Last Updated:
15 April 2025
Nexus AI CXR 1.0 is computer-aided detection (CADe) software intended for use in the evaluation of chest X-rays (CXR). The software assists in the detection of Normal X-rays and radiological abnormalities suggestive of pulmonary Tuberculosis (TB).
Nexus AI CXR 1.0 is designed to function as a decision-support tool within the radiology workflow in both clinical and screening environments. It enhances efficiency by providing the intended users with AI-generated insights to support the interpretation of chest X-rays but does not replace individual CXR reporting by a trained professional.
Nexus AI CXR 1.0 can be deployed locally (on-premises) or cloud-based, used within the radiology workflow and is designed to streamline and optimize workflow management, assisting users to evaluate findings and refer cases effectively, and responsibly when needed.
Nexus AI CXR 1.0 seamlessly integrates with DICOM compatible systems and devices offering full integration and interoperability.
Certification
CE IIb Certification pending with TÜV SÜD
Development Stage
On the Market
Deployment
Cloud or On-premises or Hybrid
Intended Age Group
15+ years
Target Setting
Nexus AI CXR can be deployed in clinics, hospitals, mobile units, immigration and healthcare settings. It supports active case finding, prevalence surveys, and lung health screening, making it ideal for diverse health and clinical use cases
Current Market
Africa
Input
Can be used to read images from any chest X-ray machine and model
Chest X-ray image format: DICOM, JPG, PNG
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
0.5 is the default threshold and can be adjusted based on requirements by clients' TB screening program.
The results are formatted in a structured PDF report. The Nexus AI CXR report includes patient details, findings summary, CAD scores, and annotated images, ensuring clarity and clinical usability.
Lung abnormalities included in the TB Score: Abscess, Air fluid level, Atelectasis, Bronchiectasis, Cavity, Consolidation, Fibrosis, Hyperinflation, Interstitial markings, Mass, Nodule, Opacity, Pleural effusion, Prominence in hilar region, Pneumothorax
The ML model is trained using X-rays with microbiological test reports and radiologist reports as reference standards
Additional findings reported by the product: Abscess, Air fluid level, Atelectasis, Bronchiectasis, Cavity, Consolidation, Fibrosis, Hyperinflation, Interstitial markings, Mass, Nodule, Opacity, Pleural effusion, Prominence in hilar region, Pneumothorax
The ML algorithm provides a general lung abnormality based on patterns from training data. It does not specify individual findings but indicate non-normal X-rays for clinical evaluation
Hardware
The Nexus AI CXR software is available as a fully offline solution, with the CAD AI installed directly on a local computer or laptop. Recommended hardware includes a multi-core processor, at least 32 GB RAM, and 1 TB of storage.
Server
The Nexus AI CXR software is available on cloud servers and can be deployed in-country (MS Azure, AWS Cloud, Google Cloud etc.) or on-premise services that meet required specifications.
Integration with X-ray Systems
Integration with PACS and Legacy Systems
Nexus fully supports DICOM standard and can integrate and receive X-Ray files transmitted via DICOM protocol from any digital (CR/DR/DDR) X-Ray system (fixed, mobile, portable, ultra-portable)
The Nexus AI CXR software can integrate to any DICOM compliant PACS or imaging systems. The AI results can aslo be sent to third party EMR and EHR systems.
Software
For on-premise Windows 10 or newer or Ubuntu 22 or newer. If using Nexus cloud services any compatible web browser (Chrome, Edge, Safari etc.)
Processing Time
Under 60 seconds depending on hardware and internet speed.
Data Sharing & Privacy
Nexus follows international data privacy standards. Data is de-identified and not shared with developers by default. A full data sharing and privacy policy is in place to ensure secure, responsible handling of all information.
Software Updates
Software updates, including routine upgrades and performance improvements, are provided annually or as needed based on post-market surveillance and system enhancements. Critical patches and newly developed modules are released promptly following successful internal testing and validation.
Price
Nexus provides volume and license based price options. Contact info@nexus-health.ai for pricing.
Product Development Method
Nexus Intelligence (Pty) Ltd collaborated with Google LLC (USA) to integrate pre trained machine learning (ML) models developed by Google for the analysis of chest radiographs, specifically for the identification of TB and Normal Chest X-rays.
Training
Data was collected from 10 countries spanning regions that included: Azerbaijan, Belarus, Georgia, Moldova, Romania, India, South Africa, China, Zambia, USA. The total amount of 460 000 CXR Dicom images were used for training the model. The training data consisted of age groups >=15 years.
Reference Standard
The TB model is trained using X-Rays using microbiological reference in combination with expert human readers (radiologist). The ML normal / abnormal model model is referenced using expert human readers (radiologist) as ground truth.
Publications
Computer-aided detection of tuberculosis from chest radiographs in tuberculosis prevelance survery in South Africa: external validation and modelled impacts of commercially available artificial intelligence software (https://www.thelancet.com/journals/landig/article/PIIS2589-7500(24)00118-3/fulltext)
Prospective multi-site validation of AI to detect tuberculosis and Chest X-Ray abnormalities (https://ai.nejm.org/doi/full/10.1056/AIoa2400018)
This website works best with browsers other than Internet Explorer.