Non-profit AI Research · Toronto, Canada
Foundation AI for Medical Imaging
Strategic Partnership Proposal
M31.Bio × IBM Canada
Dr. Bo Wang
Director, M31.Bio
Professor, UHN / University of Toronto
Creator of scGPT — Nature Methods 2024
Non-profit AI for Health Open Science
Market Opportunity
Clinical AI Has a Deployment Problem
Scale of the opportunity
$20B+
Medical imaging AI market by 2030 — from ~$2.4B today
300M+
Medical scans performed annually across North America
<30%
of Canadian hospitals with AI-assisted imaging workflows
30%
Annual growth in imaging data — outpacing radiologist capacity
Why adoption is stalling
🧩
Point solutions, not platforms
Each vendor solves one problem — a hospital needs 20 models to cover full imaging workflows
🔗
No interoperability
Siloed models block hospital-wide deployment; data structures don't transfer across systems
☁️
No enterprise-grade cloud infrastructure
Canadian hospitals need sovereign, PHIPA-compliant AI deployment — today, that's missing
🇨🇦
Policy mandate without execution
Pan-Canadian Health Data Strategy calls for scalable AI — no platform has delivered at national scale
M31.Bio
Technology Platform
One Foundation. Every Modality.
Nature Comms 2024
MedSAM — Universal Medical Image Segmentation
1.5M+ image-mask pairs · 10 modalities · 30+ cancer types · 86 internal + 60 external validation tasks · 3,300+ citations · MICCAI 2024 Young Scientist Award
2024
MedSAM2 — 3D & Video Volumetric Segmentation
85%+ reduction in annotation time · validated across 5,000 CT scans · extends to video and volumetric imaging streams
ICLR 2025 Spotlight
MorphoDiff — Cell Morphology Prediction
Top 3.26% · generative AI for cellular phenotype under drug / genetic perturbations · bridges imaging and drug discovery
ICML 2025
MedRAX — Agentic X-Ray Reasoning
Full clinical reasoning from chest X-rays · 500K+ views on X · benchmark adopted by Grok 4 / xAI evaluation team
Performance vs. point solutions
97.3%
Average Dice score across all modalities
5.2×
Faster than manual segmentation
85%
Reduction in annotation time (MedSAM2 user study)
10+
Imaging modalities from one unified model
vs. existing approaches
Single-modality specialist 1 use case
Traditional CAD tools Rule-based, no learning
M31 Foundation Models All modalities · agentic
M31.Bio
Proof Points
Published. Awarded. Deployed.
Peer-reviewed publications
Nature Communications · 2024
Segment Anything in Medical Images (MedSAM)
3,300+ citations · Top 50 AI/ML papers of 2024 · MICCAI 2024 Young Scientist Award
Nature Methods · 2023
Towards foundation models of biological image segmentation
Establishes the foundational framework for biomedical vision models
Nature Methods · 2024
The Multi-modality Cell Segmentation Challenge
Community benchmark defining state-of-the-art across modalities
Lancet Digital Health · 2024
Pan-cancer Abdominal Organ Quantification: FLARE22 Challenge
Clinical translation of foundation models at national scale
ICLR 2025 Spotlight — top 3.26%
MorphoDiff: Cell morphology prediction under chemical & genetic perturbations
ICML 2025
MedRAX: Medical Reasoning Agent for Chest X-Ray
Benchmark adopted by Grok 4 / xAI X-ray evaluation team
Hospital & industry partners
🏥
Sunnybrook Hospital
Clinical validation — radiology & oncology imaging workflows
👶
The Hospital for Sick Children (SickKids)
Pediatric imaging AI · precision diagnostics
🇺🇸
UTHealth Houston
Pan-cancer segmentation · large-scale dataset collaboration
📱
PocketHealth
Patient-facing medical imaging platform integration
Research leadership
Dr. Bo Wang — Director, M31.Bio · scGPT (Nature Methods 2024)
Dr. Jun Ma — Lead researcher · first author, MedSAM · MedSAM2 · MedRAX · MICCAI 2024 Young Scientist Award
M31.Bio
Strategic Fit
Complementary Strengths. One Mission.

IBM's enterprise reach + M31's frontier models = Canada's AI health platform

IBM Canada
✓
watsonx.ai — enterprise AI model deployment, HIPAA / PHIPA ready
✓
IBM Cloud — sovereign Canadian cloud infrastructure for hospital data
✓
Watson Health legacy — clinical trust, hospital relationships, regulatory experience
✓
IBM Research Canada — proximity to Vector Institute & Mila; shared academic network
✓
Enterprise distribution — sales channels into health systems, government, insurance
✓
Enterprise AI agent strategy — MedRAX's agentic reasoning is a direct fit for IBM's 2025 roadmap
M31.Bio
✓
MedSAM / MedSAM2 — world's most cited universal medical image segmentation models
✓
MedRAX — agentic AI for clinical reasoning; Grok 4 adopted its benchmark
✓
Clinical validation — Sunnybrook, SickKids, UTHealth; real hospital deployments, not demos
✓
Non-profit structure — enables joint CIHR, NSERC, NRC IRAP grant applications
✓
Academic credibility — U of T, UHN, Nature / ICML / ICLR; builds government trust
✓
Needs: scalable cloud, PHIPA-compliant inference, enterprise distribution at national scale
M31.Bio
Partnership Opportunities
Four Paths to Partnership
01  ·  PRODUCT
MedSAM on watsonx.ai
Deploy M31's universal segmentation models as an enterprise API on watsonx.ai. IBM distribution channels + M31 frontier models — zero integration friction for hospital customers already on IBM infrastructure.
Enterprise APIwatsonx.ai
02  ·  INFRASTRUCTURE
Scalable Canadian Hospital Inference
Run M31 model inference pipelines on IBM Cloud — PHIPA-compliant, low-latency, scaled across Canadian hospital networks. IBM Cloud as the sovereign AI backbone for M31's clinical workflows.
IBM CloudPHIPA Compliant
03  ·  RESEARCH + GRANTS
Joint Grant Applications
Co-apply to CIHR AI in Health, NSERC Alliance, and NRC IRAP programs. M31's non-profit status + IBM's industrial partnership tier creates a strong government funding profile neither organization can match alone.
CIHRNSERCNRC IRAP
04  ·  DATA + STRATEGY
National Health Data Platform
Joint data partnerships with Canadian hospital networks. M31 models + IBM health data infrastructure + Pan-Canadian Health Data Strategy alignment — building toward a national-scale AI health platform.
Pan-CanadianData Strategy
M31.Bio
Next Steps
Three Asks. This Meeting.
1
Joint pilot — Canadian hospital
Identify one hospital from M31's existing network (Sunnybrook or SickKids) to run a joint MedSAM + watsonx.ai pilot. Define success metrics, 90-day timeline, and co-publication plan.
2
Technical integration call — watsonx + MedSAM API
Schedule a technical session: IBM's watsonx engineering team + M31 research team. Map the API integration path, data pipeline architecture, and compliance requirements (PHIPA, PIPEDA).
3
Memorandum of Intent / Letter of Intent
Formalize the partnership framework with an MOI or LOI — enables joint grant applications (CIHR / NSERC) and signals commitment to hospital partners and government stakeholders.
M31.Bio
Non-profit AI Research · Toronto, Ontario · general@m31.bio
✉️
🌐
𝕏
Dr. Bo Wang
Director, M31.Bio
Professor, UHN / University of Toronto
Creator of scGPT — Nature Methods 2024
Researcher, Xaira Therapeutics
Dr. Jun Ma
Lead Researcher, M31.Bio
First author — MedSAM · MedSAM2 · MedRAX
MICCAI 2024 Young Scientist Award
M31.Bio