Medical imaging quality assurance is the program that keeps every diagnostic image clear, accurate, and safe: equipment that still performs, protocols that do not drift, people who can run both. It is not peer review of the report. It is not the daily phantom / reject-analysis test list. It is not device IQ/OQ/PQ or a healthcare data-quality program.
If you meant radiology peer review / report QA → quality assurance in radiology. If you meant QC tests (phantoms / daily / weekly / reject analysis / ACR) → quality control in radiology. If you meant device IQ/OQ/PQ → medical device validation process. If you meant device QMS / ISO 13485 → medical device quality management system. If you meant healthcare data quality → data quality in healthcare.
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.
Three pillars
Skip one pillar and the program is a costume. The machine, the recipe, and the people have to hold at the same time.
- Equipment performance. Acceptance testing on install. Routine QC with phantoms so drift shows up before a patient scan. Annual physicist eval (dose, spatial resolution, the numbers that do not care about last week’s anecdote). The full test schedule is 562.
- Process consistency. Same protocol for the same clinical question, whoever is on shift. A six-month follow-up is useless if the baseline used a different recipe. Labeling, prep, and acquisition settings are part of the program, not folklore.
- Personnel competency. Credentialing on day one, then ongoing education and a review that is about the work, not a certificate on the wall. A new protocol or a new gantry that nobody was trained on is how the other two pillars fail in public.
This is the imaging-QA program. Report peer review lives on 561. Device process validation is 699.
What the checks look like, by modality
| Modality | Typical QA check | What it proves |
|---|---|---|
| X-ray / DR | Collimator and beam alignment | Light field matches the beam; you are not irradiating anatomy you did not mean to |
| CT | CT number accuracy on a water phantom | Water stays ~0 HU. Density is a number, not a vibe |
| MRI | Signal-to-noise ratio | Useful signal still beats grain. A sudden SNR drop is a service call |
| Ultrasound | Geometric accuracy on a pin phantom | A measured centimeter is still a centimeter (biometry, organ size) |
The goal is the same on every gantry. The physics is not. Steal the schedule and the pass/fail numbers from quality control in radiology; this page only names the job.
What “good” is: SNR, spatial resolution, phantom
A radiologist’s eye is the last gate. It is not a substitute for a number you can trend.
Signal-to-noise ratio. Signal is the anatomy. Noise is grain. High SNR is how a small lesion stays visible. You can buy SNR with time or dose; the program is the written compromise, not “crank it until it looks pretty.” A step-change on the daily phantom is a hardware problem, not a bad patient.
Spatial resolution. Can the system keep two adjacent small objects apart? That is microcalcifications, inner-ear ossicles, distal vessels. Blur those and the report is guessing.
Phantom testing. Patients are not a calibration standard. A phantom is a known object you scan on a schedule so SNR and resolution have a log. That log is how a physicist sees drift before a clinic day dies. How often, which phantom, who signs — that is 562, not a second QA-program article.
Pre-scan, during, post-scan
The program follows the study, not just the gantry.
- Pre-scan. Right patient, right order, right protocol. Metal / contrast / implant flags before the table moves. Wrong protocol is a repeat and a dose you did not owe.
- During. Position, artifact watch, ALARA on CT and X-ray. The technologist is doing QC in real time; the phantom is not a substitute for a moving patient.
- Post-scan. Reconstructions as needed, then PACS. The radiologist’s image-quality check (coverage, sharpness, artifact) is still image QA. Peer review of the report is 561.
AI that flags motion, noise, or a clipped FOV before the study hits PACS is a reject helper, not a QA program. Predictive-maintenance dashboards are a maintenance tool. Neither is a reason to flatten this URL into AI in radiology.
FAQ
What is the difference between quality assurance and quality control?
QA is the program: training, protocols, documentation, the three pillars. QC is the test: today’s phantom, this week’s uniformity, the reject log. If you came here for the tests, go to quality control in radiology.
How often should equipment be tested?
Daily or weekly technologist checks catch drift. Annual (sometimes semi-annual) physicist eval is the deep pass. Exact frequency is modality- and ACR/state-specific — the table lives on 562.
Is this the same as device IQ/OQ/PQ or a QMS?
No. Imaging QA is how a department keeps scanners and studies honest. Installation / operational / performance qualification of a device line is 5503. ISO 13485 / 21 CFR 820 is 7869. Record-level data quality is 6055.
Does PYCAD build imaging QA?
No. The imaging piece, not a QA program. See the line above.