開始使用 HealthImaging 的影像集和影像影格 AWS SDK

使用必要的資源建立 AWS CloudFormation 堆疊。

    Aws::String inputBucketName;
    Aws::String outputBucketName;
    Aws::String dataStoreId;
    Aws::String roleArn;
    Aws::String stackName;

    if (askYesNoQuestion(
            "Would you like to let this workflow create the resources for you? (y/n) ")) {
        stackName = askQuestion(
                "Enter a name for the AWS CloudFormation stack to create. ");
        Aws::String dataStoreName = askQuestion(
                "Enter a name for the HealthImaging datastore to create. ");

        Aws::Map<Aws::String, Aws::String> outputs = createCloudFormationStack(
                stackName,
                dataStoreName,
                clientConfiguration);

        if (!retrieveOutputs(outputs, dataStoreId, inputBucketName, outputBucketName,
                             roleArn)) {
            return false;
        }

        std::cout << "The following resources have been created." << std::endl;
        std::cout << "A HealthImaging datastore with ID: " << dataStoreId << "."
                  << std::endl;
        std::cout << "An Amazon S3 input bucket named: " << inputBucketName << "."
                  << std::endl;
        std::cout << "An Amazon S3 output bucket named: " << outputBucketName << "."
                  << std::endl;
        std::cout << "An IAM role with the ARN: " << roleArn << "." << std::endl;
        askQuestion("Enter return to continue.", alwaysTrueTest);
    }
    else {
        std::cout << "You have chosen to use preexisting resources:" << std::endl;
        dataStoreId = askQuestion(
                "Enter the data store ID of the HealthImaging datastore you wish to use: ");
        inputBucketName = askQuestion(
                "Enter the name of the S3 input bucket you wish to use: ");
        outputBucketName = askQuestion(
                "Enter the name of the S3 output bucket you wish to use: ");
        roleArn = askQuestion(
                "Enter the ARN for the IAM role with the proper permissions to import a DICOM series: ");
    }

將DICOM檔案複製到 Amazon S3 匯入儲存貯體。

    std::cout
            << "This workflow uses DICOM files from the National Cancer Institute Imaging Data\n"
            << "Commons (IDC) Collections." << std::endl;
    std::cout << "Here is the link to their website." << std::endl;
    std::cout << "https://registry.opendata.aws/nci-imaging-data-commons/" << std::endl;
    std::cout << "We will use DICOM files stored in an S3 bucket managed by the IDC."
              << std::endl;
    std::cout
            << "First one of the DICOM folders in the IDC collection must be copied to your\n"
               "input S3 bucket."
            << std::endl;
    std::cout << "You have the choice of one of the following "
              << IDC_ImageChoices.size() << " folders to copy." << std::endl;

    int index = 1;
    for (auto &idcChoice: IDC_ImageChoices) {
        std::cout << index << " - " << idcChoice.mDescription << std::endl;
        index++;
    }
    int choice = askQuestionForIntRange("Choose DICOM files to import: ", 1, 4);

    Aws::String fromDirectory = IDC_ImageChoices[choice - 1].mDirectory;
    Aws::String inputDirectory = "input";

    std::cout << "The files in the directory '" << fromDirectory << "' in the bucket '"
              << IDC_S3_BucketName << "' will be copied " << std::endl;
    std::cout << "to the folder '" << inputDirectory << "/" << fromDirectory
              << "' in the bucket '" << inputBucketName << "'." << std::endl;
    askQuestion("Enter return to start the copy.", alwaysTrueTest);

    if (!AwsDoc::Medical_Imaging::copySeriesBetweenBuckets(
            IDC_S3_BucketName,
            fromDirectory,
            inputBucketName,
            inputDirectory, clientConfiguration)) {
        std::cerr << "This workflow will exit because of an error." << std::endl;
        cleanup(stackName, dataStoreId, clientConfiguration);
        return false;
    }

將DICOM檔案匯入 Amazon S3 資料存放區。

bool AwsDoc::Medical_Imaging::startDicomImport(const Aws::String &dataStoreID,
                                               const Aws::String &inputBucketName,
                                               const Aws::String &inputDirectory,
                                               const Aws::String &outputBucketName,
                                               const Aws::String &outputDirectory,
                                               const Aws::String &roleArn,
                                               Aws::String &importJobId,
                                               const Aws::Client::ClientConfiguration &clientConfiguration) {
    bool result = false;
    if (startDICOMImportJob(dataStoreID, inputBucketName, inputDirectory,
                            outputBucketName, outputDirectory, roleArn, importJobId,
                            clientConfiguration)) {
        std::cout << "DICOM import job started with job ID " << importJobId << "."
                  << std::endl;
        result = waitImportJobCompleted(dataStoreID, importJobId, clientConfiguration);
        if (result) {
            std::cout << "DICOM import job completed." << std::endl;

        }
    }

    return result;
}

//! Routine which starts a HealthImaging import job.
/*!
  \param dataStoreID: The HealthImaging data store ID.
  \param inputBucketName: The name of the Amazon S3 bucket containing the DICOM files.
  \param inputDirectory: The directory in the S3 bucket containing the DICOM files.
  \param outputBucketName: The name of the S3 bucket for the output.
  \param outputDirectory: The directory in the S3 bucket to store the output.
  \param roleArn: The ARN of the IAM role with permissions for the import.
  \param importJobId: A string to receive the import job ID.
  \param clientConfig: Aws client configuration.
  \return bool: Function succeeded.
  */
bool AwsDoc::Medical_Imaging::startDICOMImportJob(
        const Aws::String &dataStoreID, const Aws::String &inputBucketName,
        const Aws::String &inputDirectory, const Aws::String &outputBucketName,
        const Aws::String &outputDirectory, const Aws::String &roleArn,
        Aws::String &importJobId,
        const Aws::Client::ClientConfiguration &clientConfig) {
    Aws::MedicalImaging::MedicalImagingClient medicalImagingClient(clientConfig);
    Aws::String inputURI = "s3://" + inputBucketName + "/" + inputDirectory + "/";
    Aws::String outputURI = "s3://" + outputBucketName + "/" + outputDirectory + "/";
    Aws::MedicalImaging::Model::StartDICOMImportJobRequest startDICOMImportJobRequest;
    startDICOMImportJobRequest.SetDatastoreId(dataStoreID);
    startDICOMImportJobRequest.SetDataAccessRoleArn(roleArn);
    startDICOMImportJobRequest.SetInputS3Uri(inputURI);
    startDICOMImportJobRequest.SetOutputS3Uri(outputURI);

    Aws::MedicalImaging::Model::StartDICOMImportJobOutcome startDICOMImportJobOutcome = medicalImagingClient.StartDICOMImportJob(
            startDICOMImportJobRequest);

    if (startDICOMImportJobOutcome.IsSuccess()) {
        importJobId = startDICOMImportJobOutcome.GetResult().GetJobId();
    }
    else {
        std::cerr << "Failed to start DICOM import job because "
                  << startDICOMImportJobOutcome.GetError().GetMessage() << std::endl;
    }

    return startDICOMImportJobOutcome.IsSuccess();
}


//! Routine which waits for a DICOM import job to complete.
/*!
 * @param dataStoreID: The HealthImaging data store ID.
 * @param importJobId: The import job ID.
 * @param clientConfiguration : Aws client configuration.
 * @return  bool: Function succeeded.
 */
bool AwsDoc::Medical_Imaging::waitImportJobCompleted(const Aws::String &datastoreID,
                                                     const Aws::String &importJobId,
                                                     const Aws::Client::ClientConfiguration &clientConfiguration) {

    Aws::MedicalImaging::Model::JobStatus jobStatus = Aws::MedicalImaging::Model::JobStatus::IN_PROGRESS;
    while (jobStatus == Aws::MedicalImaging::Model::JobStatus::IN_PROGRESS) {
        std::this_thread::sleep_for(std::chrono::seconds(1));

        Aws::MedicalImaging::Model::GetDICOMImportJobOutcome getDicomImportJobOutcome = getDICOMImportJob(
                datastoreID, importJobId,
                clientConfiguration);

        if (getDicomImportJobOutcome.IsSuccess()) {
            jobStatus = getDicomImportJobOutcome.GetResult().GetJobProperties().GetJobStatus();

            std::cout << "DICOM import job status: " <<
                      Aws::MedicalImaging::Model::JobStatusMapper::GetNameForJobStatus(
                              jobStatus) << std::endl;
        }
        else {
            std::cerr << "Failed to get import job status because "
                      << getDicomImportJobOutcome.GetError().GetMessage() << std::endl;
            return false;
        }
    }

    return jobStatus == Aws::MedicalImaging::Model::JobStatus::COMPLETED;
}

//! Routine which gets a HealthImaging DICOM import job's properties.
/*!
  \param dataStoreID: The HealthImaging data store ID.
  \param importJobID: The DICOM import job ID
  \param clientConfig: Aws client configuration.
  \return GetDICOMImportJobOutcome: The import job outcome.
*/
Aws::MedicalImaging::Model::GetDICOMImportJobOutcome
AwsDoc::Medical_Imaging::getDICOMImportJob(const Aws::String &dataStoreID,
                                           const Aws::String &importJobID,
                                           const Aws::Client::ClientConfiguration &clientConfig) {
    Aws::MedicalImaging::MedicalImagingClient client(clientConfig);
    Aws::MedicalImaging::Model::GetDICOMImportJobRequest request;
    request.SetDatastoreId(dataStoreID);
    request.SetJobId(importJobID);
    Aws::MedicalImaging::Model::GetDICOMImportJobOutcome outcome = client.GetDICOMImportJob(
            request);
    if (!outcome.IsSuccess()) {
        std::cerr << "GetDICOMImportJob error: "
                  << outcome.GetError().GetMessage() << std::endl;
    }

    return outcome;
}

取得DICOM匯入任務建立的影像集。

bool
AwsDoc::Medical_Imaging::getImageSetsForDicomImportJob(const Aws::String &datastoreID,
                                                       const Aws::String &importJobId,
                                                       Aws::Vector<Aws::String> &imageSets,
                                                       const Aws::Client::ClientConfiguration &clientConfiguration) {
    Aws::MedicalImaging::Model::GetDICOMImportJobOutcome getDicomImportJobOutcome = getDICOMImportJob(
            datastoreID, importJobId, clientConfiguration);
    bool result = false;
    if (getDicomImportJobOutcome.IsSuccess()) {
        auto outputURI = getDicomImportJobOutcome.GetResult().GetJobProperties().GetOutputS3Uri();
        Aws::Http::URI uri(outputURI);
        const Aws::String &bucket = uri.GetAuthority();
        Aws::String key = uri.GetPath();

        Aws::S3::S3Client s3Client(clientConfiguration);
        Aws::S3::Model::GetObjectRequest objectRequest;
        objectRequest.SetBucket(bucket);
        objectRequest.SetKey(key + "/" + IMPORT_JOB_MANIFEST_FILE_NAME);

        auto getObjectOutcome = s3Client.GetObject(objectRequest);
        if (getObjectOutcome.IsSuccess()) {
            auto &data = getObjectOutcome.GetResult().GetBody();

            std::stringstream stringStream;
            stringStream << data.rdbuf();

            try {
                // Use JMESPath to extract the image set IDs.
                // https://jmespath.org/specification.html
                std::string jmesPathExpression = "jobSummary.imageSetsSummary[].imageSetId";
                jsoncons::json doc = jsoncons::json::parse(stringStream.str());

                jsoncons::json imageSetsJson = jsoncons::jmespath::search(doc,
                                                                          jmesPathExpression);\
                for (auto &imageSet: imageSetsJson.array_range()) {
                    imageSets.push_back(imageSet.as_string());
                }

                result = true;
            }
            catch (const std::exception &e) {
                std::cerr << e.what() << '\n';
            }

        }
        else {
            std::cerr << "Failed to get object because "
                      << getObjectOutcome.GetError().GetMessage() << std::endl;
        }

    }
    else {
        std::cerr << "Failed to get import job status because "
                  << getDicomImportJobOutcome.GetError().GetMessage() << std::endl;
    }

    return result;
}

取得影像集的影像影格資訊。

bool AwsDoc::Medical_Imaging::getImageFramesForImageSet(const Aws::String &dataStoreID,
                                                        const Aws::String &imageSetID,
                                                        const Aws::String &outDirectory,
                                                        Aws::Vector<ImageFrameInfo> &imageFrames,
                                                        const Aws::Client::ClientConfiguration &clientConfiguration) {
    Aws::String fileName = outDirectory + "/" + imageSetID + "_metadata.json.gzip";
    bool result = false;
    if (getImageSetMetadata(dataStoreID, imageSetID, "", // Empty string for version ID.
                            fileName, clientConfiguration)) {
        try {
            std::string metadataGZip;
            {
                std::ifstream inFileStream(fileName.c_str(), std::ios::binary);
                if (!inFileStream) {
                    throw std::runtime_error("Failed to open file " + fileName);
                }

                std::stringstream stringStream;
                stringStream << inFileStream.rdbuf();
                metadataGZip = stringStream.str();
            }
            std::string metadataJson = gzip::decompress(metadataGZip.data(),
                                                        metadataGZip.size());
            // Use JMESPath to extract the image set IDs.
            // https://jmespath.org/specification.html
            jsoncons::json doc = jsoncons::json::parse(metadataJson);
            std::string jmesPathExpression = "Study.Series.*.Instances[].*[]";
            jsoncons::json instances = jsoncons::jmespath::search(doc,
                                                                  jmesPathExpression);
            for (auto &instance: instances.array_range()) {
                jmesPathExpression = "DICOM.RescaleSlope";
                std::string rescaleSlope = jsoncons::jmespath::search(instance,
                                                                      jmesPathExpression).to_string();
                jmesPathExpression = "DICOM.RescaleIntercept";
                std::string rescaleIntercept = jsoncons::jmespath::search(instance,
                                                                          jmesPathExpression).to_string();

                jmesPathExpression = "ImageFrames[][]";
                jsoncons::json imageFramesJson = jsoncons::jmespath::search(instance,
                                                                            jmesPathExpression);

                for (auto &imageFrame: imageFramesJson.array_range()) {
                    ImageFrameInfo imageFrameIDs;
                    imageFrameIDs.mImageSetId = imageSetID;
                    imageFrameIDs.mImageFrameId = imageFrame.find(
                            "ID")->value().as_string();
                    imageFrameIDs.mRescaleIntercept = rescaleIntercept;
                    imageFrameIDs.mRescaleSlope = rescaleSlope;
                    imageFrameIDs.MinPixelValue = imageFrame.find(
                            "MinPixelValue")->value().as_string();
                    imageFrameIDs.MaxPixelValue = imageFrame.find(
                            "MaxPixelValue")->value().as_string();

                    jmesPathExpression = "max_by(PixelDataChecksumFromBaseToFullResolution, &Width).Checksum";
                    jsoncons::json checksumJson = jsoncons::jmespath::search(imageFrame,
                                                                             jmesPathExpression);
                    imageFrameIDs.mFullResolutionChecksum = checksumJson.as_integer<uint32_t>();

                    imageFrames.emplace_back(imageFrameIDs);
                }
            }

            result = true;
        }
        catch (const std::exception &e) {
            std::cerr << "getImageFramesForImageSet failed because " << e.what()
                      << std::endl;
        }
    }

    return result;
}

//! Routine which gets a HealthImaging image set's metadata.
/*!
  \param dataStoreID: The HealthImaging data store ID.
  \param imageSetID: The HealthImaging image set ID.
  \param versionID: The HealthImaging image set version ID, ignored if empty.
  \param outputFilePath: The path where the metadata will be stored as gzipped json.
  \param clientConfig: Aws client configuration.
  \\return bool: Function succeeded.
*/
bool AwsDoc::Medical_Imaging::getImageSetMetadata(const Aws::String &dataStoreID,
                                                  const Aws::String &imageSetID,
                                                  const Aws::String &versionID,
                                                  const Aws::String &outputFilePath,
                                                  const Aws::Client::ClientConfiguration &clientConfig) {
    Aws::MedicalImaging::Model::GetImageSetMetadataRequest request;
    request.SetDatastoreId(dataStoreID);
    request.SetImageSetId(imageSetID);
    if (!versionID.empty()) {
        request.SetVersionId(versionID);
    }
    Aws::MedicalImaging::MedicalImagingClient client(clientConfig);
    Aws::MedicalImaging::Model::GetImageSetMetadataOutcome outcome = client.GetImageSetMetadata(
            request);
    if (outcome.IsSuccess()) {
        std::ofstream file(outputFilePath, std::ios::binary);
        auto &metadata = outcome.GetResult().GetImageSetMetadataBlob();
        file << metadata.rdbuf();
    }
    else {
        std::cerr << "Failed to get image set metadata: "
                  << outcome.GetError().GetMessage() << std::endl;
    }

    return outcome.IsSuccess();
}

下載、解碼和驗證影像影格。

bool AwsDoc::Medical_Imaging::downloadDecodeAndCheckImageFrames(
        const Aws::String &dataStoreID,
        const Aws::Vector<ImageFrameInfo> &imageFrames,
        const Aws::String &outDirectory,
        const Aws::Client::ClientConfiguration &clientConfiguration) {

    Aws::Client::ClientConfiguration clientConfiguration1(clientConfiguration);
    clientConfiguration1.executor = Aws::MakeShared<Aws::Utils::Threading::PooledThreadExecutor>(
            "executor", 25);
    Aws::MedicalImaging::MedicalImagingClient medicalImagingClient(
            clientConfiguration1);

    Aws::Utils::Threading::Semaphore semaphore(0, 1);
    std::atomic<size_t> count(imageFrames.size());

    bool result = true;
    for (auto &imageFrame: imageFrames) {
        Aws::MedicalImaging::Model::GetImageFrameRequest getImageFrameRequest;
        getImageFrameRequest.SetDatastoreId(dataStoreID);
        getImageFrameRequest.SetImageSetId(imageFrame.mImageSetId);

        Aws::MedicalImaging::Model::ImageFrameInformation imageFrameInformation;
        imageFrameInformation.SetImageFrameId(imageFrame.mImageFrameId);
        getImageFrameRequest.SetImageFrameInformation(imageFrameInformation);

        auto getImageFrameAsyncLambda = [&semaphore, &result, &count, imageFrame, outDirectory](
                const Aws::MedicalImaging::MedicalImagingClient *client,
                const Aws::MedicalImaging::Model::GetImageFrameRequest &request,
                Aws::MedicalImaging::Model::GetImageFrameOutcome outcome,
                const std::shared_ptr<const Aws::Client::AsyncCallerContext> &context) {

                if (!handleGetImageFrameResult(outcome, outDirectory, imageFrame)) {
                    std::cerr << "Failed to download and convert image frame: "
                              << imageFrame.mImageFrameId << " from image set: "
                              << imageFrame.mImageSetId << std::endl;
                    result = false;
                }

                count--;
                if (count <= 0) {

                    semaphore.ReleaseAll();
                }
        }; // End of 'getImageFrameAsyncLambda' lambda.

        medicalImagingClient.GetImageFrameAsync(getImageFrameRequest,
                                                getImageFrameAsyncLambda);
    }

    if (count > 0) {
        semaphore.WaitOne();
    }

    if (result) {
        std::cout << imageFrames.size() << " image files were downloaded."
                  << std::endl;
    }

    return result;
}

bool AwsDoc::Medical_Imaging::decodeJPHFileAndValidateWithChecksum(
        const Aws::String &jphFile,
        uint32_t crc32Checksum) {
    opj_image_t *outputImage = jphImageToOpjBitmap(jphFile);
    if (!outputImage) {
        return false;
    }

    bool result = true;
    if (!verifyChecksumForImage(outputImage, crc32Checksum)) {
        std::cerr << "The checksum for the image does not match the expected value."
                  << std::endl;
        std::cerr << "File :" << jphFile << std::endl;
        result = false;
    }

    opj_image_destroy(outputImage);

    return result;
}

opj_image *
AwsDoc::Medical_Imaging::jphImageToOpjBitmap(const Aws::String &jphFile) {
    opj_stream_t *inFileStream = nullptr;
    opj_codec_t *decompressorCodec = nullptr;
    opj_image_t *outputImage = nullptr;
    try {
        std::shared_ptr<opj_dparameters> decodeParameters = std::make_shared<opj_dparameters>();
        memset(decodeParameters.get(), 0, sizeof(opj_dparameters));

        opj_set_default_decoder_parameters(decodeParameters.get());

        decodeParameters->decod_format = 1; // JP2 image format.
        decodeParameters->cod_format = 2; // BMP image format.

        std::strncpy(decodeParameters->infile, jphFile.c_str(),
                     OPJ_PATH_LEN);

        inFileStream = opj_stream_create_default_file_stream(
                decodeParameters->infile, true);
        if (!inFileStream) {
            throw std::runtime_error(
                    "Unable to create input file stream for file '" + jphFile + "'.");
        }

        decompressorCodec = opj_create_decompress(OPJ_CODEC_JP2);
        if (!decompressorCodec) {
            throw std::runtime_error("Failed to create decompression codec.");
        }

        int decodeMessageLevel = 1;
        if (!setupCodecLogging(decompressorCodec, &decodeMessageLevel)) {
            std::cerr << "Failed to setup codec logging." << std::endl;
        }

        if (!opj_setup_decoder(decompressorCodec, decodeParameters.get())) {
            throw std::runtime_error("Failed to setup decompression codec.");
        }
        if (!opj_codec_set_threads(decompressorCodec, 4)) {
            throw std::runtime_error("Failed to set decompression codec threads.");
        }

        if (!opj_read_header(inFileStream, decompressorCodec, &outputImage)) {
            throw std::runtime_error("Failed to read header.");
        }

        if (!opj_decode(decompressorCodec, inFileStream,
                        outputImage)) {
            throw std::runtime_error("Failed to decode.");
        }

        if (DEBUGGING) {
            std::cout << "image width : " << outputImage->x1 - outputImage->x0
                      << std::endl;
            std::cout << "image height : " << outputImage->y1 - outputImage->y0
                      << std::endl;
            std::cout << "number of channels: " << outputImage->numcomps
                      << std::endl;
            std::cout << "colorspace : " << outputImage->color_space << std::endl;
        }

    } catch (const std::exception &e) {
        std::cerr << e.what() << std::endl;
        if (outputImage) {
            opj_image_destroy(outputImage);
            outputImage = nullptr;
        }
    }
    if (inFileStream) {
        opj_stream_destroy(inFileStream);
    }
    if (decompressorCodec) {
        opj_destroy_codec(decompressorCodec);
    }

    return outputImage;
}

//! Template function which converts a planar image bitmap to an interleaved image bitmap and
//! then verifies the checksum of the bitmap.
/*!
 * @param image: The OpenJPEG image struct.
 * @param crc32Checksum: The CRC32 checksum.
 * @return  bool: Function succeeded.
 */
template<class myType>
bool verifyChecksumForImageForType(opj_image_t *image, uint32_t crc32Checksum) {
    uint32_t width = image->x1 - image->x0;
    uint32_t height = image->y1 - image->y0;
    uint32_t numOfChannels = image->numcomps;

    // Buffer for interleaved bitmap.
    std::vector<myType> buffer(width * height * numOfChannels);

    // Convert planar bitmap to interleaved bitmap.
    for (uint32_t channel = 0; channel < numOfChannels; channel++) {
        for (uint32_t row = 0; row < height; row++) {
            uint32_t fromRowStart = row / image->comps[channel].dy * width /
                                    image->comps[channel].dx;
            uint32_t toIndex = (row * width) * numOfChannels + channel;

            for (uint32_t col = 0; col < width; col++) {
                uint32_t fromIndex = fromRowStart + col / image->comps[channel].dx;

                buffer[toIndex] = static_cast<myType>(image->comps[channel].data[fromIndex]);

                toIndex += numOfChannels;
            }
        }
    }

    // Verify checksum.
    boost::crc_32_type crc32;
    crc32.process_bytes(reinterpret_cast<char *>(buffer.data()),
                        buffer.size() * sizeof(myType));

    bool result = crc32.checksum() == crc32Checksum;
    if (!result) {
        std::cerr << "verifyChecksumForImage, checksum mismatch, expected - "
                  << crc32Checksum << ", actual - " << crc32.checksum()
                  << std::endl;
    }

    return result;
}

//! Routine which verifies the checksum of an OpenJPEG image struct.
/*!
 * @param image: The OpenJPEG image struct.
 * @param crc32Checksum: The CRC32 checksum.
 * @return  bool: Function succeeded.
 */
bool AwsDoc::Medical_Imaging::verifyChecksumForImage(opj_image_t *image,
                                                     uint32_t crc32Checksum) {
    uint32_t channels = image->numcomps;
    bool result = false;
    if (0 < channels) {
        // Assume the precision is the same for all channels.
        uint32_t precision = image->comps[0].prec;
        bool signedData = image->comps[0].sgnd;
        uint32_t bytes = (precision + 7) / 8;

        if (signedData) {
            switch (bytes) {
                case 1 :
                    result = verifyChecksumForImageForType<int8_t>(image,
                                                                   crc32Checksum);
                    break;
                case 2 :
                    result = verifyChecksumForImageForType<int16_t>(image,
                                                                    crc32Checksum);
                    break;
                case 4 :
                    result = verifyChecksumForImageForType<int32_t>(image,
                                                                    crc32Checksum);
                    break;
                default:
                    std::cerr
                            << "verifyChecksumForImage, unsupported data type, signed bytes - "
                            << bytes << std::endl;
                    break;
            }
        }
        else {
            switch (bytes) {
                case 1 :
                    result = verifyChecksumForImageForType<uint8_t>(image,
                                                                    crc32Checksum);
                    break;
                case 2 :
                    result = verifyChecksumForImageForType<uint16_t>(image,
                                                                     crc32Checksum);
                    break;
                case 4 :
                    result = verifyChecksumForImageForType<uint32_t>(image,
                                                                     crc32Checksum);
                    break;
                default:
                    std::cerr
                            << "verifyChecksumForImage, unsupported data type, unsigned bytes - "
                            << bytes << std::endl;
                    break;
            }
        }

        if (!result) {
            std::cerr << "verifyChecksumForImage, error bytes " << bytes
                      << " signed "
                      << signedData << std::endl;
        }
    }
    else {
        std::cerr << "'verifyChecksumForImage', no channels in the image."
                  << std::endl;
    }
    return result;
}

清除資源。

bool AwsDoc::Medical_Imaging::cleanup(const Aws::String &stackName,
                                      const Aws::String &dataStoreId,
                                      const Aws::Client::ClientConfiguration &clientConfiguration) {
    bool result = true;

    if (!stackName.empty() && askYesNoQuestion(
            "Would you like to delete the stack " + stackName + "? (y/n)")) {
        std::cout << "Deleting the image sets in the stack." << std::endl;
        result &= emptyDatastore(dataStoreId, clientConfiguration);
        printAsterisksLine();
        std::cout << "Deleting the stack." << std::endl;
        result &= deleteStack(stackName, clientConfiguration);
    }
    return result;
}

bool AwsDoc::Medical_Imaging::emptyDatastore(const Aws::String &datastoreID,
                                             const Aws::Client::ClientConfiguration &clientConfiguration) {

    Aws::MedicalImaging::Model::SearchCriteria emptyCriteria;
    Aws::Vector<Aws::String> imageSetIDs;
    bool result = false;
    if (searchImageSets(datastoreID, emptyCriteria, imageSetIDs,
                        clientConfiguration)) {
        result = true;
        for (auto &imageSetID: imageSetIDs) {
            result &= deleteImageSet(datastoreID, imageSetID, clientConfiguration);
        }
    }

    return result;
}

協調步驟 (index.js).

import {
  parseScenarioArgs,
  Scenario,
} from "@aws-doc-sdk-examples/lib/scenario/index.js";
import {
  saveState,
  loadState,
} from "@aws-doc-sdk-examples/lib/scenario/steps-common.js";

import {
  createStack,
  deployStack,
  getAccountId,
  getDatastoreName,
  getStackName,
  outputState,
  waitForStackCreation,
} from "./deploy-steps.js";
import {
  doCopy,
  selectDataset,
  copyDataset,
  outputCopiedObjects,
} from "./dataset-steps.js";
import {
  doImport,
  outputImportJobStatus,
  startDICOMImport,
  waitForImportJobCompletion,
} from "./import-steps.js";
import {
  getManifestFile,
  outputImageSetIds,
  parseManifestFile,
} from "./image-set-steps.js";
import {
  getImageSetMetadata,
  outputImageFrameIds,
} from "./image-frame-steps.js";
import { decodeAndVerifyImages, doVerify } from "./verify-steps.js";
import {
  confirmCleanup,
  deleteImageSets,
  deleteStack,
} from "./clean-up-steps.js";

const context = {};

const scenarios = {
  deploy: new Scenario(
    "Deploy Resources",
    [
      deployStack,
      getStackName,
      getDatastoreName,
      getAccountId,
      createStack,
      waitForStackCreation,
      outputState,
      saveState,
    ],
    context,
  ),
  demo: new Scenario(
    "Run Demo",
    [
      loadState,
      doCopy,
      selectDataset,
      copyDataset,
      outputCopiedObjects,
      doImport,
      startDICOMImport,
      waitForImportJobCompletion,
      outputImportJobStatus,
      getManifestFile,
      parseManifestFile,
      outputImageSetIds,
      getImageSetMetadata,
      outputImageFrameIds,
      doVerify,
      decodeAndVerifyImages,
      saveState,
    ],
    context,
  ),
  destroy: new Scenario(
    "Clean Up Resources",
    [loadState, confirmCleanup, deleteImageSets, deleteStack],
    context,
  ),
};

// Call function if run directly
import { fileURLToPath } from "node:url";
if (process.argv[1] === fileURLToPath(import.meta.url)) {
  parseScenarioArgs(scenarios, {
    name: "Health Imaging Workflow",
    description:
      "Work with DICOM images using an AWS Health Imaging data store.",
    synopsis:
      "node index.js --scenario <deploy | demo | destroy> [-h|--help] [-y|--yes] [-v|--verbose]",
  });
}

部署 資源 (deploy-steps.js).

import fs from "node:fs/promises";
import path from "node:path";

import {
  CloudFormationClient,
  CreateStackCommand,
  DescribeStacksCommand,
} from "@aws-sdk/client-cloudformation";
import { STSClient, GetCallerIdentityCommand } from "@aws-sdk/client-sts";

import {
  ScenarioAction,
  ScenarioInput,
  ScenarioOutput,
} from "@aws-doc-sdk-examples/lib/scenario/index.js";
import { retry } from "@aws-doc-sdk-examples/lib/utils/util-timers.js";

const cfnClient = new CloudFormationClient({});
const stsClient = new STSClient({});

const __dirname = path.dirname(new URL(import.meta.url).pathname);
const cfnTemplatePath = path.join(
  __dirname,
  "../../../../../workflows/healthimaging_image_sets/resources/cfn_template.yaml",
);

export const deployStack = new ScenarioInput(
  "deployStack",
  "Do you want to deploy the CloudFormation stack?",
  { type: "confirm" },
);

export const getStackName = new ScenarioInput(
  "getStackName",
  "Enter a name for the CloudFormation stack:",
  { type: "input", skipWhen: (/** @type {{}} */ state) => !state.deployStack },
);

export const getDatastoreName = new ScenarioInput(
  "getDatastoreName",
  "Enter a name for the HealthImaging datastore:",
  { type: "input", skipWhen: (/** @type {{}} */ state) => !state.deployStack },
);

export const getAccountId = new ScenarioAction(
  "getAccountId",
  async (/** @type {{}} */ state) => {
    const command = new GetCallerIdentityCommand({});
    const response = await stsClient.send(command);
    state.accountId = response.Account;
  },
  {
    skipWhen: (/** @type {{}} */ state) => !state.deployStack,
  },
);

export const createStack = new ScenarioAction(
  "createStack",
  async (/** @type {{}} */ state) => {
    const stackName = state.getStackName;
    const datastoreName = state.getDatastoreName;
    const accountId = state.accountId;

    const command = new CreateStackCommand({
      StackName: stackName,
      TemplateBody: await fs.readFile(cfnTemplatePath, "utf8"),
      Capabilities: ["CAPABILITY_IAM"],
      Parameters: [
        {
          ParameterKey: "datastoreName",
          ParameterValue: datastoreName,
        },
        {
          ParameterKey: "userAccountID",
          ParameterValue: accountId,
        },
      ],
    });

    const response = await cfnClient.send(command);
    state.stackId = response.StackId;
  },
  { skipWhen: (/** @type {{}} */ state) => !state.deployStack },
);

export const waitForStackCreation = new ScenarioAction(
  "waitForStackCreation",
  async (/** @type {{}} */ state) => {
    const command = new DescribeStacksCommand({
      StackName: state.stackId,
    });

    await retry({ intervalInMs: 10000, maxRetries: 60 }, async () => {
      const response = await cfnClient.send(command);
      const stack = response.Stacks?.find(
        (s) => s.StackName === state.getStackName,
      );
      if (!stack || stack.StackStatus === "CREATE_IN_PROGRESS") {
        throw new Error("Stack creation is still in progress");
      }
      if (stack.StackStatus === "CREATE_COMPLETE") {
        state.stackOutputs = stack.Outputs?.reduce((acc, output) => {
          acc[output.OutputKey] = output.OutputValue;
          return acc;
        }, {});
      } else {
        throw new Error(
          `Stack creation failed with status: ${stack.StackStatus}`,
        );
      }
    });
  },
  {
    skipWhen: (/** @type {{}} */ state) => !state.deployStack,
  },
);

export const outputState = new ScenarioOutput(
  "outputState",
  (/** @type {{}} */ state) => {
    /**
     * @type {{ stackOutputs: { DatastoreID: string, BucketName: string, RoleArn: string }}}
     */
    const { stackOutputs } = state;
    return `Stack creation completed. Output values:
Datastore ID: ${stackOutputs?.DatastoreID}
Bucket Name: ${stackOutputs?.BucketName}
Role ARN: ${stackOutputs?.RoleArn}
    `;
  },
  { skipWhen: (/** @type {{}} */ state) => !state.deployStack },
);

複製DICOM檔案 (dataset-steps.js).

import {
  S3Client,
  CopyObjectCommand,
  ListObjectsV2Command,
} from "@aws-sdk/client-s3";

import {
  ScenarioAction,
  ScenarioInput,
  ScenarioOutput,
} from "@aws-doc-sdk-examples/lib/scenario/index.js";

const s3Client = new S3Client({});

const datasetOptions = [
  {
    name: "CT of chest (2 images)",
    value: "00029d25-fb18-4d42-aaa5-a0897d1ac8f7",
  },
  {
    name: "CT of pelvis (57 images)",
    value: "00025d30-ef8f-4135-a35a-d83eff264fc1",
  },
  {
    name: "MRI of head (192 images)",
    value: "0002d261-8a5d-4e63-8e2e-0cbfac87b904",
  },
  {
    name: "MRI of breast (92 images)",
    value: "0002dd07-0b7f-4a68-a655-44461ca34096",
  },
];

/**
 * @typedef {{ stackOutputs: {
 *   BucketName: string,
 *   DatastoreID: string,
 *   doCopy: boolean
 * }}} State
 */

export const selectDataset = new ScenarioInput(
  "selectDataset",
  (state) => {
    if (!state.doCopy) {
      process.exit(0);
    }
    return "Select a DICOM dataset to import:";
  },
  {
    type: "select",
    choices: datasetOptions,
  },
);

export const doCopy = new ScenarioInput(
  "doCopy",
  "Do you want to copy images from the public dataset into your bucket?",
  {
    type: "confirm",
  },
);

export const copyDataset = new ScenarioAction(
  "copyDataset",
  async (/** @type { State } */ state) => {
    const inputBucket = state.stackOutputs.BucketName;
    const inputPrefix = "input/";
    const selectedDatasetId = state.selectDataset;

    const sourceBucket = "idc-open-data";
    const sourcePrefix = `${selectedDatasetId}`;

    const listObjectsCommand = new ListObjectsV2Command({
      Bucket: sourceBucket,
      Prefix: sourcePrefix,
    });

    const objects = await s3Client.send(listObjectsCommand);

    const copyPromises = objects.Contents.map((object) => {
      const sourceKey = object.Key;
      const destinationKey = `${inputPrefix}${sourceKey
        .split("/")
        .slice(1)
        .join("/")}`;

      const copyCommand = new CopyObjectCommand({
        Bucket: inputBucket,
        CopySource: `/${sourceBucket}/${sourceKey}`,
        Key: destinationKey,
      });

      return s3Client.send(copyCommand);
    });

    const results = await Promise.all(copyPromises);
    state.copiedObjects = results.length;
  },
);

export const outputCopiedObjects = new ScenarioOutput(
  "outputCopiedObjects",
  (state) => `${state.copiedObjects} DICOM files were copied.`,
);

開始匯入資料存放區 (import-steps.js).

import {
  MedicalImagingClient,
  StartDICOMImportJobCommand,
  GetDICOMImportJobCommand,
} from "@aws-sdk/client-medical-imaging";

import {
  ScenarioAction,
  ScenarioOutput,
  ScenarioInput,
} from "@aws-doc-sdk-examples/lib/scenario/index.js";
import { retry } from "@aws-doc-sdk-examples/lib/utils/util-timers.js";

/**
 * @typedef {{ stackOutputs: {
 *   BucketName: string,
 *   DatastoreID: string,
 *   RoleArn: string
 * }}} State
 */

export const doImport = new ScenarioInput(
  "doImport",
  "Do you want to import DICOM images into your datastore?",
  {
    type: "confirm",
    default: true,
  },
);

export const startDICOMImport = new ScenarioAction(
  "startDICOMImport",
  async (/** @type {State} */ state) => {
    if (!state.doImport) {
      process.exit(0);
    }
    const medicalImagingClient = new MedicalImagingClient({});
    const inputS3Uri = `s3://${state.stackOutputs.BucketName}/input/`;
    const outputS3Uri = `s3://${state.stackOutputs.BucketName}/output/`;

    const command = new StartDICOMImportJobCommand({
      dataAccessRoleArn: state.stackOutputs.RoleArn,
      datastoreId: state.stackOutputs.DatastoreID,
      inputS3Uri,
      outputS3Uri,
    });

    const response = await medicalImagingClient.send(command);
    state.importJobId = response.jobId;
  },
);

export const waitForImportJobCompletion = new ScenarioAction(
  "waitForImportJobCompletion",
  async (/** @type {State} */ state) => {
    const medicalImagingClient = new MedicalImagingClient({});
    const command = new GetDICOMImportJobCommand({
      datastoreId: state.stackOutputs.DatastoreID,
      jobId: state.importJobId,
    });

    await retry({ intervalInMs: 10000, maxRetries: 60 }, async () => {
      const response = await medicalImagingClient.send(command);
      const jobStatus = response.jobProperties?.jobStatus;
      if (!jobStatus || jobStatus === "IN_PROGRESS") {
        throw new Error("Import job is still in progress");
      }
      if (jobStatus === "COMPLETED") {
        state.importJobOutputS3Uri = response.jobProperties.outputS3Uri;
      } else {
        throw new Error(`Import job failed with status: ${jobStatus}`);
      }
    });
  },
);

export const outputImportJobStatus = new ScenarioOutput(
  "outputImportJobStatus",
  (state) =>
    `DICOM import job completed. Output location: ${state.importJobOutputS3Uri}`,
);

取得映像集 IDs(image-set-steps.js - ).

import { S3Client, GetObjectCommand } from "@aws-sdk/client-s3";

import {
  ScenarioAction,
  ScenarioOutput,
} from "@aws-doc-sdk-examples/lib/scenario/index.js";

/**
 * @typedef {{ stackOutputs: {
 *   BucketName: string,
 *   DatastoreID: string,
 *   RoleArn: string
 * }, importJobId: string,
 * importJobOutputS3Uri: string,
 * imageSetIds: string[],
 * manifestContent: { jobSummary: { imageSetsSummary: { imageSetId: string }[] } }
 * }} State
 */

const s3Client = new S3Client({});

export const getManifestFile = new ScenarioAction(
  "getManifestFile",
  async (/** @type {State} */ state) => {
    const bucket = state.stackOutputs.BucketName;
    const prefix = `output/${state.stackOutputs.DatastoreID}-DicomImport-${state.importJobId}/`;
    const key = `${prefix}job-output-manifest.json`;

    const command = new GetObjectCommand({
      Bucket: bucket,
      Key: key,
    });

    const response = await s3Client.send(command);
    const manifestContent = await response.Body.transformToString();
    state.manifestContent = JSON.parse(manifestContent);
  },
);

export const parseManifestFile = new ScenarioAction(
  "parseManifestFile",
  (/** @type {State} */ state) => {
    const imageSetIds =
      state.manifestContent.jobSummary.imageSetsSummary.reduce((ids, next) => {
        return Object.assign({}, ids, {
          [next.imageSetId]: next.imageSetId,
        });
      }, {});
    state.imageSetIds = Object.keys(imageSetIds);
  },
);

export const outputImageSetIds = new ScenarioOutput(
  "outputImageSetIds",
  (/** @type {State} */ state) =>
    `The image sets created by this import job are: \n${state.imageSetIds
      .map((id) => `Image set: ${id}`)
      .join("\n")}`,
);

取得影像影格 IDs(image-frame-steps.js).

import {
  MedicalImagingClient,
  GetImageSetMetadataCommand,
} from "@aws-sdk/client-medical-imaging";
import { gunzip } from "node:zlib";
import { promisify } from "node:util";

import {
  ScenarioAction,
  ScenarioOutput,
} from "@aws-doc-sdk-examples/lib/scenario/index.js";

const gunzipAsync = promisify(gunzip);

/**
 * @typedef {Object} DICOMValueRepresentation
 * @property {string} name
 * @property {string} type
 * @property {string} value
 */

/**
 * @typedef {Object} ImageFrameInformation
 * @property {string} ID
 * @property {Array<{ Checksum: number, Height: number, Width: number }>} PixelDataChecksumFromBaseToFullResolution
 * @property {number} MinPixelValue
 * @property {number} MaxPixelValue
 * @property {number} FrameSizeInBytes
 */

/**
 * @typedef {Object} DICOMMetadata
 * @property {Object} DICOM
 * @property {DICOMValueRepresentation[]} DICOMVRs
 * @property {ImageFrameInformation[]} ImageFrames
 */

/**
 * @typedef {Object} Series
 * @property {{ [key: string]: DICOMMetadata }} Instances
 */

/**
 * @typedef {Object} Study
 * @property {Object} DICOM
 * @property {Series[]} Series
 */

/**
 * @typedef {Object} Patient
 * @property {Object} DICOM
 */

/**
 * @typedef {{
 *  SchemaVersion: string,
 *  DatastoreID: string,
 *  ImageSetID: string,
 *  Patient: Patient,
 *  Study: Study
 * }} ImageSetMetadata
 */

/**
 * @typedef {{ stackOutputs: {
 *   BucketName: string,
 *   DatastoreID: string,
 *   RoleArn: string
 * }, imageSetIds: string[] }} State
 */

const medicalImagingClient = new MedicalImagingClient({});

export const getImageSetMetadata = new ScenarioAction(
  "getImageSetMetadata",
  async (/** @type {State} */ state) => {
    const outputMetadata = [];

    for (const imageSetId of state.imageSetIds) {
      const command = new GetImageSetMetadataCommand({
        datastoreId: state.stackOutputs.DatastoreID,
        imageSetId,
      });

      const response = await medicalImagingClient.send(command);
      const compressedMetadataBlob =
        await response.imageSetMetadataBlob.transformToByteArray();
      const decompressedMetadata = await gunzipAsync(compressedMetadataBlob);
      const imageSetMetadata = JSON.parse(decompressedMetadata.toString());

      outputMetadata.push(imageSetMetadata);
    }

    state.imageSetMetadata = outputMetadata;
  },
);

export const outputImageFrameIds = new ScenarioOutput(
  "outputImageFrameIds",
  (/** @type {State & { imageSetMetadata: ImageSetMetadata[] }} */ state) => {
    let output = "";

    for (const metadata of state.imageSetMetadata) {
      const imageSetId = metadata.ImageSetID;
      /** @type {DICOMMetadata[]} */
      const instances = Object.values(metadata.Study.Series).flatMap(
        (series) => {
          return Object.values(series.Instances);
        },
      );
      const imageFrameIds = instances.flatMap((instance) =>
        instance.ImageFrames.map((frame) => frame.ID),
      );

      output += `Image set ID: ${imageSetId}\nImage frame IDs:\n${imageFrameIds.join(
        "\n",
      )}\n\n`;
    }

    return output;
  },
);

驗證影像影格 (verify-steps.js)。 AWS HealthImaging 像素資料驗證程式庫已用於驗證。

import { spawn } from "node:child_process";

import {
  ScenarioAction,
  ScenarioInput,
} from "@aws-doc-sdk-examples/lib/scenario/index.js";

/**
 * @typedef {Object} DICOMValueRepresentation
 * @property {string} name
 * @property {string} type
 * @property {string} value
 */

/**
 * @typedef {Object} ImageFrameInformation
 * @property {string} ID
 * @property {Array<{ Checksum: number, Height: number, Width: number }>} PixelDataChecksumFromBaseToFullResolution
 * @property {number} MinPixelValue
 * @property {number} MaxPixelValue
 * @property {number} FrameSizeInBytes
 */

/**
 * @typedef {Object} DICOMMetadata
 * @property {Object} DICOM
 * @property {DICOMValueRepresentation[]} DICOMVRs
 * @property {ImageFrameInformation[]} ImageFrames
 */

/**
 * @typedef {Object} Series
 * @property {{ [key: string]: DICOMMetadata }} Instances
 */

/**
 * @typedef {Object} Study
 * @property {Object} DICOM
 * @property {Series[]} Series
 */

/**
 * @typedef {Object} Patient
 * @property {Object} DICOM
 */

/**
 * @typedef {{
 *  SchemaVersion: string,
 *  DatastoreID: string,
 *  ImageSetID: string,
 *  Patient: Patient,
 *  Study: Study
 * }} ImageSetMetadata
 */

/**
 * @typedef {{ stackOutputs: {
 *   BucketName: string,
 *   DatastoreID: string,
 *   RoleArn: string
 * }, imageSetMetadata: ImageSetMetadata[] }} State
 */

export const doVerify = new ScenarioInput(
  "doVerify",
  "Do you want to verify the imported images?",
  {
    type: "confirm",
    default: true,
  },
);

export const decodeAndVerifyImages = new ScenarioAction(
  "decodeAndVerifyImages",
  async (/** @type {State} */ state) => {
    if (!state.doVerify) {
      process.exit(0);
    }
    const verificationTool = "./pixel-data-verification/index.js";

    for (const metadata of state.imageSetMetadata) {
      const datastoreId = state.stackOutputs.DatastoreID;
      const imageSetId = metadata.ImageSetID;

      for (const [seriesInstanceUid, series] of Object.entries(
        metadata.Study.Series,
      )) {
        for (const [sopInstanceUid, _] of Object.entries(series.Instances)) {
          console.log(
            `Verifying image set ${imageSetId} with series ${seriesInstanceUid} and sop ${sopInstanceUid}`,
          );
          const child = spawn(
            "node",
            [
              verificationTool,
              datastoreId,
              imageSetId,
              seriesInstanceUid,
              sopInstanceUid,
            ],
            { stdio: "inherit" },
          );

          await new Promise((resolve, reject) => {
            child.on("exit", (code) => {
              if (code === 0) {
                resolve();
              } else {
                reject(
                  new Error(
                    `Verification tool exited with code ${code} for image set ${imageSetId}`,
                  ),
                );
              }
            });
          });
        }
      }
    }
  },
);

銷毀資源 (clean-up-steps.js).

import {
  CloudFormationClient,
  DeleteStackCommand,
} from "@aws-sdk/client-cloudformation";
import {
  MedicalImagingClient,
  DeleteImageSetCommand,
} from "@aws-sdk/client-medical-imaging";

import {
  ScenarioAction,
  ScenarioInput,
} from "@aws-doc-sdk-examples/lib/scenario/index.js";

/**
 * @typedef {Object} DICOMValueRepresentation
 * @property {string} name
 * @property {string} type
 * @property {string} value
 */

/**
 * @typedef {Object} ImageFrameInformation
 * @property {string} ID
 * @property {Array<{ Checksum: number, Height: number, Width: number }>} PixelDataChecksumFromBaseToFullResolution
 * @property {number} MinPixelValue
 * @property {number} MaxPixelValue
 * @property {number} FrameSizeInBytes
 */

/**
 * @typedef {Object} DICOMMetadata
 * @property {Object} DICOM
 * @property {DICOMValueRepresentation[]} DICOMVRs
 * @property {ImageFrameInformation[]} ImageFrames
 */

/**
 * @typedef {Object} Series
 * @property {{ [key: string]: DICOMMetadata }} Instances
 */

/**
 * @typedef {Object} Study
 * @property {Object} DICOM
 * @property {Series[]} Series
 */

/**
 * @typedef {Object} Patient
 * @property {Object} DICOM
 */

/**
 * @typedef {{
 *  SchemaVersion: string,
 *  DatastoreID: string,
 *  ImageSetID: string,
 *  Patient: Patient,
 *  Study: Study
 * }} ImageSetMetadata
 */

/**
 * @typedef {{ stackOutputs: {
 *   BucketName: string,
 *   DatastoreID: string,
 *   RoleArn: string
 * }, imageSetMetadata: ImageSetMetadata[] }} State
 */

const cfnClient = new CloudFormationClient({});
const medicalImagingClient = new MedicalImagingClient({});

export const confirmCleanup = new ScenarioInput(
  "confirmCleanup",
  "Do you want to delete the created resources?",
  { type: "confirm" },
);

export const deleteImageSets = new ScenarioAction(
  "deleteImageSets",
  async (/** @type {State} */ state) => {
    const datastoreId = state.stackOutputs.DatastoreID;

    for (const metadata of state.imageSetMetadata) {
      const command = new DeleteImageSetCommand({
        datastoreId,
        imageSetId: metadata.ImageSetID,
      });

      try {
        await medicalImagingClient.send(command);
        console.log(`Successfully deleted image set ${metadata.ImageSetID}`);
      } catch (e) {
        if (e instanceof Error) {
          if (e.name === "ConflictException") {
            console.log(`Image set ${metadata.ImageSetID} already deleted`);
          }
        }
      }
    }
  },
  {
    skipWhen: (/** @type {{}} */ state) => !state.confirmCleanup,
  },
);

export const deleteStack = new ScenarioAction(
  "deleteStack",
  async (/** @type {State} */ state) => {
    const stackName = state.getStackName;

    const command = new DeleteStackCommand({
      StackName: stackName,
    });

    await cfnClient.send(command);
    console.log(`Stack ${stackName} deletion initiated`);
  },
  {
    skipWhen: (/** @type {{}} */ state) => !state.confirmCleanup,
  },
);

使用必要的資源建立 AWS CloudFormation 堆疊。

    def deploy(self):
        """
        Deploys prerequisite resources used by the scenario. The resources are
        defined in the associated `setup.yaml` AWS CloudFormation script and are deployed
        as a CloudFormation stack, so they can be easily managed and destroyed.
        """

        print("\t\tLet's deploy the stack for resource creation.")
        stack_name = q.ask("\t\tEnter a name for the stack: ", q.non_empty)

        data_store_name = q.ask(
            "\t\tEnter a name for the Health Imaging Data Store: ", q.non_empty
        )

        account_id = boto3.client("sts").get_caller_identity()["Account"]

        with open(
            "../../../../workflows/healthimaging_image_sets/resources/cfn_template.yaml"
        ) as setup_file:
            setup_template = setup_file.read()
        print(f"\t\tCreating {stack_name}.")
        stack = self.cf_resource.create_stack(
            StackName=stack_name,
            TemplateBody=setup_template,
            Capabilities=["CAPABILITY_NAMED_IAM"],
            Parameters=[
                {
                    "ParameterKey": "datastoreName",
                    "ParameterValue": data_store_name,
                },
                {
                    "ParameterKey": "userAccountID",
                    "ParameterValue": account_id,
                },
            ],
        )
        print("\t\tWaiting for stack to deploy. This typically takes a minute or two.")
        waiter = self.cf_resource.meta.client.get_waiter("stack_create_complete")
        waiter.wait(StackName=stack.name)
        stack.load()
        print(f"\t\tStack status: {stack.stack_status}")

        outputs_dictionary = {
            output["OutputKey"]: output["OutputValue"] for output in stack.outputs
        }
        self.input_bucket_name = outputs_dictionary["BucketName"]
        self.output_bucket_name = outputs_dictionary["BucketName"]
        self.role_arn = outputs_dictionary["RoleArn"]
        self.data_store_id = outputs_dictionary["DatastoreID"]
        return stack

將DICOM檔案複製到 Amazon S3 匯入儲存貯體。

    def copy_single_object(self, key, source_bucket, target_bucket, target_directory):
        """
        Copies a single object from a source to a target bucket.

        :param key: The key of the object to copy.
        :param source_bucket: The source bucket for the copy.
        :param target_bucket: The target bucket for the copy.
        :param target_directory: The target directory for the copy.
        """
        new_key = target_directory + "/" + key
        copy_source = {"Bucket": source_bucket, "Key": key}
        self.s3_client.copy_object(
            CopySource=copy_source, Bucket=target_bucket, Key=new_key
        )
        print(f"\n\t\tCopying {key}.")

    def copy_images(
        self, source_bucket, source_directory, target_bucket, target_directory
    ):
        """
        Copies the images from the source to the target bucket using multiple threads.

        :param source_bucket: The source bucket for the images.
        :param source_directory: Directory within the source bucket.
        :param target_bucket: The target bucket for the images.
        :param target_directory: Directory within the target bucket.
        """

        # Get list of all objects in source bucket.
        list_response = self.s3_client.list_objects_v2(
            Bucket=source_bucket, Prefix=source_directory
        )
        objs = list_response["Contents"]
        keys = [obj["Key"] for obj in objs]

        # Copy the objects in the bucket.
        for key in keys:
            self.copy_single_object(key, source_bucket, target_bucket, target_directory)

        print("\t\tDone copying all objects.")

將DICOM檔案匯入 Amazon S3 資料存放區。

class MedicalImagingWrapper:
    """Encapsulates AWS HealthImaging functionality."""

    def __init__(self, medical_imaging_client, s3_client):
        """
        :param medical_imaging_client: A Boto3 Amazon MedicalImaging client.
        :param s3_client: A Boto3 S3 client.
        """
        self.medical_imaging_client = medical_imaging_client
        self.s3_client = s3_client

    @classmethod
    def from_client(cls):
        medical_imaging_client = boto3.client("medical-imaging")
        s3_client = boto3.client("s3")
        return cls(medical_imaging_client, s3_client)


    def start_dicom_import_job(
        self,
        data_store_id,
        input_bucket_name,
        input_directory,
        output_bucket_name,
        output_directory,
        role_arn,
    ):
        """
        Routine which starts a HealthImaging import job.

        :param data_store_id: The HealthImaging data store ID.
        :param input_bucket_name: The name of the Amazon S3 bucket containing the DICOM files.
        :param input_directory: The directory in the S3 bucket containing the DICOM files.
        :param output_bucket_name: The name of the S3 bucket for the output.
        :param output_directory: The directory in the S3 bucket to store the output.
        :param role_arn: The ARN of the IAM role with permissions for the import.
        :return: The job ID of the import.
        """

        input_uri = f"s3://{input_bucket_name}/{input_directory}/"
        output_uri = f"s3://{output_bucket_name}/{output_directory}/"
        try:
            job = self.medical_imaging_client.start_dicom_import_job(
                jobName="examplejob",
                datastoreId=data_store_id,
                dataAccessRoleArn=role_arn,
                inputS3Uri=input_uri,
                outputS3Uri=output_uri,
            )
        except ClientError as err:
            logger.error(
                "Couldn't start DICOM import job. Here's why: %s: %s",
                err.response["Error"]["Code"],
                err.response["Error"]["Message"],
            )
            raise
        else:
            return job["jobId"]

取得DICOM匯入任務建立的影像集。

class MedicalImagingWrapper:
    """Encapsulates AWS HealthImaging functionality."""

    def __init__(self, medical_imaging_client, s3_client):
        """
        :param medical_imaging_client: A Boto3 Amazon MedicalImaging client.
        :param s3_client: A Boto3 S3 client.
        """
        self.medical_imaging_client = medical_imaging_client
        self.s3_client = s3_client

    @classmethod
    def from_client(cls):
        medical_imaging_client = boto3.client("medical-imaging")
        s3_client = boto3.client("s3")
        return cls(medical_imaging_client, s3_client)


    def get_image_sets_for_dicom_import_job(self, datastore_id, import_job_id):
        """
        Retrieves the image sets created for an import job.

        :param datastore_id: The HealthImaging data store ID
        :param import_job_id: The import job ID
        :return: List of image set IDs
        """

        import_job = self.medical_imaging_client.get_dicom_import_job(
            datastoreId=datastore_id, jobId=import_job_id
        )

        output_uri = import_job["jobProperties"]["outputS3Uri"]

        bucket = output_uri.split("/")[2]
        key = "/".join(output_uri.split("/")[3:])

        # Try to get the manifest.
        retries = 3
        while retries > 0:
            try:
                obj = self.s3_client.get_object(
                    Bucket=bucket, Key=key + "job-output-manifest.json"
                )
                body = obj["Body"]
                break
            except ClientError as error:
                retries = retries - 1
                time.sleep(3)
        try:
            data = json.load(body)
            expression = jmespath.compile("jobSummary.imageSetsSummary[].imageSetId")
            image_sets = expression.search(data)
        except json.decoder.JSONDecodeError as error:
            image_sets = import_job["jobProperties"]

        return image_sets


    def get_image_set(self, datastore_id, image_set_id, version_id=None):
        """
        Get the properties of an image set.

        :param datastore_id: The ID of the data store.
        :param image_set_id: The ID of the image set.
        :param version_id: The optional version of the image set.
        :return: The image set properties.
        """
        try:
            if version_id:
                image_set = self.medical_imaging_client.get_image_set(
                    imageSetId=image_set_id,
                    datastoreId=datastore_id,
                    versionId=version_id,
                )
            else:
                image_set = self.medical_imaging_client.get_image_set(
                    imageSetId=image_set_id, datastoreId=datastore_id
                )
        except ClientError as err:
            logger.error(
                "Couldn't get image set. Here's why: %s: %s",
                err.response["Error"]["Code"],
                err.response["Error"]["Message"],
            )
            raise
        else:
            return image_set

取得影像集的影像影格資訊。

class MedicalImagingWrapper:
    """Encapsulates AWS HealthImaging functionality."""

    def __init__(self, medical_imaging_client, s3_client):
        """
        :param medical_imaging_client: A Boto3 Amazon MedicalImaging client.
        :param s3_client: A Boto3 S3 client.
        """
        self.medical_imaging_client = medical_imaging_client
        self.s3_client = s3_client

    @classmethod
    def from_client(cls):
        medical_imaging_client = boto3.client("medical-imaging")
        s3_client = boto3.client("s3")
        return cls(medical_imaging_client, s3_client)


    def get_image_frames_for_image_set(self, datastore_id, image_set_id, out_directory):
        """
        Get the image frames for an image set.

        :param datastore_id: The ID of the data store.
        :param image_set_id: The ID of the image set.
        :param out_directory: The directory to save the file.
        :return: The image frames.
        """
        image_frames = []
        file_name = os.path.join(out_directory, f"{image_set_id}_metadata.json.gzip")
        file_name = file_name.replace("/", "\\\\")
        self.get_image_set_metadata(file_name, datastore_id, image_set_id)
        try:
            with gzip.open(file_name, "rb") as f_in:
                doc = json.load(f_in)
            instances = jmespath.search("Study.Series.*.Instances[].*[]", doc)
            for instance in instances:
                rescale_slope = jmespath.search("DICOM.RescaleSlope", instance)
                rescale_intercept = jmespath.search("DICOM.RescaleIntercept", instance)
                image_frames_json = jmespath.search("ImageFrames[][]", instance)
                for image_frame in image_frames_json:
                    checksum_json = jmespath.search(
                        "max_by(PixelDataChecksumFromBaseToFullResolution, &Width)",
                        image_frame,
                    )
                    image_frame_info = {
                        "imageSetId": image_set_id,
                        "imageFrameId": image_frame["ID"],
                        "rescaleIntercept": rescale_intercept,
                        "rescaleSlope": rescale_slope,
                        "minPixelValue": image_frame["MinPixelValue"],
                        "maxPixelValue": image_frame["MaxPixelValue"],
                        "fullResolutionChecksum": checksum_json["Checksum"],
                    }
                    image_frames.append(image_frame_info)
            return image_frames
        except TypeError:
            return {}
        except ClientError as err:
            logger.error(
                "Couldn't get image frames for image set. Here's why: %s: %s",
                err.response["Error"]["Code"],
                err.response["Error"]["Message"],
            )
            raise
        return image_frames


    def get_image_set_metadata(
        self, metadata_file, datastore_id, image_set_id, version_id=None
    ):
        """
        Get the metadata of an image set.

        :param metadata_file: The file to store the JSON gzipped metadata.
        :param datastore_id: The ID of the data store.
        :param image_set_id: The ID of the image set.
        :param version_id: The version of the image set.
        """

        try:
            if version_id:
                image_set_metadata = self.medical_imaging_client.get_image_set_metadata(
                    imageSetId=image_set_id,
                    datastoreId=datastore_id,
                    versionId=version_id,
                )
            else:
                image_set_metadata = self.medical_imaging_client.get_image_set_metadata(
                    imageSetId=image_set_id, datastoreId=datastore_id
                )
            with open(metadata_file, "wb") as f:
                for chunk in image_set_metadata["imageSetMetadataBlob"].iter_chunks():
                    if chunk:
                        f.write(chunk)

        except ClientError as err:
            logger.error(
                "Couldn't get image metadata. Here's why: %s: %s",
                err.response["Error"]["Code"],
                err.response["Error"]["Message"],
            )
            raise

下載、解碼和驗證影像影格。

class MedicalImagingWrapper:
    """Encapsulates AWS HealthImaging functionality."""

    def __init__(self, medical_imaging_client, s3_client):
        """
        :param medical_imaging_client: A Boto3 Amazon MedicalImaging client.
        :param s3_client: A Boto3 S3 client.
        """
        self.medical_imaging_client = medical_imaging_client
        self.s3_client = s3_client

    @classmethod
    def from_client(cls):
        medical_imaging_client = boto3.client("medical-imaging")
        s3_client = boto3.client("s3")
        return cls(medical_imaging_client, s3_client)


    def get_pixel_data(
        self, file_path_to_write, datastore_id, image_set_id, image_frame_id
    ):
        """
        Get an image frame's pixel data.

        :param file_path_to_write: The path to write the image frame's HTJ2K encoded pixel data.
        :param datastore_id: The ID of the data store.
        :param image_set_id: The ID of the image set.
        :param image_frame_id: The ID of the image frame.
        """
        try:
            image_frame = self.medical_imaging_client.get_image_frame(
                datastoreId=datastore_id,
                imageSetId=image_set_id,
                imageFrameInformation={"imageFrameId": image_frame_id},
            )
            with open(file_path_to_write, "wb") as f:
                for chunk in image_frame["imageFrameBlob"].iter_chunks():
                    f.write(chunk)
        except ClientError as err:
            logger.error(
                "Couldn't get image frame. Here's why: %s: %s",
                err.response["Error"]["Code"],
                err.response["Error"]["Message"],
            )
            raise


    def download_decode_and_check_image_frames(
        self, data_store_id, image_frames, out_directory
    ):
        """
        Downloads image frames, decodes them, and uses the checksum to validate
        the decoded images.

        :param data_store_id: The HealthImaging data store ID.
        :param image_frames: A list of dicts containing image frame information.
        :param out_directory: A directory for the downloaded images.
        :return: True if the function succeeded; otherwise, False.
        """
        total_result = True
        for image_frame in image_frames:
            image_file_path = f"{out_directory}/image_{image_frame['imageFrameId']}.jph"
            self.get_pixel_data(
                image_file_path,
                data_store_id,
                image_frame["imageSetId"],
                image_frame["imageFrameId"],
            )

            image_array = self.jph_image_to_opj_bitmap(image_file_path)
            crc32_checksum = image_frame["fullResolutionChecksum"]
            # Verify checksum.
            crc32_calculated = zlib.crc32(image_array)
            image_result = crc32_checksum == crc32_calculated
            print(
                f"\t\tImage checksum verified for {image_frame['imageFrameId']}: {image_result }"
            )
            total_result = total_result and image_result
        return total_result

    @staticmethod
    def jph_image_to_opj_bitmap(jph_file):
        """
        Decode the image to a bitmap using an OPENJPEG library.
        :param jph_file: The file to decode.
        :return: The decoded bitmap as an array.
        """
        # Use format 2 for the JPH file.
        params = openjpeg.utils.get_parameters(jph_file, 2)
        print(f"\n\t\tImage parameters for {jph_file}: \n\t\t{params}")

        image_array = openjpeg.utils.decode(jph_file, 2)

        return image_array

清除資源。

    def destroy(self, stack):
        """
        Destroys the resources managed by the CloudFormation stack, and the CloudFormation
        stack itself.

        :param stack: The CloudFormation stack that manages the example resources.
        """

        print(f"\t\tCleaning up resources and {stack.name}.")
        data_store_id = None
        for oput in stack.outputs:
            if oput["OutputKey"] == "DatastoreID":
                data_store_id = oput["OutputValue"]
        if data_store_id is not None:
            print(f"\t\tDeleting image sets in data store {data_store_id}.")
            image_sets = self.medical_imaging_wrapper.search_image_sets(
                data_store_id, {}
            )
            image_set_ids = [image_set["imageSetId"] for image_set in image_sets]

            for image_set_id in image_set_ids:
                self.medical_imaging_wrapper.delete_image_set(
                    data_store_id, image_set_id
                )
                print(f"\t\tDeleted image set with id : {image_set_id}")

        print(f"\t\tDeleting {stack.name}.")
        stack.delete()
        print("\t\tWaiting for stack removal. This may take a few minutes.")
        waiter = self.cf_resource.meta.client.get_waiter("stack_delete_complete")
        waiter.wait(StackName=stack.name)
        print("\t\tStack delete complete.")




class MedicalImagingWrapper:
    """Encapsulates AWS HealthImaging functionality."""

    def __init__(self, medical_imaging_client, s3_client):
        """
        :param medical_imaging_client: A Boto3 Amazon MedicalImaging client.
        :param s3_client: A Boto3 S3 client.
        """
        self.medical_imaging_client = medical_imaging_client
        self.s3_client = s3_client

    @classmethod
    def from_client(cls):
        medical_imaging_client = boto3.client("medical-imaging")
        s3_client = boto3.client("s3")
        return cls(medical_imaging_client, s3_client)


    def search_image_sets(self, datastore_id, search_filter):
        """
        Search for image sets.

        :param datastore_id: The ID of the data store.
        :param search_filter: The search filter.
            For example: {"filters" : [{ "operator": "EQUAL", "values": [{"DICOMPatientId": "3524578"}]}]}.
        :return: The list of image sets.
        """
        try:
            paginator = self.medical_imaging_client.get_paginator("search_image_sets")
            page_iterator = paginator.paginate(
                datastoreId=datastore_id, searchCriteria=search_filter
            )
            metadata_summaries = []
            for page in page_iterator:
                metadata_summaries.extend(page["imageSetsMetadataSummaries"])
        except ClientError as err:
            logger.error(
                "Couldn't search image sets. Here's why: %s: %s",
                err.response["Error"]["Code"],
                err.response["Error"]["Message"],
            )
            raise
        else:
            return metadata_summaries


    def delete_image_set(self, datastore_id, image_set_id):
        """
        Delete an image set.

        :param datastore_id: The ID of the data store.
        :param image_set_id: The ID of the image set.
        """
        try:
            delete_results = self.medical_imaging_client.delete_image_set(
                imageSetId=image_set_id, datastoreId=datastore_id
            )
        except ClientError as err:
            logger.error(
                "Couldn't delete image set. Here's why: %s: %s",
                err.response["Error"]["Code"],
                err.response["Error"]["Message"],
            )
            raise