The measurement-and-validation layer for AI medical imaging.
NorthLuma Metrology Pro is the hosted SaaS layer above the open-source mcx-msa library. Decompose your AI imaging device's output variance across subject, operator, device, illumination, and algorithm. Generate uncertainty budgets, fairness audits, and audit-trail-signed reports aligned to FDA 21 CFR 820.72(a), EU MDR Annex I §17.1, ISO 13485 §7.6, and the EU AI Act window.
The library ships chapter-by-chapter on PyPI between May 2026 and January 2027. The hosted Pro tier opens a private beta after the v1.0 release. Join the waitlist below.
GR&R MATLAB Toolkit v0.5 — ship today, while we build the hosted SaaS.
The underlying methodology toolkit that powers the forthcoming NorthLuma Metrology Pro SaaS, packaged as a standalone MATLAB toolkit. Eight Measurement-System Analysis methods auto-dispatched, audit-ready HTML reports, plus a 10-document Compliance Pack v1.0 (SRS, SAD, V&V, Test Report, SBOM, IEC 62304 Class A, ISO 14971 Risk Analysis, Release Notes, User Manual, Cybersecurity Statement). Buy it now and you get free upgrades through v1.0 (Jan 2027).
- 11 MATLAB functions (~7,840 LOC)
- 8 MSA methods auto-dispatched
- 5 worked examples
- 6 reference CSV datasets
- 10-doc Compliance Pack v1.0
- ~117 KB of user manuals
- AIAG MSA 4th Edition
- FDA 21 CFR 820.72(a)
- EU MDR Annex I §17.1
- ISO 13485 §7.6
- IEC 62304 Class A
- VDA 5 / ISO 14971
- Single-org commercial license
- Free upgrades through v1.0
- 30-day support window
- Private beta access to SaaS
- Roadmap influence on v0.6+
- Refund if blocking bug found
For comparison: Minitab MSA add-on is ~€1,800/year per seat. Q-DAS qs-STAT is ~€2,400/year. A single MSA-audit consulting engagement is €1,500–6,000.
Pricing is intentionally low for v0.5 early access. Price increases to €299 at v0.7 release (Aug 2026) and €499 at v1.0 (Jan 2027). Early buyers keep their €199 price for life.
The artifact every regulator wants is statistical, not documentational.
Today AI-medical-imaging device makers demonstrate measurement-system capability by hand: spreadsheets, consultants at $150–500 per hour per submission, and PDF templates assembled by ALM tools that were built for documentation, not statistics. We see three pains repeatedly:
- 1. The EU AI Act forcing function is real, the date is in motion.Article 15 requires declared accuracy as a labelling claim. The medical-device deadline window is in active flux as of May 2026 — Aug 2026 / Aug 2027 / Aug 2028 are all on the table depending on Council vs ratified text. Sophisticated readers should verify against the EU AI Office source before quoting any specific date. The forcing function does not move; the date does.
- 2. FDA 21 CFR 820.72(a) audits fail when uncertainty is hand-waved.The clause is short: measurement equipment shall be “capable of producing valid results.” It survived the 2026-02-02 QMSR transition verbatim. Demonstrating “capable” for an AI algorithm in 2026 means showing which fraction of variance comes from where — subject, operator, device, illumination, algorithm — and producing an uncertainty budget per AIAG MSA 4th Edition conventions. PDFs assembled from compliance templates do not show this. Statistics do.
- 3. ISO 13485 §7.6 calibration cycles are skipped on AI devices because nobody knows what to calibrate.A CT scanner has phantoms. An rPPG algorithm does not. The library and the hosted layer make AI-device calibration concrete — Fitzpatrick I-VI fairness audits, drift over time, robustness to illumination, ANOVA-decomposed components.
Three capabilities at v1.0
The hosted Pro tier ships once the open-source library reaches v1.0 (January 2027). Three capabilities anchor the offering:
Upload a validation study (parts × operators × replicates, long format). Get an AIAG-anchored verdict — capable (%GR&R < 10%), marginal (10–30%), or reject (>30%) — with ndc ≥ 5, Cgk ≥ 1.33, and a tabulated uncertainty budget. Two-way crossed ANOVA Method 1 and Average & Range Method 2 supported. Verified against three published worked examples to 3-decimal-place agreement.
One PDF closes five regulations and one industry convention: FDA 21 CFR 820.72(a), EU MDR Annex I §17.1, ISO 13485 §7.6, EU AI Act Article 15, IEC 62304 (Edition 2 expected late 2026), and AIAG MSA 4th Edition thresholds. The export is the artifact. The signature on the audit trail is cryptographic.
Pro hosts the same statistics core as the MIT-licensed mcx-msa library. Pro adds enterprise integration: multi-user collaboration, saved validation runs, regulator-ready report templates, audit-trail signing, white-glove onboarding, EU data residency. Algorithms remain open. Integrations and report templating are commercial. By design.
Built on our open-source library
The library ships in eight releases on PyPI, from May 2026 through January 2027, each building toward full AI-assisted Measurement-System Analysis for medical imaging. v1.0 lands in January 2027.
| Release | Module | Target |
|---|---|---|
| v0.1 | Skeleton + GR&R ANOVA core (Method 1) + Average & Range (Method 2) | May 2026 |
| v0.2 | Monte-Carlo phantom simulation | May 2026 |
| v0.3 | SFDI optical-property estimation | Jun 2026 |
| v0.4 | PSD / Welch / Bartlett spectral analysis | Aug 2026 |
| v0.5 | Wavelet-based denoising | Aug 2026 |
| v0.6 | Full GR&R / ANOVA + uncertainty budget core | Oct 2026 |
| v0.7 | rPPG validation suite + Fitzpatrick I-VI fairness | Nov 2026 |
| v0.8 | Audit-protocol export (ISO 13485 / EU AI Act / FDA SaMD) | Dec 2026 |
| v1.0 | v1.0 release — full pipeline on three modalities | Jan 2027 |
Repository: github.com/northluma/mcx-msa. License: MIT. Documentation: northluma.com/mcx-msa. The library is the credibility seed; the hosted Pro tier is the regulated-production layer above it.
Pricing — published intent, not yet billed
Three tiers at v1.0. Today the page is a waitlist; pricing below is published intent, not yet billed. Free during the open-source v1.0 release window (January 2027) — all Pro features open in exchange for case-study and regulatory-validation feedback.
| Tier | Price | Audience & capability |
|---|---|---|
| Free / open source | $0 | Researchers, students, individual engineersFull mcx-msa library on PyPI; community support |
| Pro | $99 / seat / month | Single quality engineer at a regulated medical-device makerLibrary + saved validation runs + basic ISO 13485 reports |
| Team | $499 / team / month | Regulated team of 3–10 engineersMulti-user collaboration + full template suite + unlimited runs |
| Enterprise | Contact for pricing | Medical-device manufacturer (50+ engineers)SLA, white-glove onboarding, regulator-ready report templates, cryptographically-signed audit trail. EU data residency configurable. |
Hosted on EU infrastructure (GDPR Article 9 special-category aware). Data residency configurable per customer.
Be on the design-partner shortlist
Drop your work email. We do not spam. We send one update per mcx-msa release (eight emails through January 2027) and one when the Pro tier opens a private beta. Counterparties under NDA receive a separate channel.
No spam. Unsubscribe anytime. EU data residency.
Who is building this
NorthLuma Metrology is led by a single founder with a Magna Electronics measurement-system-analysis background and a five-to-seven-year commitment to building the category. The founder identity is disclosed under NDA at design-partner stage.
NorthLuma AB is a Swedish aktiebolag headquartered in Linköping — Sweden's medtech corridor, walking distance from Sectra HQ and Region Östergötland. The company runs lean by design.
Frequently asked
Is NorthLuma Metrology FDA-cleared, ISO-certified, or EU-AI-Act-compliant?+
No. NorthLuma Metrology is a methodology library and a forthcoming hosted SaaS layer. Neither is a medical device. Customers use the artifacts the platform produces (variance decomposition, uncertainty budgets, audit-protocol PDFs) inside their own quality-management system to support their own regulatory submissions. We hold the methodology to AIAG MSA 4th Edition conventions and align reports to the regulator clauses listed below; we do not represent that the platform itself is “cleared” or “certified”. ISO 13485 certification of NorthLuma AB as a quality-management system begins post-revenue.
Which regulatory clauses does the audit-protocol export close?+
Five regulations + one industry convention + GUM appendix: FDA 21 CFR 820.72(a) — measurement equipment “capable of producing valid results”; EU MDR Annex I §17.1 — repeatability, reliability, performance; ISO 13485 §7.6 — control of monitoring and measuring equipment; EU AI Act Article 15 — declared accuracy as a labelling claim (date contested as of May 2026; verify against the EU AI Office source); IEC 62304 — software lifecycle (Edition 2 publication expected late 2026); AIAG MSA 4th Edition — industry-convention thresholds (%GR&R < 10% / 10–30% / > 30%; ndc ≥ 5; Cgk ≥ 1.33).
How does this differ from Ketryx, Veeva Vault Quality, MasterControl, or Greenlight Guru?+
In our research those vendors own horizontal-process compliance — quality management, design history files, document control, audit trails. None ship a per-algorithm metrology layer for AI medical imaging today. NorthLuma Metrology integrates underneath a horizontal eQMS (the audit-protocol PDF can attach to a Ketryx or Vault record); it does not replace the eQMS. Vertical statistical-rigor layer + horizontal process eQMS is the deployment pattern we believe most medical-device makers will land on.
How does this differ from Minitab, JMP, or Q-DAS?+
Those tools own industry-agnostic MSA / DOE / SPC. They are excellent at the statistics. They are not regulator-fluent: no out-of-the-box FDA-21-CFR-820.72 / ISO 13485 §7.6 / EU AI Act Article 15 audit-export templates, no medical-device-specific fairness modules (Fitzpatrick I-VI demographic stratification), no AI-uncertainty-quantification primitives (calibration drift, robustness to illumination). NorthLuma Metrology specialises the same statistical core for AI medical imaging.
What ships when?+
The open-source mcx-msa library ships chapter-by-chapter from May 2026 to January 2027 on PyPI. The hosted Pro tier opens a private beta after the January 2027 v1.0 release. Public general availability is planned for Q3 2027. The waitlist on this page is the access channel for the private beta. Re-verify against the release roadmap above; we keep it current.
This page is reviewed quarterly. Last reviewed: 2026-05-09. Next review: 2026-08-09.
Specific market-size figures cited on the public NorthLuma site are drawn from publicly summarized industry-analyst projections; sophisticated readers should run independent diligence. Specific regulatory-deadline dates (EU AI Act medical-device window) are in active flux as of May 2026; confirm against the EU AI Office source before quoting.