Product:

TiSepX TB

Version 1

Company HQ: 

Seoul, South Korea

Download Product Profile

Last Updated:

September 2nd, 2021

Determining the activity of pulmonary tuberculosis on chest radiographs is difficult. Medical IP's TiSepX TB deep learning model could determine the activity of tuberculosis and differentiate active TB from healed TB on chest radiographs comparable to expert clinicians. The model-driven activity score reflected bacilli burden and treatment response.
Medical IP has also other products available that allow for the detection of COVID-19 or pneumonia.

Certification

Pending (Korea MFDS)

Development Stage

On the market

Deployment

Completely online, Completely offline

Intended Age Group

20+ years

Target Setting

Primary health centres, General hospital (above primary level), Teleradiology companies, Government/public sector, e.g. national TB program, Private sector, High-TB burden countries

Current Market

South Korea, Less Developed Countries with ODA programs

Input

Chest X-ray image format: JPEG, PNG, DICOM, Neuroimaging Informatics Technology Initiative files (NII), bitmap (BMP)

Chest X-ray type: Posterior-anterior CXR, Anterior-posterior CXR, portable

Output

Heat map, Probability score for identifying active TB from healed TB / Score trajectory during anti-TB treatment


Lung findings identified by AI:

Algorithm does not specify abnormalities


Lung findings included in TB score:

Abscess, Airfluid level, Blunted costophrenic angle,   Cavity, Chest wall invasion/destruction, Consolidation, Interstitial   markings, Loculated pleural effusion, Lymphadenopathy, Mass, Nodule, Opacity,   Pleural effusion, Prominence in hilar region, Probability score for   identifying active TB from healed TB


Disease scores identified (*detection of other diseases on CXR beyond TB could come with additional cost):

COVID-19, pneumonia


Format: PNG

ai4hlthlogoNEWtoUse.png

Hardware

Online: Stable internet access and internet browsers (i.e. Chrome, Safari) are enough to execute the program for online use.
Offline: Memory supporting Compute Unified Device Architecture (CUDA)

Hardware Requirements: TiSepX Lung / TiSepX COVID / TiSepX TB -> 8GB GPU

Validation

If the product is integrated with the machine (works as a part of the machine), a separate validation is normally required. If the product works as a separate software, the X-ray machine does not require validation before use. The product works well with X-ray images using most digital and computed radiographic machines but cannot be assured if X-ray is acquired using film radiography.

Hardware

Online: Stable internet access and internet browsers (i.e. Chrome, Safari) are enough to execute the program for online use.
Offline: Memory supporting Compute Unified Device Architecture (CUDA)

Hardware Requirements: TiSepX Lung / TiSepX COVID / TiSepX TB -> 8GB GPU

Server

Main server is located in South Korea, a local server can be set up if required.

Integration

Software

Software Requirements: NVIDIA CUDA required
Main server is located in South Korea, a local server can be set up if required.

Processing Time

10-20 seconds (depending on the X-ray image resolution and size)

Data Sharing & Privacy

Server location (for online product):
Main server is located in South Korea, a local server can be set up if required.

Data shared with manufacturer? No
De-identification (option to deidentify?): Yes
The TiSepX TB server anonymizes all personal information before uploading original images to the server.

Price

Volume-based pricing models are available.
Please contact company for quote: sales@medicalip.com

Software Updates

As soon as the module is developed, an update is made after the test. Regular updates happen twice a year, but irregular updates might happen if abnormality function is discovered.
Please contact company for information: tech@medicalip.com

Product Development Method

Unsupervised deep learning (AEs/SAE, RBMs, DBNs)

Training

9836 adult CXRs

Reference Standard

Human reader, Chest CT

Publications

Seowoo Lee, Soon Ho Yoon, et al. Deep Learning to Determine the Activity of Pulmonary Tuberculosis on Chest Radiographs. Radiology. 2021. Available from: https://doi.org/10.1148/radiol.2021210063

Does this page require updates? Send us a message:

Designed by Tasneem Naheyan

This website works best with browsers other than Internet Explorer.