Your images leave.
The patient stays.
Annalise.ai and Aidoc keep coming up in strategy meetings. Legal keeps saying no because the images have to leave the network with patient identifiers still attached. PixelIQ removes that blocker - inline, on-prem, before any study reaches any AI vendor.
Before and after PixelIQ.
Everything currently standing between your practice and an Annalise.ai or Aidoc partnership.
Manual de-identification is not a DICOM strategy.
It is a liability waiting to surface.
Most radiology practices rely on PACS export settings and hope. Burnt-in pixel text in CT scouts, private creator tags, and secondary capture objects are invisible to header-only tools - and invisible to the practice until a breach notification arrives.
The PixelIQ pipeline.
Runs inside your network. Nothing raw ever leaves.
Three things that change immediately.
PixelIQ deployed as DICOM proxy
Our team installs the agent between your PACS and the first downstream destination. Radiologists notice nothing. IT does not reconfigure modalities. Typically 4–8 hours on-site.
Annalise.ai or Aidoc connected
Your existing AI vendor relationship connects to the governed pipeline. They receive exactly what the data-use agreement permits. The legal blocker is architecturally removed.
First audit log ready
Every study that moves generates a cryptographic audit entry - timestamped, signed, exportable. From day one you have a compliant record of every data movement, ready for any review.
What PixelIQ covers.
Pixel-level PHI detection
A pixel-level detection model trained on radiology imaging - patient banners, accession overlays, clinic watermarks - identifies and removes PHI invisible to any header-only tool. Covers CT, MR, DX, MG, CR, XA, NM, and all secondary capture objects.
PS3.15 header scrub
All E.1 and E.2 profile attributes processed, configurable retain/remove/pseudonymise per tag.
SR & KO handling
Structured Reports, Key Objects, and Presentation States processed natively - not skipped.
Policy-gated routing
Each destination receives only what its data-use agreement permits. Enforced at network layer.
Immutable audit chain
Every anonymisation decision cryptographically signed. Tamper-evident log for HREC and TGA.
Built for PACS integration, not alongside it.
PixelIQ runs as a transparent DICOM proxy. Your PACS does not know it is there. No workflow changes for radiologists. No integration project for IT.
What radiology teams usually ask us.
Annalise.ai keeps coming up in our AI strategy meetings. Why can't we just share studies directly with them?
You can - with PixelIQ in place. Annalise.ai, Aidoc, and similar vendors receive a governed, PS3.15-compliant dataset. The raw studies with patient identifiers never leave your network. You retain the audit trail. The contract review cycle shortens dramatically because the data-sharing risk is architecturally removed.
Our PACS already has de-identification settings. Isn't that enough?
For most deployments, no. PACS de-identification tools process header fields reliably, but they systematically miss burnt-in pixel text on scout images and overlays, private creator tags from vendors like Siemens and GE, and secondary capture objects (SC SOPs). These surfaces carry patient identifiers that don't appear in standard DICOM headers - and are invisible to the practice until a breach notification arrives.
What about burnt-in pixel text on CT scouts and DX images?
That is the surface PixelIQ was specifically built for. A pixel-level detection model trained on radiology imaging - patient banners, accession number overlays, clinic watermarks - detects and removes pixel-level PHI across every SOP class. Header-only tools cannot do this.
How does PixelIQ interact with our existing PACS and radiology workflow?
PixelIQ operates as a transparent DICOM proxy. Your PACS sends studies the same way it always has. Radiologists see no change. IT does not need to reconfigure modalities. PixelIQ sits inline and processes studies before they reach any downstream destination - on-prem, sub-50ms per study.
What is our liability position under the Australian Privacy Act if something goes wrong?
Under APP 11, you are required to take reasonable steps to protect the personal information you hold. Sending identifiable DICOM to an AI vendor without architectural controls is difficult to defend as 'reasonable'. With PixelIQ, PHI never reaches the external recipient - which is the only framing that clearly satisfies the obligation.
How long does a trial deployment take?
A typical sandbox deployment - where we connect PixelIQ to your PACS, configure the anonymisation profile, and run a live demo against your study formats - takes 4–8 hours on-site. There is no disruption to production workflows during the trial.