MorphoStack is a lab morphometry toolkit. Treat outputs as exploratory until calibration, segmentation quality, and (where relevant) profile-specific validation are in place.
- TIFF/LSM stacks without embedded spacing fall back to 1 × 1 × 1 µm.
- Surface area, volume, and perimeter in µm are not biological under default spacing.
- CZI metadata is read when present — still confirm against the acquisition log.
- Manual overrides appear in manifests as
voxel_source: override.
- Seed (circle or polygon) selects one connected component on the seed frame.
- Tracking uses overlap-first association across Z, with centroid-distance fallback.
- Tracking can fail when objects move quickly, touch neighbors, or leave the ROI.
- UI debug overlay and manifest
trackingfields show per-frame centroids and losses. - MorphoStack does not auto-segment every object in a crowded field.
| Profile | Status |
|---|---|
| Vesicle | Primary threshold-contour path + optional mesh |
| RBC | Same engine with RBC-oriented defaults; biconcavity/thickness metrics not validated |
| Active surfaces | Experimental surfel refinement — compare to threshold before paper use |
- Meshes come from marching cubes on aligned contour masks.
- Empty or tiny contours → empty/skipped mesh.
- Exported geometry uses the active voxel calibration; uncalibrated exports are for shape review only.
- Browser mesh preview may be decimated; use CSV/export files for numbers.
- One run targets one seeded object.
- Neighbor-merge warnings fire when tracked area jumps relative to the seed frame.
- Watershed pre-split (active-surfaces path) helps but does not guarantee separation of strongly touching cells.
- Not a full ImageJ/Fiji replacement or multi-object instance segmenter for whole fields.
- Not a deep-learning segmentation suite (Cellpose/StarDist are intentionally out of scope as required deps).
- Not a substitute for manual scientific review of contours on publication datasets.