Last Updated:
24 August 2021

CAD4TB is CE certified since 2015 by the stringent Dutch regulatory body and new releases come out on a yearly basis. To date, more than 8 million people have been screened with the software across 40 countries.
CAD4TB is a state-of-the-art artificial intelligence (AI) software product that takes a frontal chest radiograph as input and produces several outputs which include an automated quality assessment of the chest image, a heat map indicating possible areas of abnormalities, and a score between 0 and 100 related to the likelihood that the X-ray image demonstrates abnormalities associated with tuberculosis (TB). It can work both online with CAD4TBCloud and offline with the CAD4TBbox. The offline CAD4TB solution is highly relevant for settings with high daily throughput, limited or no stable internet connectivity or for programmes coping with restrained budgets for data exchange over the mobile phone network.
CAD4TB is mainly used as a rapid and accurate triaging tool for TB disease and does not require skilled human readers to determine if the subject is a presumptive TB case requiring further confirmative testing. Based on the set threshold of the CAD4TB tool, local staff can ask the presumptive TB patients to submit sputum and immediately perform a GeneXpert MTB/RIF test. Abnormal chest images from patients who were Xpert MTB/RIF-negative can be read by a skilled human reader to determine what steps to take next (for a clinical diagnosis of TB and/or other lung diseases).
The latest version, CAD4TB 7, has been commercially available since April 2021 with enhanced specificity, among other features, compared with previous versions to further increase savings on Xpert MTB/RIF cartridges. The latest CAD4TB platform also assists TB programmes with overall improved AI performance, an integrated TB screening portal, and advanced management monitoring tools. The new features include digital modules for registration, symptom screening with easy to modify questionnaires, CAD4TB 7 score, patient reporting, and the possibility to seamlessly integrate data from GeneXpert systems. Furthermore, an ‘insights’ module shows statistical representations of the data, including distribution of CAD4TB results based on demographics, such as gender and age.
Additionally, CAD4TB can be complemented with the CAD4COVID-XRay software (CE certified) to support triaging COVID-19 and to enable bi-directional screening for TB and COVID-19.
Certification
CE 0344 marked Class IIb
Development Stage
On the market. Technology readiness level 9
Deployment
Online & offline
Intended Age Group
4+ years
Target Setting
Outreach and mobile health services, primary healthcare facilities, general hospitals (above primary level), occupational health programmes, teleradiology companies, government/public sector, e.g. national TB programme, private sector
Current Market
CAD4TB is in use in over 40 countries worldwide. A complete list can be found on the website: https://www.delft.care
Input
»» Can be used to read images from any kind of chest X-ray machine
»» Chest X-ray image format: DICOM, PNG, JPEG.
By using an app (SNAP4CAD), analogue X-ray images can be used as well.
»» Chest X-ray type: PA (posterior-anterior) chest X-ray (from stationary, portable, and ultra-portable X-ray systems).
Output
The CAD4TB results consist of:
»» Abnormality score for TB
»» Heat map
»» Binary classification “TB” or "not TB"
The new CAD4TB platform includes a full report with screening results as well as an advanced management dashboard to monitor screening progress.




Hardware
CAD4TB comes as a complete product and is installed on a mini computer called the CAD4TBbox.
Validation
The software has been validated on a variety of X-ray machines (over 30 different machines), so a separate validation for each specific X-ray machine is not mandatory. However, as a free-of-charge optional service, Delft Imaging can advise the user on the threshold setting for optimal diagnostic output for their specific TB screening programmes also taking the available budget and operational setting into account.
Hardware
CAD4TB comes as a complete product and is installed on a mini computer called the CAD4TBbox.
Server
All leading cloud computing providers are supported. In country or on-premise servers are available upon request.
Integration
For version 6 onwards, it is possible to send the heat map and abnormality score in a DICOM format to a picture archiving and communication system (PACS).
Software
CAD4TB 7 runs on Linux. Currently, Ubuntu LTS 16.04 or newer is preferred.
Processing Time
Less than 20 seconds for the CAD4TBbox. Cloud result processing time depends on internet connectivity.
Data Sharing & Privacy
»» Server location: All leading cloud computing providers are supported – with multiple data centers around the world. In country or on-premise servers are also available upon request.
»» No data are shared with the developer. As a company headquartered in the Netherlands, the developer follows the General Data Protection Regulation (GDPR) of the European Union. Each customer receives a Data Processing Agreement that provides clear insight on how their data are processed and used. The developer does not use the data for any other purpose than to provide the CAD4TB result for the customer, unless given explicit permission in writing. Any data processed by CAD4TB remain the property of the customer.
»» Data are de-identified before they are sent to the server in order to make sure no personal information is sent over the internet, ensuring compliance with privacy regulations.
Price
»»CAD4TB is available in two ways (per-use basis or perpetual license). Perpetual license for CAD4TB software, CAD4TBbox and related services are available on the Stop TB Partnership’s Global Drug Facility (GDF) catalogue. Please see the Diagnostics Catalogue of GDF for more details.
»» In parallel, CAD4TB is sold on a per use basis and available offline and/or online. Bundle prices depend on volume and typically range from US $ 0.45 to US $ 0.95 depending on the order quantity (higher-volume orders are available at a lower price).
»» Offline: Upfront cost for a CAD4TBbox is US $ 6000 (for 1 CAD4TBbox, prices decrease based on quantity) and is inclusive of shipment, remote support and 2-year warranty on the hardware.
»» Online: Upfront cost for cloud service (for primary or secondary review and/or archiving purposes) is US $2 500 and entails the setup of the cloud server, the installation of the upload programme that allows images to be sent, compressed and de-identified if required, and remote support.
»» There is no pricing difference for public or private sector.
»» Prices can be further customized on a project basis depending on project objectives and goals.
Software Updates
»» Annual software updates
»» Updates are included in the price
»» No additional costs, as long as hardware permits
Product Development Method
Supervised deep learning (CNN, RNN) plus manual feature engineering
Training
The product was trained on 1 million chest X-rays from several countries and continents
Reference Standard
Human reader, culture and GeneXpert MTB/RIF results
Publications
Over 50 scientific publications and references. A complete list of all publication related to CAD4TB can be found on https://www.delft.care/publications/
Recent publications related to CAD4TB:
»» Murphy, K., Habib, S.S., Zaidi, S.M.A. et al. Computer aided detection of tuberculosis on chest radiographs: An evaluation of the CAD4TB v6 system. Sci Rep. 2020;10:5492. https://doi.org/10.1038/s41598-020-62148-y
»» Habib, S.S., Rafiq, S., Zaidi, S.M.A. et al. Evaluation of computer aided detection of tuberculosis on chest radiography among people with diabetes in Karachi Pakistan. Sci Rep. 2020;10:6276. https://doi.org/10.1038/s41598-020-63084-7
»» Kim, H-Y., Zishiri, V., Page-Shipp, L. et al. Symptom and digital chest X-ray TB screening in South African prisons: yield and cost-effectiveness. International Union Against Tuberculosis and Lung Diseases. 2020;24(3):295–302. https://doi.org/10.5588/ijtld.19.0214
»» Khan, F.A., Majidulla, A., Tavaziva, G. et al. Chest x-ray analysis with deep learning-based software as a triage test for pulmonary tuberculosis: a prospective study of diagnostic accuracy for culture-confirmed disease. Lancet Digit Health. 2020;2(11):e573-e581. https://doi.org/10.1016/S2589-7500(20)30221-1
»» Wali, A., Khan, D., Safdar, N. et al. Prevalence of tuberculosis, HIV/AIDS, and hepatitis; in a prison of Balochistan: a cross-sectional survey. BMC Public Health. 2019;19:1631. https://doi.org/10.1186/s12889-019-8011-7
This website works best with browsers other than Internet Explorer.