Create an AWS HealthImaging datastore with aws medical-imaging create-datastore --datastore-name MY_DATASTORE and note the returned datastoreId
Upload source DICOM P10 files to an S3 input bucket and grant the HealthImaging service role s3:GetObject on the input bucket and s3:PutObject on the output bucket
Start the import job with aws medical-imaging start-dicom-import-job --datastore-id DATASTORE_ID --input-s3-uri s3://INPUT_BUCKET/prefix/ --output-s3-uri s3://OUTPUT_BUCKET/results/ --data-access-role-arn arn:aws:iam::ACCOUNT:role/ROLE
Poll job status with aws medical-imaging get-dicom-import-job --datastore-id DATASTORE_ID --job-id JOB_ID until jobStatus is COMPLETED
Inspect the job-output-manifest.json in the output S3 bucket to review SUCCESS and FAILURE sub-folders for per-instance import results
Known gotchas
HealthImaging stores images in its own HTJ2K-encoded image set format, not as raw DICOM P10 — pixel data is transcoded at import time; retrieve via GetImageSetMetadata and GetImageFrame, not STOW-RS
Conflicting Study/Series/Instance UIDs between newly imported data and existing image sets cause the new instances to be placed in non-primary image sets rather than failing outright — check the FAILURE folder for metadata conflicts
The service IAM role must be trusted by the medical-imaging.amazonaws.com service principal; a missing trust relationship is a common cause of import job failures
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp