EndoAngel AI Platform

Technology

State-of-the-art deep learning models optimized for real-time medical image analysis — designed to integrate seamlessly into existing clinical workflows.

100K+

Clinical
Cases

800K+

Annotated
Images

<40ms

Inference
Latency

Full GI

Detection & QC
Gastroscopy · Colonoscopy · EUS

Scope
Compatible

Major brands
supported

Core Technology

AI Architecture

EndoAngel uses state-of-the-art deep learning models optimized for real-time medical image analysis, trained on large-scale multi-center endoscopy datasets with expert-validated ground truth labels.

🧠

Deep Learning Framework

  • Convolutional Neural Networks (CNN)
  • Multi-task: detection, segmentation & classification
  • Federated learning for continuous improvement
  • Trained on 100,000+ annotated images

Real-time Processing

  • Inference latency <40ms per frame
  • Hardware-accelerated GPU processing
  • Zero impact on endoscopy workflow
  • Supports 1080p input resolution

📈

Training Data Quality

  • Multi-center data from leading hospitals
  • Expert-validated ground truth labels
  • Diverse patient populations
  • Multiple equipment brands and settings

Upper GI AI — Gastroscopy Interface

Colonoscopy AI — Detection Interface

Colonoscopy AI — Detection Interface
1

Procedure Time Recording: Records the inspection time and the esophageal review time during the procedure.

2

Real-Time Operation Scoring: Monitors blind spots and automatically scores the physician’s operation according to standardization guidelines.

3

Blind Spot Monitoring: Indicates the number of anatomical sites already covered and those yet to be examined.

4

Early Cancer Recognition: Intelligently identifies high-risk lesions, highlighted with a red bounding box and voice reminder.

5

Model Illumination: Real-time reminder of all 28 gastric anatomical sites with illumination of the corresponding model area.

6

Automatic Image Collection: Representative images of sites and lesions are automatically captured and saved during the procedure.

EndoAngel Lower GI Quality Control

Colonoscopy AI — Withdrawal Interface

Colonoscopy AI — Withdrawal Interface
1

Procedure Time Recording: Records the intubation time and withdrawal time during the endoscopic procedure.

2

Real-Time Bowel Preparation Scoring: Intestinal preparation scores (eBBPS) are calculated and displayed in real time.

3

Withdrawal Speed Monitoring: Monitors withdrawal speed in real time and displays it on a panel to alert the endoscopist.

4

Adenoma / Polyp Recognition: Intelligently identifies high-risk lesions in real time, highlighted with a red bounding box.

5

Ileocecal Valve Reminder: Automatically identifies arrival at the ileocecal valve and illuminates the intestinal model accordingly.

6

Automatic Image Collection: Representative images of sites and lesions are automatically captured and saved during the procedure.

EndoAngel Upper GI Quality Control Indicators

Quality Control IndexCorresponding EndoAngel Function
Sufficient esophageal endoscopy time (>2 min)Automatic esophageal timing
Adequate inspection time for stomach and duodenum (>5 min)Automatic gastric cavity timing
Recording and feedback of inspection time (>90%)Automatic time recording
Systematic observation of entire esophagus and stomach (intragastric views >22 images)Blind spot monitoring (26 zones) and standardization of endoscopist operations
Comprehensive and accurate acquisition of endoscopic imagesAutomatic image capture
Training endoscopists to identify suspicious lesions under white lightReal-time assistance under white light to indicate suspicious lesions
White-light endoscopy combined with image-enhanced endoscopy (e.g., NBI) for suspicious lesionsReal-time assistance under NBI to indicate suspicious lesions

EndoAngel Lower GI Quality Control Indicators

Quality Control IndexCorresponding EndoAngel Function
Cecal intubation rate: screened patients ≥95% / all patients ≥90%Real-time withdrawal speed monitoring
Withdrawal process must be slow; withdrawal time in negative colonoscopy (≥6 min)Real-time withdrawal speed monitoring
Recording of withdrawal time (>90%)Automatic recording and timing
Polyp detection rate (≥40%)Real-time assistance for polyp recognition
Adenoma detection rate (>25%)Real-time assistance for polyp type identification and adenoma diagnosis
Pedunculated polyps and serrated polyps <2 cm not suitable for surgical resection should attempt endoscopic resection firstReal-time identification of polyp types
Bowel preparation assessment (Boston standard)Automatic bowel preparation scoring

Blind Spot Monitoring During Upper GI Endoscopy (26 Zones)

Blind spot monitoring upper GI endoscopy 26 zones

Withdrawal Speed Monitoring

Withdrawal speed monitoring colonoscopy

Integration Design

Integration Design

EndoAngel connects via HDMI pass-through — no modification to existing equipment, no IT integration required. Works with Olympus, Fujifilm and Pentax systems.

Seamless Workflow

  • HDMI pass-through: plug and play installation
  • No modification to existing endoscopy equipment
  • Compatible with Olympus, Fujifilm, Pentax
  • Works with existing monitor or secondary screen

Deployment Options

  • Standalone bedside unit
  • Department-level server (multi-room support)
  • Cloud-connected for remote updates and analytics
EndoAngel AI instrument detection and measurement during endoscopic procedure

Technical Specifications

Parameter Value
Inference latency<40ms per frame
Input resolution1080p supported
Compatible scopesOlympus, Fujifilm, Pentax
ConnectionHDMI pass-through
AI model updatesOTA via secure connection
CertificationsNMPA (China) · CE Mark (EU)
Data processing100% local — no external transmission
ComplianceGDPR · local data protection regulations

Data Security

All image processing is performed locally at the hospital. No patient data is transmitted externally at any time.

  • All processing performed locally at the hospital
  • No patient data transmitted externally
  • DICOM-compatible data handling
  • GDPR and local data protection compliant

Continuous Improvement

EndoAngel is continuously improving through secure OTA updates and ongoing clinical validation studies with partner hospitals.

  • Regular AI model updates via secure OTA
  • Performance analytics dashboard
  • Ongoing clinical validation studies
  • Feedback loop with clinical partners

See the Technology in Action

Schedule a live demonstration at your facility — no commitment required.