Digital volume correlation in Dragonfly with SPAM DVC
Prepare reference and deformed CT volumes, choose a DVC workflow, and inspect displacement and strain fields using the SPAM DVC plugin in Dragonfly.
Digital volume correlation (DVC) estimates internal displacement from the grayscale texture of two 3D images of the same specimen. The Prototype Apps SPAM DVC panel runs the SPAM computation in WSL2 and returns displacement and optional strain Channels to Dragonfly.
Start with suitable CT data
Use a reference scan and a deformed scan of the same specimen, with comparable voxel spacing and sufficient stable 3D texture. Pores, grains or other internal features provide information for correlation. Weak texture or changing image appearance makes the result harder to interpret.
Remove overall translation and rotation with rigid alignment before DVC. Do not use deformable registration for this preparation: it can absorb the deformation you intend to measure. Cropping to a common size is not registration.
Prepare the SPAM environment
Enable the SPAM DVC plugin in Dragonfly and prepare the supplied WSL2 compute environment, or configure a compatible SPAM environment following the manual. The computation is CPU based; a GPU is not required.
Check the WSL distribution, Python interpreter, worker path and Windows job folder shown in the panel. Choose an output directory with enough space for the input volumes and results. An optional processing ROI limits the analysis to its bounding box.
Run the first analysis
- Select the aligned Reference Channel and Deformed Channel, and an optional Processing ROI.
- Start with Local Subset mode. Choose a subset that contains enough texture and a node step appropriate to the desired sampling density.
- Review geometry resampling, grayscale preprocessing and whether strain fields should be computed. Resampling onto the reference grid accounts for geometry differences; it does not replace registration.
- Confirm the WSL backend and output settings, then run DVC and follow the log.
- Inspect the returned DVC Channels and the saved result files. If report generation is enabled, review the generated Word report alongside the image results.
How parameters affect interpretation
A larger subset uses more texture but averages over a larger region and can blur localized changes. A smaller node step produces denser output and more computation; it does not by itself guarantee finer independent spatial resolution.
Global FEM is an alternative that describes the field on a mesh. Mesh size, smoothing and regularisation affect the displacement and strain you see. Compare settings on a small representative region before committing to a full volume.
Review displacement before interpreting strain
Check displacement fields for discontinuities, boundary artefacts and implausible patterns, and review the run log. Strain derives from changes in displacement and can magnify measurement noise. A completed run alone does not establish measurement accuracy.
The outputs include DVC_* Channels in Dragonfly and result files in NPZ/JSON format, with an optional Word report. Preserve the reference/deformed scan identities, physical calibration, preprocessing and correlation settings when reporting an analysis.