A Product of Clearflow Systems

Federated medical imaging at the global scale

FedScan connects hospitals, community health centres, and rural clinics through a privacy-preserving federated learning network — enabling AI-assisted diagnosis without centralising patient data.

4,817
DICOM studies
10
Federation sites
1.000
AUC-ROC (round 20)
5
Expert tiers
Image delivery pipeline
How a DICOM scan travels from scanner to expert browser — click any node to learn more
🏥
Health centre
DICOM acquisition
→
🔒
Anonymise
Strip PHI headers
→
☁️
STOW-RS
Upload to GCP
→
🗄️
Cloud archive
GCP Healthcare API
→
📡
WADO-RS
Pixel fetch
→
🖥️
Browser viewer
W/L + AI overlay
→
👨‍⚕️
Expert review
Verdict + escalate
GCP DICOM Store — Live | Study: FS-2026-04817 | WADO-RS: us-central1 | AI: Pneumonia 91.3%
W:1500 / L:−600 · Lung window
AXIAL · Slice 28/48
Consolidation 91%
CORONAL · Slice 18/32
GGO 73%
SAGITTAL · Slice 14/28
MPR Reference
Slice: 28 W/L: 1500/−600
Patient & study
Patient ID
PT-04817
Age / Sex
47 y / M
Site
Riverside CHC
Scanner
AnyScan S
Noise σ
90 HU
GCP Store
us-central1
FL model result
Pneumonia detected
RLL consolidation + GGO
91.3%
Consolidation frac.
0.78
GGO fraction
0.63
Mean lung HU
−412
Air fraction
0.32
Expert review
Verdict
✓
Confirm
AI correct
✎
Modify
Adjust finding
✕
Reject
AI incorrect
↑
Escalate
Second opinion
Severity
Mild
Moderate
Severe
Critical
Clinical notes
↑ Escalate to — select expert
RP
Dr. R. Patel
Consultant Radiologist · T3
Online
KN
Dr. K. Nakamura
Sr. Radiologist · T4
In review
SO
Prof. S. Okonkwo
Chief of Radiology · T5
Online
Review hierarchy — current position
T1
Dr. Lopes
T2
Dr. Mensah
T3
Dr. Patel
T4
Dr. Nakamura
T5
Prof. Okonkwo

Submit medical images

Upload DICOM studies to the FedScan GCP Healthcare API store for AI classification and expert review

📁
Click to select DICOM files or drag and drop
.dcm files · DICOM Part 10 · Any modality
GCP STOW-RS endpoint
POST https://healthcare.googleapis.com/v1/projects/fedscan-prod-001/locations/{region}/datasets/fedscan-dataset/dicomStores/fedscan-dicom/dicomWeb/studies
TRANSFER PROTOCOL
DICOMweb STOW-RS
HTTPS · TLS 1.3 · 256-bit AES
ANONYMISATION
Auto de-identification
DICOM PS3.15 Profile E
AI INFERENCE
Automatic on ingest
FedAvg round 20 model

Expert registration

Join the FedScan network of radiologists and provide AI-assisted second opinions to health centres worldwide

Select your expert tier
T1
Resident / Junior
T2
Specialist Radiologist
T3
Consultant Radiologist
T4
Senior Radiologist
T5
Chief of Radiology
Certifications held
Privacy & data governance
Your account will be verified within 48 hours. You will receive login credentials and onboarding instructions at your registered email.

Hospital & facility portal

Register your facility as a FedScan client to access AI-assisted image classification and expert review services

Federated learning participation
Community
Free
Up to 20 studies/month
AI inference included
T1–T2 expert review
48h response time
Standard
$49/mo
Up to 100 studies/month
Priority AI inference
T1–T3 expert review
8h response time
Advanced
$149/mo
Unlimited studies
Instant AI inference
T1–T5 expert review
2h response time
Data governance agreement
Your facility ID and GCP credentials will be emailed within 24 hours. Technical onboarding support is included with all plans.

System flowcharts

Click any flowchart to view in full detail

📡
Flowchart 1
Image delivery pipeline — from scanner to browser via GCP Healthcare API and DICOMweb
👨‍⚕️
Flowchart 2
Expert review decisions — confirm, modify, reject, or escalate with retraining label generation
↑
Flowchart 3
Escalation hierarchy — five-tier expert chain from resident to chief of radiology
Flowchart 1 — Image delivery pipeline
DICOM file journey from rural health centre to expert browser via Google Cloud Healthcare API
Rural / community health centre
CT / MRI scan acquired
↓
Anonymise DICOM
Strip PHI · DICOM PS3.15 Profile E
↓
STOW-RS upload
HTTPS POST → GCP Healthcare API
↓
Google Cloud DICOM archive
fedscan-dicom · us-central1 · DICOMweb
↓
FL model inference
FedAvg round 20 · auto on ingest
↓
Expert notification
Alert queued to next available expert
↓
WADO-RS pixel fetch
Browser requests slice data via HTTPS GET
↓
Browser canvas viewer
W/L 1500/−600 · AI overlay · multi-planar
↓
Expert reviews in browser
No plugin · no download · any device
Flowchart 2 — Expert review decisions
Four verdict paths and how each generates a retraining label for the FL model
Expert opens study
Views DICOM · FL result · radiomic features
↓
Review scan + AI result
Adjust W/L · scroll slices · compare
↓
Verdict
type?
↓
Confirm
Soft label
Weight 1.0×
Modify
Hard override
Weight 1.5×
Reject
Error label
Weight 2.0×
Escalate
Consensus
Weight 2.5×
↓
Retraining buffer
(features, label, weight) tuples accumulate
↓
FL coordinator triggers new round
Distributes buffer · FedAvg aggregation
↓
Updated global model deployed
Cloud inference server refreshed
Flowchart 3 — Expert escalation hierarchy
Five-tier escalation chain — each tier can resolve or pass upward with full annotation trail
T1
Tier 1 — Resident / Junior Doctor
First AI-assisted review · generates initial verdict
Entry point
↓ escalate if uncertain
T2
Tier 2 — Specialist Radiologist
Reviews T1 notes + original scan · full AI report
Can resolve
↓ escalate if complex
T3
Tier 3 — Consultant Radiologist
Full annotation trail · all prior expert notes
Can resolve
↓ escalate if rare / unusual
T4
Tier 4 — Senior Radiologist
Complex, rare, and high-stakes cases
Can resolve
↓ escalate only for exceptional cases
T5
Tier 5 — Chief of Radiology
Terminal authority · multi-disciplinary if needed
Final authority
↓ case closed · EHR updated · label added to FL buffer
Diagnosis confirmed · records updated
Expert verdict → retraining label → FL improvement cycle