Last Updated:
25th January 2025
Lunit INSIGHT CXR is an artificial intelligence solution that detects 10 different radiologic findings on chest X-rays. It also comes with a dedicated TB engine, which generates a TB Analysis Score for enhanced TB screening performance. In a head-to-head study conducted by International Organization for Migration using a global multicenter database, Lunit INSIGHT CXR demonstrated the highest accuracy (AUC) when analyzed against microbiological reference standard(MRS).1
The WHO Consolidated Guidelines on Tuberculosis, published in 2021, listed Lunit INSIGHT CXR as one of the computer-aided detection software that can be used in place of human readers for TB screening in individuals aged 15 years and older.
1. Gelaw, Sifrash Meseret, et al. “Diagnostic Accuracy of Three Computer-Aided Detection Systems for Detecting Pulmonary Tuberculosis on Chest Radiography When Used for Screening: Analysis of an International, Multicenter Migrants Screening Study.” PLOS Global Public Health, 2022
Certification
CE MDR, Class IIa
Korea MFDS
Development Stage
On the Market
Deployment
Online & Offline
Intended Age Group
CE MDR : 4 years or older
Korea MFDS : 14 years or older
Target Setting
Primary health centres, general hospital (above primary level), teleradiology companies, government/public sector, examples national TB programmes, private sector
Current Market
55+ countries
Input
Can read images from any digital CXR machine or model
Chest X-ray image format: DICOM
Chest X-ray type: posterior-anterior chest X-ray, anterior-posterior chest X-ray
Output
Output includes:
Heatmap
Dichotomous output indicating whether TB is likely present or absent
Dichotomous output indicating whether each abnormality is likely present or absent
Probability score for TB
Probability score for each abnormality
Location of each abnormality
Auto Comparison for comparing current and prior images (Pleural Effusion, Consolidation, Pneumothorax)
The default cut-off probability score is 15 and this can be adjusted for each radiologic finding (15, 20, 25, 30, 35, 40, 45)
Results are formatted in a structured report.
The TB Score is determined by a dedicated TB engine that is separate from the engine for the other abnormalities
Additional findings reported by the product: Atelectasis, Calcification, Consolidation, Fibrosis, Mass, Nodule, Opacity, Pleural effusion, Pneumothorax, Cardiomegaly, Mediastinal widening, Pneumoperitoneum.
Hardware
Please contact contact@lunit.io for information.
Server
Server Operating Environment
OS Windows 10, version 21H2 / Windows 10 Enterprise LTSC 2021
Linux kernel 3.10 or higher
CPU i5-9500 (6core, 3.0GHz)
RAM 16 GB
Disk SSD 256GB
GPU NVIDIA T400 4GB
Network Environment 100 Mbps or higher
Integration with X-ray Systems
Integration with PACS and Legacy Systems
Lunit INSIGHT CXR can be integrated with various x-ray systems such as FujiFilm, GE, Samsung, Carestream, etc. If the system can send/receive images via DICOM protocol, Lunit INSIGHT CXR can be directly integrated with the system. Otherwise, Lunit INSIGHT CXR can be integrated via PACS or API.
Lunit INSIGHT CXR can be integrated with various PACS such as FujiFilm, Agfa, GE, Philips, etc. which communicates via DICOM C-Store.
Software
Please contact contact@lunit.io for information.
Processing Time
About 5-20 seconds per X-ray
Data Sharing & Privacy
Server location: Amazon Web Services is used. Local or national servers can be set up if required
Data are not shared with the developer
There is an option to de-identify data. The Lunit DICOM Gateway anonymizes all personal information before transferring original images to the analysis server
Software Updates
Software is updated at least once a year.
When Lunit INSIGHT CXR is updated, Lunit or its distribution partner will inform the customer about the update details.
Product Development Method
The artificial intelligence CADe software, Lunit INSIGHT CXR, was developed based on deep learning algorithms. As a subset of machine learning, deep learning is designed to mimic the human brain with neurons that transmit information to the next neurons using synaptic strength.
Training
The deep learning algorithms of Lunit INSIGHT CXR have been trained by approximately 220,000+ chest radiography cases
Reference Standard
For data curation, radiologists reviewed all chest radiography images and excluded some images with an incorrect label. To prevent the AI algorithms from missing small radiologic findings, the ground truth for training dataset is based on chest CT images of each patient.
This website works best with browsers other than Internet Explorer.