Skip to content
한국어 English 日本語 中文
Request a Demo

AI analysis software
for chest X-ray reading

Classifies 14 abnormal findings, localizes lesions and visualizes AI attention — all in one report.

01Overview

AI that supports the reader

TORAH Vision AI uses a ResNet50-based deep learning model to analyze high-resolution chest X-rays, presenting clinicians with the likelihood and location of abnormal findings. The clinician makes the final decision; the AI supports the process.

01

14 abnormal findings

From aortic enlargement to pneumothorax and pulmonary fibrosis — 14 findings and No Finding in a single analysis.

02

Grad-CAM visualization

Heatmaps overlaid on the original image show the regions the AI relied on.

03

Lesion localization

Per-finding bounding boxes indicate the location of suspected abnormalities.

04

High-resolution image processing

Uses the full grayscale depth of high-resolution medical images.

02Analysis Findings

14 abnormal findings + No Finding

The analysis classes defined in the product specification, grouped by anatomical region.

No. Finding Region Description
01 Aortic Enlargement Heart · great vessels Widening of the aortic shadow beyond the normal range.
02 Atelectasis Lung parenchyma Incomplete expansion of part of the lung, resulting in volume loss.
03 Calcification Lung parenchyma Calcium deposits in the lung parenchyma, lymph nodes or other structures, appearing as high-density foci.
04 Cardiomegaly Heart · great vessels An enlarged cardiac silhouette relative to the thoracic dimensions.
05 Consolidation Lung parenchyma Alveolar air replaced by fluid or cellular material, producing homogeneous opacity.
06 ILDInterstitial Lung Disease Lung parenchyma Changes in the pulmonary interstitium presenting as reticular or linear opacities.
07 Infiltration Lung parenchyma Ill-defined opacity spreading through the lung parenchyma.
08 Lung Opacity Lung parenchyma An area of increased attenuation, appearing whiter than normal lung.
09 Nodule/MassNodule / Mass Lung parenchyma A round or oval opacity within the lung, classified as a nodule or mass by size.
10 Other Lesion Other Abnormal findings not covered by the other categories.
11 Pleural Effusion Pleura Fluid in the pleural space, typically blunting the costophrenic angle.
12 Pleural Thickening Pleura Thickened pleura, seen as opacity along the outer lung margin.
13 Pneumothorax Pleura Air in the pleural space causing partial or complete lung collapse.
14 Pulmonary Fibrosis Lung parenchyma Fibrotic change of lung tissue with reticular opacities and volume loss.
— No Finding Normal class Studies in which none of the 14 abnormal findings is detected are classified separately as normal.

Descriptions are general radiological definitions provided for reference and do not represent the diagnostic performance of the product.

03Analysis Output

Actual analysis output

Review the original image, the Grad-CAM attention heatmap and the combined lesion-and-heatmap view side by side.

Original image
01
Original imageThe chest X-ray as uploaded
Grad-CAM heatmap
02
Grad-CAM heatmapAI attention regions shown in color
BBox · Grad-CAM combined
03
BBox · Grad-CAM combinedLesion locations with attention regions

Excerpted from a TORAH Vision AI analysis report screen.

04Workflow

From image input to results

A five-step analysis process delivers results ready for clinical review.

  1. STEP 01

    Image input

    Upload a chest X-ray in DICOM or another supported format.

  2. STEP 02

    Preprocessing

    The high-resolution image is normalized and prepared for analysis.

  3. STEP 03

    AI analysis

    The deep learning model analyzes the image and classifies findings.

  4. STEP 04

    Grad-CAM visualization

    Regions the AI attended to are rendered as a heatmap.

  5. STEP 05

    Review

    Review the report with per-finding scores and visualizations.

05Specifications

Product specifications

Product
TORAH Vision AI
Type
Medical AI software supporting chest X-ray reading
Input
Chest X-ray (DICOM, JPEG, PNG)
Image processing
High-resolution medical image processing
Model
ResNet50-based deep learning
Analysis classes
14 abnormal findings + No Finding
Output
Per-finding scores, bounding boxes, Grad-CAM heatmap, integrated report
Quality system
ISO 13485, Digital Medical Device GMP

This product is software that supports clinicians in reading chest X-rays. The final clinical decision must always be made by a qualified clinician.

Contact

Let's talk about
medical AI for your institution

Product adoption, demo requests, clinical collaboration or international partnerships — our team responds within one business day.

Get in Touch View Certifications