Turning Real-World Radiology Into TAO Demand
RadTensor is a vendor-neutral radiology subnet on Bittensor that routes de-identified imaging exams from hospital PACS to competing AI miners -- the first enterprise-scale fiat subnet.
Read the Full 205-Section Technical ProposalRadiology AI is broken by centralization
Billions of imaging exams per year. Billions in AI spend. Yet healthcare providers are locked into expensive, inflexible systems that block innovation and limit access.
Vendor Lock-In
Hospitals are trapped in rigid, single-vendor AI contracts with proprietary PACS integrations. Switching costs are enormous, and innovation is stifled.
Prohibitive Costs
Implementation costs run $100K-$500K+ per site. Per-study AI fees of $3-8 add up to six-figure annual costs even at moderate volumes.
Legacy Infrastructure
PACS and VNA systems were never designed to connect to modern AI networks. Every AI vendor requires a custom, expensive point-to-point integration.
Compliance Gaps
Sending imaging data to external AI services without robust de-identification, encryption, and audit trails creates unacceptable regulatory risk.
RadTensor inverts this model. Instead of centralized SaaS, it exposes radiology AI as a competitive, decentralized marketplace on Bittensor -- vendor-neutral, usage-based, and compliant by design.
Patent-Pending Gateway Architecture
Based on US Patent Application No. 63/720,155 (QRad), RadTensor bridges PACS/VNA systems to a decentralized AI marketplace on Bittensor.
RadTensor Gateway
Node 003Central hub orchestrating imaging data flow between radiology systems and the Bittensor subnet. Vendor-neutral, modular, API-driven.
End-to-End Data Flow
Purpose-built for the Bittensor network
RadTensor maps directly onto Bittensor's miner/validator architecture with a radiology-specific proof-of-effort mechanism designed for clinical accuracy.
AI Miners: Compete on Clinical Accuracy
Miners on the RadTensor subnet execute specialized AI models on de-identified radiology tasks. Each miner can specialize in particular modalities, body regions, or task types.
Task Specialization
- Chest X-ray triage
- CT brain hemorrhage detection
- Mammography screening
- QA on image orientation & labeling
Standardized I/O
- De-identified image data or derived feature tensors
- Modality & body region metadata
- Per-label pathology probabilities
- Anomaly flags & heatmap references
Performance Metrics
- Diagnostic accuracy (AUROC)
- Latency & throughput
- Calibration quality
- Compliance with PHI constraints
Task Assignment & Reward Loop
This loop runs continuously, ensuring RadTensor's incentive structure remains closely aligned with clinical value.
The fiat-to-TAO revenue engine
Radiology imaging is mandatory, central to healthcare, and funded by national health systems and insurance. This is not speculative traffic -- it is embedded, reimbursable clinical demand.
3.6B+
Annual Imaging Exams
Global diagnostic imaging volume across X-ray, CT, MRI, ultrasound, and PET modalities.
$45B+
AI Imaging Market by 2030
Projected AI in medical imaging market with high double-digit CAGR from the mid-2020s.
$100K-500K
Per-Site AI Cost Today
Current implementation costs including PACS integration, licenses, and support contracts.
$3-8
Per-Study AI Fee
Current per-study pricing from centralized vendors, creating six-figure annual bills at moderate volumes.
RadTensor Revenue Scenarios
Conservative estimates at $3-8 net protocol revenue per exam
| Phase | Global Coverage | Annual Exams | Est. Revenue |
|---|---|---|---|
| Proof of Value | 0.1% | 3.6M | $10-30M |
| Regional Penetration | 1% | 36M | $100-300M |
| Mature Phase | 3% | 108M | $300M-900M |
Why RadTensor is Different
Unlike speculative or crypto-native subnets, RadTensor is backed by mandatory clinical workflows. Imaging volumes are not discretionary -- they are embedded in healthcare delivery. This means continuous, predictable, high-value external revenue flowing into the Bittensor network.
The Fiat-to-TAO Bridge
Institutions pay per-exam in fiat. RadTensor converts this into TAO-denominated rewards for miners and validators. Even 1% of global imaging volume creates hundreds of millions in annual protocol revenue -- a direct, sustainable bridge between healthcare spending and the Bittensor economy.
Patent-pending gateway technology
RadTensor's gateway architecture is based on US Patent Application No. 63/720,155 (QRad), covering the vendor-neutral radiology integration system that bridges PACS to decentralized AI.
US Patent Application No. 63/720,155
QRad Vendor-Neutral Radiology Integration System
Controlled by Project Lead -- Michael John Parker
System Workflow Overview
End-to-end workflow showing imaging data flow from PACS through the RadTensor gateway, security modules, AI/quantum processing, validation, and reintegration into clinical systems.
High-Level Functional Architecture
Simplified overview: imaging data from modalities to PACS, through the RadTensor gateway to AI/quantum processing, returning validated results to radiologists.
Marketplace & Provider Selection
Healthcare providers configure workloads via the marketplace interface. Miners and external services compete as computational providers in a vendor-neutral ecosystem.
Secure Data Flow Framework
Integration between PACS, marketplace, cloud connectivity, encryption, compliance logging, and dynamic routing to AI/quantum processors.
Processing & Validation Pipeline
Data processing workflow from the core node through cloud connectivity, routing, security modules, validation checkpoints, to AI and quantum processors.
Reintegration & End-User Delivery
Final stage: validated outputs flow from RadTensor back to clinical data stores and radiologists, with optional human-in-the-loop validation.
Why the Patent Matters for Bittensor
Defensible Moat
Patent-pending architecture gives RadTensor a defensible position in the radiology AI integration space, protecting the subnet's value proposition.
Open Subnet, Protected Gateway
The subnet itself is open on Bittensor, but the gateway IP ensures commercial deployments benefit the RadTensor ecosystem specifically.
Enterprise Credibility
Healthcare institutions require IP protection and legal clarity. The patent signals institutional-grade commitment to potential adopters.
From testnet to global radiology standard
RadTensor targets high-need, high-volume radiology segments where AI augmentation delivers clear clinical value and institutions are already exploring AI tools.
Proof of Value
Targeted Early Adopters
- Deploy gateway at 3-5 high-volume teleradiology groups
- Focus on chest X-ray triage and CT brain hemorrhage
- Demonstrate technical, regulatory, and financial viability
- Process millions of exams/year through testnet then mainnet
Regional Penetration
Integration Partnerships
- Partner with PACS vendors and enterprise imaging providers
- Expand to outpatient imaging centers and regional networks
- Standardize QA and triage protocols across institutions
- Handle tens of millions of exams per year
Mature Scale
De Facto Standard
- Function as standard routing layer for radiology AI
- Multi-vendor, multi-region coverage
- Deeper integration with reporting and reimbursement
- Nontrivial percentage of global imaging volume
Target Early Adopters
Teleradiology Groups
Large overnight and overflow reading operations with high volume and clear AI-augmentation ROI.
Outpatient Imaging Centers
High-throughput CT and MRI facilities seeking cost-effective AI triage and QA.
Regional Hospital Networks
Networks looking to standardize AI-assisted triage and quality protocols across sites.
Early Participation Incentives
Preferential Pricing
Below-market per-exam rates for early adopter institutions that deploy the RadTensor gateway.
Increased Reward Shares
Higher TAO allocation for early data contributors providing labeled cases for gold-standard evaluation sets.
Clinical Advisory Council
Early adopters join a council that shapes task definitions, evaluation criteria, and the RadTensor roadmap.
Explore the codebase
The subnet protocol is fully open. The patent-pending gateway creates a defensible commercial layer while keeping miner/validator participation decentralized.
protocol.py
radtensor/protocol.py
Synapse definitions -- RadTensorTask, HealthCheck, ComplianceAudit
reward.py
radtensor/reward.py
Composite scoring function: accuracy, latency, calibration, format
miner.py
radtensor/neurons/miner.py
Reference miner with model registry and task-specific inference handlers
validator.py
radtensor/neurons/validator.py
Reference validator with gold-standard scoring, canary jobs, weight updates
mechanism-design.md
radtensor/docs/mechanism-design.md
Full incentive and mechanism design specification
gateway-architecture.md
radtensor/docs/gateway-architecture.md
Patent-pending gateway architecture: modules, data flow, security model
6
Core Files
3
Synapse Types
4
Scoring Dimensions
The most direct path from fiat to TAO
RadTensor converts billions in real-world radiology spend into Bittensor network value. Join us in building the first enterprise-scale medical imaging subnet.
$10K
Hackathon Prize
1000 TAO
~$260K Discretionary Fund
$5K
Compute Credits
Direct
Bitstarter Accelerator Entry