ImageSim is an accredited, competency-based learning platform that trains clinicians to correctly interpret visually diagnosed tests — radiographs, ECGs, ultrasound and clinical photographs. 247 Labs has been part of the engineering team behind it for more than seven years.

What began in 2019 as the replacement of an end-of-life Flash application is now a platform serving over 3,000 health care providers in more than 65 countries, carrying accreditation from four professional colleges and generating data that appears in peer-reviewed journals.

The distance between what ImageSim looks like — an online course catalogue — and what it has to be underneath is the whole story of this engagement.

Visually diagnosed tests are among the most commonly ordered in front-line medicine, and the person ordering them is usually not a radiologist or a cardiologist. It is an emergency physician, a resident, a family doctor, a nurse practitioner. They make the call, and the patient is managed on that call. Disagreement between a front-line read and an expert read is a documented source of medical error — and in pediatrics the margin is thinner still, because a growth plate is not a fracture but can look like one to an eye that has not seen enough of them.

ImageSim’s answer to that gap was established over roughly eleven years of research at the Hospital for Sick Children and the University of Toronto: rather than waiting for cases to arrive one patient at a time, present hundreds of them in a structured sequence, force a commitment on every one, give feedback immediately, and repeat until performance crosses a defined standard.

By 2019 the evidence was settled and the software was the bottleneck. The existing platform was a Flash application with a Flash-based case loader behind it, months from end-of-life, carrying a content library that had taken years to assemble. The brief called for a full rebuild on modern web technology, migration of the existing content, and — the part that mattered most day to day — a far easier way for clinicians to author new cases and courses.

The harder constraint was invisible in the brief. The case-mix ratio, the feedback timing, the randomization behaviour, the mastery threshold: these look like configuration values. They are the clinical intervention, each one traceable to a published finding. Rebuilding the platform meant preserving an evidence base, not just porting features.

247 Labs specified and delivered a monolithic Laravel application on PHP 7, with Vue.js driving the purchasing flow and Stripe handling payments inline rather than bouncing users to a hosted checkout. The case authoring tool was built on Fabric.js, so a content admin can draw hotspots, highlights, arrows and annotations directly onto a diagnostic image in the browser — the same canvas machinery that captures where a learner believes a finding is located.

A three-metric competency engine. Performance is defined by accuracy, sensitivity and specificity, and competency requires clearing an administrator-set threshold on all three simultaneously. A learner who calls everything abnormal scores brilliantly on sensitivity and fails medicine; one who calls everything normal looks fine on specificity and misses fractures. Only the three together describe a clinician you would trust. Thresholds are set per course rather than hard-coded, which is what allows the catalogue to run standards tuned to clinical consequence — 80% on dental bitewings, 85% on chest radiographs and ECG, 90% on the elbow, pelvis, ankle and child-abuse identification courses.

Competency that latches. Once a learner meets the standard, the determination holds even if subsequent performance drifts below it. The achievement is a fact about a moment that was reached, not a live scoreboard that punishes continued practice — a deliberate pedagogical decision that prevents the system from discouraging practice past certification.

Eight user roles reflecting how medical education actually buys. Trainee, faculty, program director, super-director, manager, coordinator, platform admin and case admin. A program director purchases seats in bulk for a cohort; a coordinator handles logistics without seeing grades; a manager runs daily operations without spending money; content authorship is delegated per course to the clinicians who own that subject matter.

Assessment integrity over the open web. The answer key cannot reach the browser before the learner commits. If it does, the assessment is worthless and so is every research finding drawn from it. Mitigations run from bundling and minification through strictly same-origin API scoping to the fundamental rule of never shipping the diagnosis until an answer is locked in.

Engineering for the environments it actually runs in. Non-functional requirements assumed roughly 1 Mbps and set a battery budget of six hours on mobile and two on a notebook under continuous use — because the platform is used in resource-limited settings, and the media in question is high-resolution diagnostic imagery, now including video for the ultrasound courses. Database collation handles French, Spanish and other keyboard entry; account retrieval is case-insensitive; surnames run past thirty characters. In a platform used by clinicians in dozens of countries, name handling is not an edge case.

“For the past 7 years, we have entrusted 247 Labs to help us set up and run a large number of research studies in addition to assisting in expanding the capabilities of our global online medical educational platform. We have relied on 247 Labs team to provide a range of services that included software developers, UI and UX support, dev ops and backend expertise in order to generate a cross platform solution for us. We have been very happy with the results.”