Notes on bringing medical technology into everyday clinical use: regulation, adoption, imaging operations and AI governance.
Governing the Algorithm · Article 11 of 12
The EU AI Act and the FDA reached similar governance expectations independently. What their convergence means for SaMD companies that sell on both sides of the Atlantic.
Governing the Algorithm · Article 10 of 12
ISO 42001 does for AI governance what ISO 13485 did for quality: gives buyers an auditable way to verify it. Why healthcare AI vendors will increasingly need it.
Governing the Algorithm · Article 9 of 12
A validation study is a snapshot. Hospitals increasingly expect ongoing performance reporting after go-live, and vendors need a post-market surveillance plan that delivers it.
Governing the Algorithm · Article 8 of 12
A PCCP sets out in advance how a model may change and how each change is validated. For AI that keeps learning, that plan can matter more than the original clearance.
Announcement
We have signed a Memorandum of Understanding with NeoPathology Corp. to explore U.S. integration of an FDA-cleared lung imaging AI application, pairing their clinical network with our integration and operations work.
Governing the Algorithm · Article 7 of 12
"Human-in-the-loop" only protects patients if the clinician can actually disagree. How interface defaults turn claimed oversight into automation bias, and what exercisable oversight requires.
Breast Imaging
Dr. Kimberly Beavers, a fellowship-trained, board-certified breast imaging radiologist at NYU Langone Health and an active voice in American radiology, joins us as Founding Clinical Advisor.
Governing the Algorithm · Article 6 of 12
Every deployed model drifts as equipment, populations and protocols change. Drift is expected; failing to detect it is the governance failure.
Governing the Algorithm · Article 5 of 12
A strong overall accuracy figure can hide worse performance for specific patient groups. If subgroup performance was never tested, nobody knows the gap is not there.
Governing the Algorithm · Article 4 of 12
In healthcare AI, bad input data produces bad clinical output. Why data governance belongs with clinical safety rather than in an engineering back office.
Governing the Algorithm · Article 3 of 12
When an AI-assisted decision goes wrong, accountability runs through vendor, hospital and clinician. Why that chain has to be mapped before an incident, not after one.
Governing the Algorithm · Article 2 of 12
Most AI vendors describe their model when asked about governance. A model is an asset; governance is the management system that decides how it learns, changes and is monitored.
Governing the Algorithm · Article 1 of 12
A 510(k) assesses one version of a product, on one dataset, at one moment. Why being cleared says little about how an AI is governed after that moment.
Governing the Algorithm · Introduction
Launch compliance is giving way to operational compliance. Why AI, which can drift and be retrained after clearance, needs governance in a way drugs and fixed devices never did.
Beyond FDA · Part 12 of 12
The series in one frame: stalled pilots and products that never scaled trace back to four pillars. FDA clearance, clinical validation, interoperability and reimbursement.
Beyond FDA · Part 11 of 12
The Predetermined Change Control Plan lets AI models change within a pre-agreed plan instead of a new submission each time. Why it may be the biggest shift since the 510(k).
Beyond FDA · Part 10 of 12
ACR alignment is about how radiology practices, not only detection rates. Reporting structure, follow-up and workflow expectations that AI vendors tend to overlook.
Beyond FDA · Part 9 of 12
"Is this aligned with ACR guidance?" is now a buying-committee question in radiology. Why ACR alignment carries weight even though it is not a legal requirement.
Beyond FDA · Part 8 of 12
By deployment, the model is the easy part. PACS approval, EHR data governance and infrastructure decide the go-live date, and few teams budget time for them.
Beyond FDA · Part 7 of 12
Deals often die in IT security review, not in clinical evaluation. CIOs buy risk reduction, and products need to be built for that question from the start.
Announcement
Elliot Silverman, a radiology director and healthcare consultant with 50 years in American medical imaging, joins our Advisory Board to guide clinical evaluation, KOL engagement and teleradiology work.
Beyond FDA · Part 6 of 12
A slightly less accurate AI that fits PACS, RIS and EHR will often beat a better one that needs workarounds. What interoperability actually requires, item by item.
Beyond FDA · Part 5 of 12
FDA clearance, clinical validation and clinical adoption answer three different questions for three different audiences. Why conflating them stalls healthcare AI companies.
Beyond FDA · Part 4 of 12
Marketing rarely sells healthcare AI; clinicians vouching to other clinicians does. What key opinion leaders actually provide, and why investors watch it too.
Beyond FDA · Part 3 of 12
Hospitals now ask whether they can trust a model in their own environment. Why clinical validation is overtaking benchmark accuracy as the real differentiator.
Beyond FDA · Part 2 of 12
A great product is not a business until someone can pay for it. Why reimbursement, economic value and the hospital CFO decide the fate of many AI tools.
Beyond FDA · Part 1 of 12
FDA clearance answers whether a product can be marketed. Hospitals ask why they should adopt it. The first lesson in the series: clearance is not market access.
Healthcare AI
An introduction to the Beyond FDA series: what international healthcare AI and MedTech companies tend to learn about the U.S. market only after they arrive.
MedTech
Seven lessons U.S. healthcare teaches MedTech founders the hard way, from clearance not equalling market access to why hospitals buy workflows rather than algorithms.
Anurag Velekkatt Sunil Kumar joins Qscription to lead MedTech delivery: SaMD and clinical AI integration programs for U.S. hospital networks.
Benchmark accuracy no longer wins on its own. Why multi-site validation, representative populations and measured workflow impact now decide whether health systems trust an AI product.
Strategy
Detection is becoming a commodity. The next decade of imaging AI will be decided by workflow orchestration, radiologist productivity and outcomes the enterprise can measure.
Strategy
Five signs a healthcare AI startup has built a feature rather than a company, and the moves that turn a single algorithm into a business that owns a workflow.
Accuracy gets the attention; hidden costs decide the ROI. Seven operational costs, from workflow friction to infrastructure overhead, that surface after clearance and stall enterprise radiology AI.
The FDA asks whether a product is safe. The hospital CIO asks what it exposes the enterprise to. Five operational questions that end AI deals after clearance.
Clearance gets the press release. Five operational gaps, in workflow, IT security and support, explain why so many cleared AI products never reach meaningful clinical use.
Announcement
Dr. Nisha Chandran, a board-certified pediatrician and medical director with more than 20 years in clinical leadership, joins our Medical Advisory Board.
Qscription opens its operational office in Jackson, Wisconsin, a base for the team's integration and deployment work with U.S. health systems.
Clearance is one layer. Ten operational pillars, from establishment registration and QMS maturity to PACS integration and post-market duties, that decide whether a foreign SaMD survives hospital review.
A radiology AI company with a 510(k), clinical value and global customers still could not land a U.S. contract. The gap was ecosystem readiness, not the product.
A clinically strong, FDA-cleared radiology AI product with international customers, unable to win a single U.S. hospital. What was missing, and how enterprise readiness closes the gap.
Why cleared SaMD platforms stall in U.S. hospital pilots: the security, workflow and quality-system gaps that surface in four phases, and what successful market entries do differently.
Why AI MedTech companies with strong results abroad slow down at the U.S. border, and how a deployment-led strategy with local integration support changes the trajectory.
International AI MedTech companies with proven products often stall in U.S. hospitals. Five friction points where traditional go-to-market plans break, and the case for a deployment-led approach.
Announcement
Why we started writing in public: patterns we see across many FDA-regulated MedTech teams, practical tooling notes, and quick reads on regulatory news as it lands.