High-resolution image processing, deep learning, explainable visualization and multi-class classification — the foundation of TORAH Vision AI.
01Core Technology
Each technology is designed so clinicians can understand and verify what the AI reports.
High-Resolution Imaging
Uses the wide grayscale range of medical images so that subtle differences in density inform the analysis.
Deep Learning
A ResNet50-based convolutional neural network extracts image features from chest X-rays and classifies findings.
Explainable AI
Heatmaps highlight the regions the AI relied on, so clinicians can verify results directly on the image.
Multi-class AI
A single analysis classifies 14 abnormal findings together with No Finding for a comprehensive assessment of each study.
02Development Pipeline
Every stage, from data management to release, is controlled under our medical device quality management system.
Medical images are acquired with personal information removed.
Specialists label findings to build the training dataset.
Deep learning models are trained and optimized.
Model performance is verified on separate data.
Products are released and maintained under QMS procedures.
03Research
Mortality-risk prediction combining chest X-ray and clinical data — Excellence Award for presentation
Presented AI-based prognosis prediction for bacteremia patients with lung disease
Exhibition and demonstration at Korea ECMO Symposium & The 8th BEST 2025
Contact
Product adoption, demo requests, clinical collaboration or international partnerships — our team responds within one business day.