Zapx Labs
Home/Products/RadQA
RadiologyPilotBuilt by Zapx Labs
RadQA logo

RadQA

Catch laterality conflicts, uncommunicated critical findings, protocol deviations, and image-quality defects before they leave the department — with evidence marked on the report and a human approving every write-back.

Register for the pilotWork With Us

RadQA is a quality-assurance assistant for imaging departments. It is not a diagnostic device and does not replace radiologist reporting or clinical judgement. Validate against your own systems and local governance before production use.

Radiology

RadQA

Live preview
RadQA live product screenshot

Who it’s for

Radiology and imaging departments, QA leads, radiographers, and radiologists who want systematic quality assurance without adding manual review load.

Problem

Radiology QA is mostly retrospective sampling done by hand. Discrepancies, template leftovers, missed critical-result communication, and acquisition errors slip through because nobody has time to check every study.

How it works

  1. 1

    Connect a source — DICOMweb PACS, HL7 v2 RIS feed, FHIR R4 server, or a local DICOM folder.

  2. 2

    QA flavours run on each study: report QA, protocol QA, image QA, and peer review, combining deterministic rules with model-backed checks.

  3. 3

    Findings land in a worklist with evidence highlighted; the agent helps investigate, and any write-back needs explicit approval.

Key features

  • Report QA: laterality, impression vs findings, critical-result communication, follow-up, template leftovers
  • Protocol QA: missing series, contrast and slice-thickness mismatches, body part and modality checks
  • Image QA: markers, exposure, collimation, clipped anatomy, rotation, burned-in identifiers
  • RADPEER-style peer review with deterministic sampling
  • Pluggable connectors: FHIR, DICOMweb, HL7 v2, local files, demo data
  • Claude-powered agent with approval-gated external actions
  • Runs on the desktop; API keys stay in the OS keychain

Inside RadQA

Screens from the current build. Details may change as the product develops.

RadQA report QA findings on a CTA chest study, including a critical finding with no documented communication
Report QA — a critical finding of bilateral pulmonary emboli is flagged because nothing in the report records that it was communicated, alongside two nodules described without a follow-up recommendation. Every finding quotes the sentence it came from.
RadQA in dark mode showing protocol QA findings on a real CT pulmonary angiogram
Protocol QA — the template for a CT pulmonary angiogram expects IV contrast and specific series. RadQA flags the missing contrast, the acquisitions that never arrived, and a slice thickness above the diagnostic limit.
RadQA image viewer in dark mode showing slice 18 of a real head and neck CT
Images — DICOM is decoded in the app, with stack scroll, window and level, zoom and pan, and image-QA findings overlaid as regions on the pixels. Reading happens in dark mode, where the app follows your system theme.
The RadQA agent explaining findings on an open case and proposing next steps
The agent knows which case is open. Ask it to review the findings and it explains the likely root cause, ranks what matters clinically, and proposes next steps for you to approve or decline.
RadQA analytics dashboard in dark mode showing case volume, mean QA score, turnaround and the critical communication gap
Analytics — volume, mean QA score, turnaround, SLA breaches, and the critical-communication gap across a chosen range, filterable by modality and by connector.
RadQA connectors page listing configured sources and available connector types
Connectors — studies and reports come in from FHIR servers, DICOMweb PACS, HL7 v2 RIS feeds over MLLP, a local folder, or the built-in demo dataset. Write-back goes out the same way.

Example journey

Step 1

Connect your systems

Point RadQA at PACS, RIS, or FHIR, or start with the built-in demo dataset.

Step 2

Run QA

Studies and reports are checked automatically as they arrive or on demand.

Step 3

Review the worklist

Open a case to see findings with evidence marked directly on the report and images.

Step 4

Act with oversight

Ask the agent to investigate; addenda and write-backs need explicit approval.

Get updates

Current status: Pilot

FAQ

RadQA is built by Zapx Labs.