Medical Imaging Archive Migration and Exit on AWS HealthImaging

The problem: your archive is the lock-in

Nobody renews a PACS because they love it. They renew because the archive is inside it.

Ten or twenty years of studies sit in a vendor database, often in a proprietary wrapper rather than clean DICOM Part 10 on disk. The migration quote is priced at what the data is worth to you, not what the work costs, with a timeline in quarters. The hardware is on a refresh cycle, the DR copy has never been restored end to end, and the archive only grows, because no clinician will sign off deleting anything.

So the exit question, "how do we get our data out", has no credible answer. That is why the renewal gets signed.

What AWS HealthImaging changes

A DICOM-native managed service with an open API surface, not a black box.

• DICOMweb throughout: QIDO-RS to search, WADO-RS to retrieve, STOW-RS to store, and BulkData URIs so large elements come down separately from metadata.
• Frame-level access. Pixel data is addressable per frame; metadata returns as JSON independently of pixels. Viewers stream progressively instead of pulling whole studies, and AI pipelines read the archive directly with no PACS in the path.
• Any SOP class UID, including retired and private, with all private attributes preserved. AWS's own statement, and the difference between a migration that completes and one that stalls on the first vendor-specific object.
• Sixteen transfer syntaxes accepted on import, stored as HTJ2K lossless by default, or JPEG 2000 Lossless if set at datastore creation.
• Deterministic image set identifiers. Re-importing a study returns the same identifier, which is what makes the programme reversible.
• It behaves the same at 5 GB and at 100 TB. No SAN sizing, no capex cycle, no re-architecture as you grow. Storage auto-tiers to Archive Instant Access, roughly 93 percent cheaper than active, with no code change and no loss of access.

What we have actually done

Delivered systems, not a reference architecture.

• Multiple archive migrations in the 20 TB and above range, across legacy PACS and departmental estates.
• A full cardiology archive taken out of one vendor's estate, through HealthImaging for normalisation and clinical retention review, and into a different vendor's system. Roughly 25,000 retained studies, over 2,000 archive days, 131,492 files reconciled against counts at both ends.

Modalities and object types

Angiography and interventional (XA), ultrasound including lossy-compressed, CT, MR, CR and DX, secondary capture, structured reports, presentation states and non-image objects. On live archives we have carried cardiac electrophysiology waveforms and vendor private SOP classes, which generic tooling skips silently, and the usual reason a count reconciliation fails at the end of a migration.

Why the cleanup is the hidden win

Most archive projects move everything, because deciding what not to move feels harder than paying to keep it. It is the other way round.

On a delivered archive, clinical review retired 58 percent of studies. That is permanently less storage, permanently less protected health information to defend, a smaller breach surface, a retention position you can evidence to a privacy officer, and a migration that moves half the data.

It only works if the process is provable. A ledger row is written for every image set before it is deleted and never after. Every run follows a dry run that reconciles with zero deletions. Execution needs a typed count-matching confirmation. The restore path is tested on real studies before anything is removed. The audit trail carries no patient identifiers, so it can go straight to an auditor.

Export: the direction nobody else sells

Retrieval out of HealthImaging to DICOM Part 10 in whatever transfer syntax the receiving system demands, with per-batch manifests written last as the commit marker, count and checksum reconciliation at both ends, and resume after interruption. Acceptance is reconciliation, not an opinion.

Support

Enquiries through this listing are answered within two business days, with a scoping call and a written proposal. After purchase you get a named engagement lead, weekly status reporting and support during New Zealand business hours. Response targets by severity are in the statement of work attached to each offer.

Documentation

https://www.easycoder.co 

Infrastructure costs

AWS HealthImaging, storage, retrieval and data transfer are billed by AWS directly to you, separately from this engagement. Incumbent vendor licensing stays between you and that vendor until the source is decommissioned. Our access is through least-privilege IAM roles we document and hand over.