Label Management

Every operation here except Label and Graph from labels takes a coded image and returns one; Graph from labels returns a mask. Seeds ignore mask matters when an event already reaches outside the mask. In the reference's Grow labels example, with it set to No everything outside the allowed area is deleted, and with Yes the full seed bodies are kept.

Source: StrataQuest 8.1 Engines Reference, ch. 6.6 (pp. 181–205).

Splitting Joined Objects

Use Cut bottlenecks when two structures were detected as one event because a thin area joins them, which the reference calls under-segmentation. Its example is a pair of colon structures joined by a narrow neck. Bottleneck -max length (px) caps how long a cut may be, and Bulge -min area (px) rejects cuts that would leave a piece too small. When objects touch along a broad front rather than a neck, a cut-based split has nothing to find; that is the kind of case WatershedLoading... and the distance mapLoading... are usually used for.

Filling and Renumbering

Use Fill labels when labeled objects have interior holes from uneven staining or thresholding, so area covers the whole object. On a mask, before labeling, the matching operation is Fill holesLoading....

Run a Re-Label operation after anything that changes the label list. The reference names Remove labels as a source of missing labels, so Re-Label (missing) follows it. It names Fuse events as a source of duplicates, so Re-Label (missing, duplicates)Loading... follows Fuse eventsLoading.... Duplicates are the worse problem: in image analysis generally, two separate bodies with one ID are measured as one object.

Boundary Modification

Use Grow labels to expand events beyond their detected boundary, for example to approximate the territory around each nucleus. Because each pixel goes to the closest label, neighbors meet roughly halfway between them instead of overlapping. Add a mask to keep growth inside tissue. The Grow (global)Loading... engine does related work in a different place: it builds measurement masks in the layer's Measurements Masks section, with Grow, Shrink and Ring modes.

Use Shrink labels to focus measurements on the core of each object and leave out the edge pixels, where neighboring signal may mix in.

Spatial Analysis

Use Graph from labels to draw connections between events that lie within a chosen distance of each other. The reference's example connects nuclei closer than 15 pixels, and the yellow lines overlaid on the tissue show which cells are near which. The output is a mask: good to view, or to process further with other mask operations, but it doesn't list neighbors per cell by itself. In general terms, this is a distance-threshold graph; with Distance mode set to contours, two large cells can connect even when their centers are far apart.