Five weeks from first call to 420 studies
Subgroup analysis is what tells a developer where an algorithm is weak — which pathologies, which image quality, which confounding findings. That requires a complete and consistent label set on every study, not a headline accuracy figure.
The client needed 900 non-contrast head CT studies labelled by experienced US neuroradiologists, split into three priority batches with staggered deadlines. The most time-critical batch, 420 studies, was due on 10 June — five weeks after the first request. Recruitment, onboarding and reading for that batch all had to fit inside that window.
Four fellowship-trained neuroradiologists, each approved by the client
Four US board-certified radiologists, all fellowship-trained in neuroradiology, each meeting the client's requirement of ten or more years of experience. The client approved every profile before onboarding.
The original plan called for three readers. To protect the 10 June deadline, Medcase proposed and onboarded a fourth to spread the workload — a change to the staffing plan made while the clock was running, not after the first batch slipped.
Onboarding compressed to about a week
A live onboarding session would have forced every reader to wait for a shared date. Instead, a recorded training video prepared by the client replaced it, so each reader could start as soon as their contract was in place.
Readers received the project annotation protocol and the video, then completed two training studies in the annotation platform. The client reviewed every training study before the reader moved to production. All training studies passed review, and all four readers finished training within about a week of the video.
A complete label set on every study
Each study was labelled by one expert reader, as specified by the client's study design. For each study, readers recorded:
Image quality
An overall rating, plus the presence of motion, beam-hardening, metal, noise and ring artifacts.
Pathologies
Acute intracranial hemorrhage with location, chronic or subacute hemorrhage, hydrocephalus, herniation, midline shift grade, acute ischemia, edema, hyperdense vessel, chronic infarct, mass, pneumocephalus and ventricular findings.
Confounders
Degenerative confounders — atrophy, small vessel disease and intracranial calcifications with location — and surgical confounders, including hardware and other surgical signs.
Batched by priority, read in parallel
The client uploaded the 900 studies as three priority batches, so the most time-critical studies were read and delivered first. As readers finished their Priority 2 assignments they moved straight on to Priority 3 rather than waiting for the batch to close, which is why the second and third batches landed on the same day.
Validation built into the annotation tool
Medcase set up automated validation in the annotation platform that checked each study against the annotation protocol before it could be submitted, so incomplete or inconsistent labels were caught and corrected by the reader rather than found later.
The client reviewed each delivered batch. No rework was requested on any of them.
Project volume and timeline
Work package | Count |
|---|---|
Production studies (single axial soft-tissue series each) | 900 |
Priority 1 batch (highly time-critical) | 420 |
Priority 2 batch (medium) | 359 |
Priority 3 batch (low) | 121 |
Readers per study | 1 |
Training studies per reader, reviewed by the client | 2 (plus 1 optional) |
Phase | Dates (2026) | Client deadline |
|---|---|---|
Client request and first meeting | 6–7 May | — |
Reader recruitment, four readers contracted | 7–18 May | — |
Training and client review | 19–28 May | — |
Production reading | 21 May – 30 June | — |
Priority 1 delivered (420 studies) | 2 June | 10 June |
Priority 2 delivered (359 studies) | 30 June | 30 June |
Priority 3 delivered (121 studies) | 30 June | 31 July |
First request to final delivery: about eight weeks. All three batches were delivered by 30 June, completing the full dataset a month before the final batch's own deadline.
Results: 900 studies, every deadline met or beaten
- Study-level labels delivered for all 900 non-contrast head CT studies by 30 June, supporting subgroup analysis of the client's intracranial hemorrhage and midline shift algorithms.
- Priority 1, the 420 highly time-critical studies, delivered eight days before its 10 June deadline.
- Priority 2 delivered on its deadline; Priority 3 delivered a month ahead of its own.
- Four senior US neuroradiologists recruited, trained and in production within about two weeks of the first request.
- All training studies passed client review, and no rework was requested on any production batch.