| Operation | Input | Description |
|---|---|---|
| Cut bottlenecks | Coded image | Identifies thin connections (isthmuses) between label regions and severs them, splitting objects connected through narrow necks into separate events. |
| Fill labels | Coded image | Fills enclosed holes within labeled objects — background pixels completely surrounded by a single label become part of that label. |
| Graph from labels | Coded image | Constructs a spatial adjacency graph: nodes = labeled events, edges = shared borders between events. Output enables neighborhood analysis. |
| Grow labels | Coded image + radius | Expands each label's boundary outward by the specified distance (morphological dilation per label). Labels grow independently; conflicts at meeting points are resolved by proximity. |
| Re-Label (missing + duplicates) | Coded image | Full re-indexing: resolves duplicate label assignments and removes gaps from deleted labels. Output has contiguous labels from 1 to N. |
| Re-Label (missing) | Coded image | Lighter re-indexing: renumbers labels sequentially to fill gaps from removed objects, but does not check for duplicates. |
| Shrink labels | Coded image + radius | Contracts each label's boundary inward by the specified distance (morphological erosion per label). Very small labels may disappear entirely. |
BOM Group
Label Management
Refining and manipulating coded images after labeling
View
Definition
Post-labeling operations for refining coded images — cutting bottleneck connections, filling interior holes, growing or shrinking label boundaries, re-indexing labels, and constructing spatial adjacency graphs — the cleanup toolkit between raw labeling and final measurement.
Eight Refinement Operations
Cut, fill, grow, shrink, re-label, and graph
Topology Modification
Change how objects are connected and shaped
Label Index Maintenance
Re-indexing after object removal or splitting
Adjacency Graph
Build neighborhood relationships between labeled objects
Operations Reference
Simplified
Cut bottlenecks separates objects joined by thin bridges. Fill labels fills internal holes. Grow/Shrink expands or contracts object boundaries. Re-Label renumbers objects after removals. Graph builds a neighbor map showing which objects touch each other.
When to Use Which
Splitting Incorrectly Merged Objects
Cut bottlenecks — When thresholding or labeling merged two objects through a thin connecting bridge. The operation identifies the narrowest point of connection and cuts there, producing two separate labels. Most useful for tissue structures like glands connected through thin tissue bridges.
Cleaning Up Labels
Fill labels — When labeled objects have internal holes from uneven staining or thresholding artifacts. Filling ensures that area measurements reflect the true object extent, not just the consistently-stained portions.
Re-Label — After any operation that removes labels (cut, filter by size, etc.), run Re-Label to renumber everything sequentially. Use the full variant (missing + duplicates) after cut bottlenecks, which can create duplicate label numbers.
Boundary Modification
Grow labels — Expand labels to approximate cell territory beyond the detected nuclear boundary. Similar in concept to the Grow engine but operates within the BOM pipeline.
Shrink labels — Contract labels to focus measurements on the core of each object, excluding edge pixels that might be contaminated by background signal.
Spatial Analysis
Graph from labels — When you need to know which objects are neighbors. The adjacency graph enables questions like: 'which cells are adjacent to cell X?' or 'how many neighbors does each cell have?' This is the foundation for spatial interaction analysis within the BOM.
Simplified
Use Cut bottlenecks to split merged objects. Use Fill labels to close internal holes. Use Re-Label after any operation that changes the number of objects. Use Grow/Shrink to expand or contract object boundaries. Use Graph to build a neighborhood map for spatial analysis.