FIND is seeking collaborators who are willing to provide chest-X-rays and clinical data from finalized projects or studies to the ECAD-TB project, to contribute to the evaluation of CAD4TB and other CADs to inform WHO policy guidance on this topic.
What is the ECAD-TB project?
A large proportion of persons with active tuberculosis (TB) do not have classical TB symptoms, while TB abnormalities can be detected early on using chest X-ray (CXR). There is growing evidence of the diagnostic utility of computer-aided detection (CAD) for TB diagnosis using CXR. Chest X-rays are an important tool to screen and triage individuals for TB in both resource-rich and resource-limited settings. However, in many high-disease burden countries, access to CXRs of acceptable quality and, more importantly, human resources to read the CXRs, remain limited. CAD-based technologies for the automated interpretation of CXRs have been gaining interest. At a recent World Health Organization (WHO)-sponsored meeting, it was decided that additional high-quality data on the utility of CAD for both triage testing and screening for TB are needed to issue formal WHO recommendations.
FIND and a number of FIND partners are working to generate such data under a project called “evaluation of CAD solutions for TB detection” (ECAD-TB project). The aim of the ECAD-TB project is to establish a standard panel of CXR files that include a representative spectrum of TB-associated lung pathology, pathology of relevant differential diagnoses, and distribution of patients according to their gender, age, HIV-status and geographical origin. This standard panel will contain images that were taken because of a triage use-case (diagnostic work-up in individuals seeking care owing to symptoms) or screening use-case (due to health provider–initiated, systematic screening activity). The standard panel will be used to conduct an independent evaluation of the performance of current and future CAD solutions to identify abnormalities suggestive of TB.
What kind of project / studies are we looking for?
FIND is looking for finalized projects and studies that have collected digital CXR images (DICOM files) and demographic and clinical information from individuals that underwent CXR because of screening or triage activity, defined as follows:
Screening projects: CXR is performed to evaluate an individual regardless of symptoms, and who is not seeking care by him/herself due to TB symptoms. Screening use-cases are also sometimes referred to as health provider-initiated screening or active case finding in certain high risk groups.
Triage projects: CXR is performed to evaluate an individual with signs and symptoms of TB, who is seeking care. Triage use-cases have also been described as patient-initiated evaluations or passive case finding.
Does your project of study meet our criteria? See questions below.
|Eligibility criteria for project / studies contributing data to the ECAD-TB project||YES / NO|
|Individuals in your screening project are screened irrespective of whether they had symptoms (e.g. screening was provider-initiated and not patient-initiated).
Individuals in your triage projects had signs and symptoms compatible with active TB and were seeking care (e.g. patient-initiated evaluation).
|CXR images in the form of DICOM files are available for a cohort of individuals who were tested within a defined period.|
|DICOM files have NOT been shared with CAD software developers for the use of training CAD4TB or any other CAD software. DICOM files may be read with the CAD software, but proof is required from the collaborator and CAD developer that these images were not shared and used to train the CAD application.|
|As a minimum, the following variables are known for all individuals in your cohort from which a sample will be drawn: age, gender and, if done, test result of smear, culture and/or Xpert are required.|
|You are willing to create de-identified DICOM files which only include a unique identification number that can link the DICOM file with the record containing the corresponding de-identified demographic and clinical information of an individual.|
|You are willing to prepare a dataset with personal level data of the sampled individuals, containing the same unique identification number as on the de-identified DICOM image.|
|You have obtained the required ethical approval which allows you to share de-identified DICOM files and clinical data for the purpose of the ECAD-TB project.|
|You agree that sharing of data will be done in full compliance with any applicable laws and regulations governing protected health information and other personal data (including the upcoming EU General Data Protection Regulation; www.eugdpr.org).|
If you have answered YES to all the questions and wish to collaborate in our ECAD-TB project,
please contact us at: Sandra.Kik2@finddx.org
What do we expect from our collaborators?
- Agree with FIND on the collaboration, data sharing and usage agreement (the “Agreement”).
- Sign the Agreement with FIND.
- Review and agree on the study protocol, as prepared by FIND.
- Agree with FIND on study or projects that are eligible and for which data will be contributed.
- Fill out sampling table with total number of participants (cases and non-cases) for each identified study or project (Appendix B of the study protocol).
- Agree with FIND on timelines for preparation and submission of data.
- Upon receipt of sampling number, fetch the agreed upon sample of DICOM files.
- De-identify DICOM files (remove them from any personal identifying information) and keep or assign a unique person identification number that can link the file to the demographic and clinical data for the same person.
- Prepare a dataset (in .csv or other agreed upon format) with individual level data (variables outlined in Appendix C of the study protocol) of the sampled individuals. This dataset should be separated from any personal identifying information, but contain the unique person identification number that can link the DICOM file to the information in the dataset.
- Prepare a document outlining the study level information, including a description of the operationalization (definitions and coding) of the variables in the individual level dataset.
- Upload the data to the secure FIND server and notify us.
- Review draft versions of the study results that are prepared by FIND and be available to answer interim questions related to the data provided.
- Review, comment and approve final draft of the study report.
- Review, comment and approve final manuscript for publication.
What do we offer collaborators in return?
- Each institute is allowed to contribute up to 2 authors on any publication that arises from the ECAD-TB study using the data from your institute. For authorship rules we will follow the international standards on authorship, which can be found here: http://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html.
- In addition, FIND will pay a compensation fee of 10 USD per record.