Quality assurance in radiology here is practice / report QA: peer review, discrepancy, turnaround on critical results, and a culture that treats a miss as a lesson. It is not the imaging-QA program (scanners, SNR, phantoms). It is not the daily QC test list. It is not “what is AI in radiology.”
If you meant imaging QA as a program (equipment / phantoms / ACR) → medical imaging quality assurance. If you meant QC tests (daily / weekly / reject analysis) → quality control in radiology. If you meant what is AI in radiology → artificial intelligence in radiology. If you meant worklist / interruptions → radiology workflow optimization.
If the work is overlaying a QC mark or running an imaging model inside a department QA/QC stack, PYCAD is the imaging piece (viewer / model), not a QA/QC vendor.
What this job is
Image acquisition can be perfect and the report still wrong. This page is the second half: did the interpretation hold, was the critical result out in time, and does the department learn when it did not.
A “recurring error in acquisition” is a real problem. It is not this URL. Acquisition / phantom / ACR live on 517 and 562.
A program that is not a blame ritual
| Component | What it does | What “working” looks like |
|---|---|---|
| Peer review | A second radiologist reads a sample (or a flagged case) for learning, not a scoreboard | Discrepancies discussed; people still send cases in |
| Quality metrics | Discrepancy rate, TAT on critical results, peer-review scores — numbers that change a roster or a protocol | A bottleneck you can name, not a dashboard nobody opens |
| Workflow integration | QA hooks in the RIS so data is not a side spreadsheet | Fewer re-keys; the worklist job is still 694 |
| Quality champions | Someone in the reading room who owns the ritual | Participation without a compliance memo |
Peer review that is only a hunt for misses makes people hide cases. Peer review that is a conference — “here is the pattern, here is the next protocol” — is the job.
Metrics that are not SNR
| Metric | What it tells you | What to do with it |
|---|---|---|
| Discrepancy rate | Share of reports that needed an amendment | Start here. Then slice by modality, finding, and reader — a raw % does not name a cause |
| Turnaround time (TAT) | Exam complete → report available, especially critical results | A queue problem, a staffing problem, or a “everything is STAT” problem |
| Peer-review scores | Accuracy / completeness on the sample | Training and protocol change, not a public ranking |
| Patient-facing signals | Communication and wait, not HU | Useful for the front desk and the critical-result path; not a substitute for discrepancy |
A chest radiograph discrepancy cluster on subtle nodules is a training or protocol change (or a detection tool). It is not a phantom score. Do not steal 517’s SNR chapter onto this URL.
Shortage is a constraint, not a costume
The Royal College of Radiologists’ 2023 clinical radiology census put the UK consultant shortfall around 30% — on the order of 2,000 posts below a safe establishment. That is a staffing fact, not a reason to skip peer review. The departments that keep QA under a short list do three unglamorous things:
- Triage the sample. High-risk studies and known miss patterns get the second look. Random 5% on everything is how the ritual dies when the list is long.
- Peer learning, not a tribunal. A short conference beats a silent score in a folder.
- Support staff own the logistics. Case pull, RIS fields, the TAT clock — if the radiologist is also the clerk, QA is the first thing that slips.
Worklist design is radiology workflow optimization. This page does not become a staffing explainer.
AI is not this product
A detection model that flags a candidate on the worklist is AI in radiology, not a report-QA platform and not a PYCAD peer-review product. Use it as a second set of eyes on a named finding. Do not confuse a vendor overlay with a peer-review program. PYCAD does not ship “AI-powered peer review assistance.”
FAQ
Is this the same as medical imaging QA?
No. Imaging QA is equipment + process + personnel + image metrics. That is 517. This page is whether the report held.
Where do daily phantom tests go?
Does PYCAD sell radiology QA?
No. No peer-review product either. See the line above.