Reference: every SP Classify menu item¶
All under Extensions → SP Classify, in menu order.
| Item | Requires | Action |
|---|---|---|
| Binary Classifiers... | Project | Opens the binary classifier manager: create, open or delete one classifier per marker. |
| Composite Classification... | Project + ≥1 trained binary classifier | Applies trained binary classifiers and assigns composite labels. |
| Class Control... | Project | Add, delete, merge and undo-merge classes. |
| Channel Mapping (Review Display)... | Open image | Sets which image channels review shows for each class. Saved per project, with CSV import and export. See Marker table format. |
| Select Features... | Project | Choose which measurement columns are used for training. |
| Clustering Normalisation | Project | Sets an arcsinh or square-root transform per feature, with a shared cofactor, for clustering only. The classifier uses raw values. See §4.2. |
| Project Prediction Summary... | Project | Cohort QC: anomaly score and flags for each image. See §8. |
| Intensity Heatmaps... | Open image with detections | Heatmap of mean marker intensity per class (z-score colours), for one image or the whole project. PNG and CSV export. See §9. |
| Image Pixel Prescreen... | Project | Whole-image QC without cells: per-channel pixel statistics compared across the project, with flags for background-heavy, saturated, weak-signal and intensity-outlier images. CSV export. See §17. |
| Scatter Plots and Clustering... | Open image with detections | PCA/UMAP scatter plot with k-means or Leiden clustering, annotation and class gating, and cluster → class assignment. Scope switches to clustering across several images. See §11. |
| Cellular Neighborhoods... | Open image with cells in at least 2 classes | Groups cells by the cell types around them (k-means on neighbourhood composition), for one image or the whole project. See §18. |
| Batch Normalisation... | Project | Aligns each marker's intensity across images (UniFORM). Fit and QC, then use the corrected values in clustering and the classifier, or write them as (batchnorm) columns. See §19. |
| Generate Distance Measurements... | Project | Distances to annotations (signed), to other classes, and to the nearest cell of the same class, for the selected images. See §10. |
| Export → Cell Table... | Open image with detections | One CSV per selected image. See §12.1. |
| Export → Ground Truth... | Open image with labels (multi-class) | CSV of labels and feature values that you can import into another project. |
| Export → Active Binary Ground Truth... | Binary mode active + open image with labels | As above, for the active marker only. |
| Import → Marker Table... | Open image | Loads a cell type → markers CSV used to switch channels during review. The Channel Mapping editor is the recommended way to edit it. |
| Import → Ground Truth... | Open image (multi-class) | Imports labels by spatial match, or as training data only. |
| Import → Active Binary Ground Truth... | Binary mode active + open image | As above, for the active marker only. |
| Utility Scripts → Filter Cells by Size & Circularity... | Open image with cells | Removes cells outside optional area and circularity limits, in the current image. See §13.1. |
| Utility Scripts → Resolve Hierarchy... | Open image or project | Rebuilds parent/child relationships between objects, for the current image or the whole project. See §13.2. |
| Utility Scripts → Lock All Annotations | Open image or project | Locks every annotation, in the current image or in all project images, so it cannot be moved or edited by accident. See §13.7. |
| Utility Scripts → [TEST] Import GeoJSON Objects... | Open image | Imports objects from a GeoJSON or gzipped GeoJSON file into the current image. Small to medium files only. See §13.4. |
| Utility Scripts → [TEST] Export Annotation Regions... | Open image | Exports annotation regions from the current image as polygon-masked OME-TIFF files. One image, small to medium regions only. See §13.5. |
| Utility Scripts → Delete Measurements by Keyword... | Open image or project | Destructive: deletes detection measurements whose names contain a keyword. Shows the matches and asks you to confirm. See §13.3. |
| Utility Scripts → Reset Project State... | Project | Destructive: deletes the project's celltune/ folder (labels, models, predictions, settings). Writes a backup zip first and asks you to type RESET. Can also clear SP Classify label points and cell classifications from every image. See §13.6. |
| How to Cite... | Nothing (always available) | Shows the references to cite for SP Classify and for the methods you used. |