While attempting to import programmatically a specific “subdirectory” of DICOM objects from an S3 bucket into a HealthImaging data store, I’m getting a “No file matches the provided input S3 URI” error. Meanwhile, I can visually confirm on the S3 console that there are files in that specific bucket including the full “subdirectory” prefix. It seem that the files are there but cannot be found.

The desired steps are:

• Programmatically upload one or more DICOM files to an S3 bucket (bucketname/level1prefix/level2prefix). This step uses TranferUtility and UploadAsync.

• Once upload is complete, invoke StartDICOMImportJobAsync to import from the S3 bucket to an existing HealthImaging data store.

The call to StartDICOMImportJobAsync works fine when the InputS3Uri looks like this:

s3://bucketname/level1prefix/

But it fails when the InputS3Uri looks like this:

s3://bucketname/level1prefix/level2prefix/

I’d like it to work the second way because I may have multiple S3 upload jobs that use the same level1prefix (e.g., Accession Number or Study Instance UID) but with unique prefixes at the second level. The level2prefix is a unique job id GUID.

To troubleshoot, I tried the same import in the AWS HealthImaging Console and could successfully Import DICOM data from bucketname/level1prefix/level2prefix into my data store.

I also tried adding a wait after the S3 upload step before attempting the import in case this is a timing issue, but that made no difference. I also tried a ListObjects call to see what was in the sub-bucket (it claims it doesn’t exist) while visually confirming in the S3 console that the sub-bucket (with objects) does exist.

Since it works from the console and not the API call, it feels like maybe it is either a timing issue or a prefix parsing error where maybe the API call only takes everything up to the second delimiter (forward slash) or otherwise fails to read my mind. It’s entirely possible that my expectation of what should work is incorrect, but I couldn’t find anything in the online docs that provides clarity.

Is there a way to programmatically import from an S3 sub-bucket into a HealthImaging data store? Anyone have experience with import from S3 to HealthImaging who could provide a glimmer of insight?