAD technologies and new versions of existing products when these enter the market supports policy development and regulatory decisions.
Results from an evidence report that we prepared for WHO, assessing the performance of three CAD technologies using data from the archive, has already informed a recommendation as part of the updated WHO screening guidelines, which stipulate that CAD may be used as an alternative to human reader interpretation of chest X-ray for pulmonary TB in individuals aged 15 years or more.
What is the timescale?
This work is expected to continue until 2022.
Partners and funding
This project is being implemented by FIND and our partners, including (in alphabetical order): Amsterdam Institute for Global Health and Development, Academic Medical Center, Amsterdam, The Netherlands; Aurum Institute, Parktown, South Africa; Bamenda Regional Hospital, Bamenda, Cameroon; Bill and Melinda Gates Foundation, Seattle, US; Center for Disease Control and Prevention, Atlanta, US; Departments of Epidemiology, Biostatistics & Occupational Health, McGill University, Montreal, Canada; Department of Medicine, Faculty of Medicine, McGill University, Montreal, Canada; Department of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden; GGD Groningen, Groningen, Netherlands; International Organization for Migration (IOM), Geneva, Switzerland; International Organization for Migration (IOM), Manila, Philippines; Japan Anti-Tuberculosis Association (JATA), Tokyo, Japan; KNCV Tuberculosis Foundation, The Hague, The Netherlands; Liverpool School of Tropical Medicine, Liverpool, United Kingdom; London School of Hygiene and Tropical Medicine, London, United Kingdom; Ludwig Maximilian University of Munich, Munich, Germany; Research Center Borstel, Sülfeld, Germany; Tuberculosis Reference Laboratory Bamenda, Bamenda, Cameroon; University College London, London, United Kingdom; University of Cape Town Lung Institute, Cape Town, South Africa; University Medical Center Groningen, Groningen, Netherlands; University of Heidelberg, Heidelberg, Germany; and University of Oxford, Oxford, UK.
The project is funded by a grant from the Ministry of Foreign Affairs of the Netherlands.
For more information please email@example.com.