12.25. Registration Pipeline¶
Registration pipeline takes a data folder and serveral registration parameters as input. Input data can be either in DICOM or nifti format. Pipeline uses image-registration operator for registering input datsets to an input target data.
- NV_REG_TARGET: Registration target name. (Folder name if the input is in Dicom format, image name if input is in nifti format)
- NV_REG_SEQUENCE: Sequence of registration operations, separated by “::”.
- NV_OUT_SEQUENCE: Sequence of registration data in case of multi channel output, separated by “::”
- NV_REG_INPUT: Input file format. Supported types: dicom, nifti
- NV_REG_OUTPUT: Output file format. Supported types: nifti_series, nifti
- NV_REG_TRANSFORM: Registration Transformation, supported types: affine
- NV_REG_METRIC: Registration Metric. Supported types: mattes_mutual
- NV_REG_MM_BINS: Mattes Mutual information metric bin size, typical range 10-60
- NV_REG_MM_SAMPLING_PERCENTAGE: Mattes Mutual Sampling percentage, typical value .01 to .5
- NV_REG_OPTIMIZER: Optimizer, Supported types: reg_step_grad_desc
- NV_REG_RSGD_ITERATIONS: Iterations of regular step gradient descent optimzer
- NV_REG_RSGD_MAXSTEP: Regular step gradient descent optimzer learning rate, typical range 1.0 to 5.0
- NV_REG_RSGD_MINSTEP: Regular step gradient descent optimzer minimum step 0.001
- NV_REG_RSGD_RELAXATION: Regular step gradient descent optimzer relaxation factor
- NV_REG_INTERPOLATOR: Interpolator in Registration. Supported types: linear
12.25.3. Execution Steps on Clara Platform¶
The following are the execution steps for registration pipeline:
Platform Installation: Ensure that platform is installed successfully and Clara CLI is installed correctly.
Required Operator: Following operator must be installed. If the operator is not present in the system, either build it locally or pull it from https://ngc.nvidia.com/containers.
Update the pipeline definition with appropriate container image and tag information. Ensure that container image and tag are present in the system before executing the pipeline.
Execute following in sequence using Clara cli:
clara create pipeline -p <path to registration pipeline> clara create jobs -n <name> -p <pipeline ID> -f <path to input data folder> clara start job -j <JOB ID>
Once job is complete, check payloads for the registered output