AWS HealthImaging pixel data verification - AWS HealthImaging

AWS HealthImaging pixel data verification

During import, HealthImaging provides built-in pixel data verification by checking the lossless encoding and decoding state of every image. This feature ensures that images decoded using HTJ2K decoding libraries always match the original DICOM P10 images imported into HealthImaging.

  • The image onboarding process begins when an import job captures the original pixel quality state of DICOM P10 images before they are imported. A unique immutable Image Frame Resolution Checksum (IFRC) is generated for each image using the CRC32 algorithm. The IFRC checksum value is presented in the job-output-manifest.json metadata document. For more information, see Understanding import jobs.

  • After the images are imported into a HealthImaging data store and transformed into image sets, the HTJ2K-encoded image frames are immediately decoded and new IFRCs are calculated. HealthImaging then compares the full resolution IFRCs of the original images against the new IFRCs of the imported image frames to verify accuracy.

  • A corresponding per-image descriptive error condition is captured in the import job output log (job-output-manifest.json) for you to review and verify.

To verify pixel data

  1. After importing your medical imaging data, view the per-image set descriptive success (or error condition) captured in the import job output log, job-output-manifest.json. For more information, see Understanding import jobs.

  2. Image sets are comprised of metadata and image frames (pixel data). Image set metadata contains information about associated image frames. Use the GetImageSetMetadata action to get metadata for an image set. For more information, see Getting image set metadata.

  3. The PixelDataChecksumFromBaseToFullResolution contains the IFRC (checksum) for the full resolution image. For images stored in their original transfer syntax 1.2.840.10008.1.2.4.203, 1.2.840.10008.1.2.4.91, 1.2.840.10008.1.2.4.50, and 1.2.840.10008.1.2.1 (Binary Segmentation only) the checksum is calculated on the original image. For images that are stored in HTJ2K Lossless with RPCL, the checksum is calculated on the decoded full resolution image. For more information, see Supported transfer syntaxes.

    Following is example metadata output for the IFRC that is generated as part of the import job process and recorded to job-output-manifest.json.

    "ImageFrames": [{
"ID": "67890678906789012345123451234512",
"PixelDataChecksumFromBaseToFullResolution": [
{
    "Width": 512,
    "Height": 512,
    "Checksum": 2510355201
}
]

    For images stored in their original transfer syntax 1.2.840.10008.1.2.4.203, 1.2.840.10008.1.2.4.91, 1.2.840.10008.1.2.4.50, and 1.2.840.10008.1.2.1 (Binary Segmentation only) the MinPixelValue and MaxPixelValue will not be available. The FrameSizeInBytes indicates the size of the original frame.

    "PixelDataChecksumFromBaseToFullResolution": [
   {"Width": 512, "Height": 512, "Checksum": 1379921327 }
],
"MinPixelValue": null,
"MaxPixelValue": null,
"FrameSizeInBytes": 429

    For images stored in HTJ2K Lossless with RPCL, the FrameSizeInBytes indicates the size of the decoded image frame.

    "PixelDataChecksumFromBaseToFullResolution": [
   {"Width": 512, "Height": 512, "Checksum": 1379921327 }
],
"MinPixelValue": 11,
"MaxPixelValue": 11,
"FrameSizeInBytes": 1652

  4. For instances containing video, HealthImaging performs a lightweight codec validation that verifies the transfer syntax specified in the DICOM metadata matches the video codec.

    HealthImaging computes an IFRC checksum value on the original video object using the CRC32 algorithm. The IFRC checksum value is recorded to the job-output-manifest.json, and persisted in the HealthImaging metadata. As with images stored in their original transfer syntax (described above), the MinPixelValue and MaxPixelValue will not be available. The FrameSizeInBytes indicates the size of the original frame.

  5. HealthImaging verify pixel data, access the Pixel data verification procedure on GitHub and follow the instructions in theREADME.md file to independently verify lossless image processing by the various HTJ2K decoding libraries that are utilized by HealthImaging. After the full image is loaded, you can compute the IFRC for raw input data on your end and compare it with the IFRC value provided in the HealthImaging metadata to verify the pixel data.