The LuCIM Volume Lab is a deliberately compact bridge between the foundational volume algorithms and future medical or atmospheric applications. It generates a deterministic, non-diagnostic CT-like phantom containing air, tissue, lung-like cavities, cortical structures, calibration inserts, and an explicitly missing acquisition wedge. No patient data, external asset, decoder, or network request is involved.
Three synchronized panels show axial, coronal, and sagittal slices. Cyan crosshairs connect their selected coordinates. Anatomy applies an adjustable density window directly to the resident source values; Threshold overlays the current classification; and Components colors the deterministic sparse roots produced after three-dimensional closing and six-connected labeling.
Changing the density threshold writes one GPU scalar and re-encodes the already compiled graph:
- calibrated volume thresholding;
- ball-footprint binary closing;
- bounded connected-component stabilization; and
- fixed-capacity voxel counts and three-dimensional index bounds.
Persistent source values, masks, component labels, and region rows remain GPU-resident and feed the slice renderer directly. The inspector reads back only convergence, iteration count, and overflow flags. An invalid observation remains separate from valid air or background and appears as a hatched wedge rather than being silently converted to zero.
This tranche keeps its tri-planar renderer and synthetic source inside the example. A later CAT tranche can add physical region measurements and a licensed NIfTI or preprocessed source adapter; the hurricane track can separately pressure-test chunking, three-dimensional halos, and time steps before any example-local abstraction is promoted into a public API.