This website works best with browsers other than Internet Explorer.

Product:

​

Version 1

XrayAME

Company HQ: 

Antwerp, Belgium

Last Updated:

April 23rd, 2020

XrayAME is a computer-aided diagnostic for chest X-rays that can be used as screening and triage test to identify tuberculosis (TB) patients from chest X-rays in under-resourced communities.

Certification

Pending (expected Q1 2021)

Development Stage

Online web interface: on the market; mobile application: under development (targeted for release in Q3 2020)

Deployment

Online & offline

Intended Age Group

18+ years

Target Setting

Primary health centers, teleradiology companies, government/public sector, e.g. national TB program

Current Market

Pakistan, the Philippines, South Africa, Brazil, Uganda, Cambodia

Input

»» Compatible with any digital chest X-ray machine (vendor neutral)

»» Chest X-ray image format: JPEG, PNG, DICOM

»» Chest X-ray type: posterior-anterior chest X-ray, anterior-posterior chest X-ray

»» Other requirements: none

Output

Structured report including: 

»» Heat map, 

»» Dichotomous output only indicating whether TB is likely present or likely absent, 

»» Probability score for TB 

»» The default threshold score for "Highly likely of TB" is 75. Any score above 40 is considered "Suspicious". Any score below 40 is considered "Unlikely TB". All score thresholds can be adjusted per application based on the required True Positive and False Positive Rate at a given threshold and for a given patient population.

Server

The cloud server requires 32 GB RAM, 8 cores and 500 GB disk space. The offline device requires 4 GB RAM.

Integration

EPCON uses DICOM C-Store and all other legacy integrations will need to be discussed on a case-by-case basis.

Software

Any modern web browser

Processing Time

20 seconds

Hardware

For cloud solution: a web interface running on a desktop PC or tablet.
For offline solution: a device with power supply that hosts a DICOM C-Store and local user interface is provided.

Validation

Recommended: calibration of score thresholds of the software output based on a reference set (500) of classified images obtained from the given machine and patient population

Data Sharing & Privacy

»» Server Location (for online product): current server located in South Africa, but local or national servers can be set up if required

»» No DICOM pixel data or metadata are shared with the manufacturers during use of the product

»» De-identification: No DICOM metadata is saved by the tool and only the extracted images are displayed. However, identifying information printed on the image is not removed

Software Updates

»» Quarterly software updates

»» Updates are included in the license fee. No cost difference between public and private sector

»» Extra costs: none. Updates are included in support costs

Price

»» Pricing is adjusted in line with project scope and availability of budget. In a commercial setting, transactional pricing models can be applied. For social impact projects EPCON supports free CAD readings as part of the CAD4Good program

»» Public/private pricing difference: hardware and travel costs are required for the offline solution. Support requirements may be more in offline settings

»» Upfront installation/set up costs: travel and installation support of offline devices. Server set-up and running costs for national or local installations. Offline device installation includes hardware and set-up costs and range from €1000 to €5000

Product Development Method

Supervised deep learning (CNN, RNN)

Training

The product was trained on 10 000 chest X-rays from South Africa, Nepal and China

Reference Standard

GeneXpert and human reader

Publications

»» Peer-reviewed publications are not yet available.

Does this page require updates? Send us a message:

Designed by Tasneem Naheyan

© 2020 Stop TB Partnership