Healthimaging Diagnostics

AWS HealthImaging Diagnostics

When to use

Any AWS HealthImaging investigation — datastore creation, DICOM import jobs, image set retrieval, metadata queries, copy/deletion operations, access control, encryption configuration, or integration issues.

Investigation workflow

Step 1 — Collect and triage

aws medical-imaging list-datastores --query 'datastoreSummaries[*].{Name:datastoreName,Id:datastoreId,Status:datastoreStatus}'
aws medical-imaging list-dicom-import-jobs --datastore-id <datastore-id> --query 'jobSummaries[*].{JobId:jobId,JobName:jobName,Status:jobStatus}'
aws medical-imaging search-image-sets --datastore-id <datastore-id> --search-criteria '{"filters":[{"operator":"EQUAL","values":[{"DICOMPatientId":"<patient-id>"}]}]}'

Step 2 — Domain deep dive

aws medical-imaging get-datastore --datastore-id <datastore-id>
aws medical-imaging get-dicom-import-job --datastore-id <datastore-id> --job-id <job-id>
aws medical-imaging get-image-set --datastore-id <datastore-id> --image-set-id <image-set-id>

Step 3 — Detailed investigation

aws cloudtrail lookup-events --lookup-attributes AttributeKey=EventSource,AttributeValue=medical-imaging.amazonaws.com --max-results 20
aws medical-imaging get-image-set-metadata --datastore-id <datastore-id> --image-set-id <image-set-id>
aws medical-imaging get-image-frame --datastore-id <datastore-id> --image-set-id <image-set-id> --image-frame-information '{"imageFrameId":"<frame-id>"}' outfile.jph

Read references/guardrails.md before concluding on any HealthImaging issue.

Tool quick reference

Gotchas: AWS HealthImaging

• HealthImaging stores DICOM data in datastores. Each datastore is an isolated container — image sets cannot be shared across datastores without copy operations.
• DICOM import jobs read from S3 and validate DICOM compliance. Invalid DICOM files are rejected per-file, not per-job — partial imports are possible.
• Image sets are immutable versioned collections. Updates create new versions; previous versions are retained until explicitly deleted.
• Metadata queries use DICOM tag-based search criteria. Not all DICOM tags are indexed — only patient, study, and series level tags are searchable.
• Copy operations are asynchronous and cross-datastore. They create independent copies — changes to the source do not propagate to the copy.
• Deletion is a two-step process: delete the image set, then optionally purge. Deleted image sets may be recoverable within a retention window.
• HealthImaging supports AWS KMS encryption. Datastores use either AWS-managed or customer-managed KMS keys configured at creation time and cannot be changed later.

Anti-hallucination rules

1. Always cite specific datastore IDs, job IDs, or image set IDs as evidence.
2. DICOM import failures can be partial — never assume the entire job failed from a single file error.
3. Image set versions are immutable. Never claim in-place modification is possible.
4. Metadata search is limited to indexed DICOM tags. Never claim arbitrary tag search is supported.
5. Spend no more than 2 minutes on any single hypothesis. Pivot if inconclusive.

12 runbooks