From Citrix and paper questionnaires to 900 doctors in a browser

The situation

Circle Institute trains cardiologists, radiologists and technicians in cardiac MRI, and does it well enough that doctors fly in for the courses. That was the ceiling: however many people fit in a room.

Their remote option worked like this. A doctor dialled into a server through Citrix to view DICOM studies. Then they answered questions on a printed sheet. Grading was manual. A retake meant repeating the whole arrangement. Every additional participant was additional work for a human being.

What a generic LMS could not do

The obvious answer was an off-the-shelf learning platform. The obvious answer does not handle cardiac MRI studies. You cannot teach diagnosis from a video, and no standard e-learning tool lets a doctor scroll through real patient scan data in a browser.

The constraint that shaped everything

The platform had to earn CME accreditation from the Ärztekammer. That is not a badge you apply for once the software is finished. It dictates architecture.

It meant full anonymisation of real patient scan data, separate databases for DICOM studies and user records, a formal authentication and authorisation concept including 2FA, and auditable completion records with defensible retake rules.

Building that in afterwards would have meant rebuilding. I designed it in from the start.

The decision that saved a low six-figure budget

They needed a DICOM viewer running in a browser. There were three options.

Build one from scratch: roughly six figures and about a year, to arrive at a solved problem. License one: a comparable range, and nothing on the market actually fit this use case on the web. Or take proven open-source components and put the effort where the value was.

I took the third route. Orthanc as the DICOM server, OHIF as the viewer, wrapped in a custom Next.js shell, with two DICOM servers integrated behind it. Doctors interact with real scan data: no plugin, no desktop client, no Citrix.

The money that would have gone into rebuilding a viewer went into the thing that actually differentiates the product: the training experience around it.

What got built

A full learning environment rather than a course player. Structured progression with gamified case bundles. Interactive questionnaires and diagnosis modules against real studies. Automated grading and re-attempts. Content sold as subscriptions, so institutes can buy individual indications on demand.

I chose AWS for the infrastructure: load-balanced EC2, CloudFront, S3, Lambda. Global delivery and GDPR-compliant data residency both had to hold at once.

Where it is now

Around 900 medical professionals across multiple countries train on the platform, roughly 200 of them paying, and there has been no marketing spend at all. Every user arrived through the institute's existing network.

It is CME-accredited by the Ärztekammer. It launched in early 2024. The nearest comparable platform, from a large German institute, is still in soft launch.

Development has been continuous since January 2024.