This page walks through two complete analysis workflows in ariadne.ai SPATIAL, from data upload to statistical results. These examples cover the most common use cases in spatial biology. For a detailed description of the methods used in each step, see the Statistical analysis page.
Goal: go from raw image data to a quantitative, per-cell expression table that is ready for cell type identification and spatial analysis.
Typical use cases: any multiplexed immunofluorescence experiment (CODEX, Phenocycler, SignalStar), spatial transcriptomics (Xenium, Visium HD), or serial IHC dataset.
Navigate to spatial.ariadne.ai and log in. On the dataset management screen, click + DATASET to create a new dataset and upload your image files. SPATIAL accepts all common imaging formats.
Organize multi-dataset experiments into folders (e.g. one folder per patient or condition). This is important later for cohort-based comparisons in the Plotting job.
Skip this step if all your markers were acquired in a single imaging session with no channel misalignment or if they have been aligned already.
If your experiment spans multiple staining cycles, channels acquired in later cycles will have small spatial shifts or deformations relative to the first cycle. Registration corrects these before any further analysis.
Check the registration quality in the viewer by toggling the registered layers on and off. Always use the registered versions of your layers in all subsequent steps.
Cell segmentation converts the raw image into a map of individual cells — the foundation of all single-cell spatial analysis.
If you only have a nuclear marker available, use Nucleus Expansion instead: the algorithm segments the nuclei and expands the masks outward to approximate full cell boundaries.
Inspect the segmentation result in the viewer by enabling the segmentation layer. Each cell should be cleanly outlined.
With segmented cells, you can now measure how much of each marker is present inside each cell.
When this job completes, SPATIAL creates a Mapping — a per-cell table with one row per cell and one column per marker. This table is the basis of all downstream analysis.
With the Mapping ready, you can define cell populations. There are two complementary approaches — you can use either or both.
Use the right panel to display marker intensity histograms and scatter plots. Draw gates to define cell populations based on co-expression of markers (e.g. gate on CD3⁺ CD8⁺ to define cytotoxic T cells, or CD68⁺ to define macrophages). Each gate creates a Selection — a named group of cells used in all downstream spatial analysis jobs.
For datasets with many markers (10+), run a t-SNE or UMAP job. When the job completes, draw gates around clusters of interest directly on the scatter plot in the right panel to define Selections. The cells belonging to each Selection are simultaneously highlighted in the image viewer, letting you see where each cluster is located in the tissue.
Alternatively, run Automatic Cell Type Detection to let SPATIAL assign cell type labels automatically. This adds a Categorical Property to your Mapping that can be used directly as a grouping variable without manual gating.
With Selections defined, you are ready to quantify the spatial relationships between cell populations.
To find which pairs of cell populations tend to be spatially close or avoid each other:
→ Run Neighborhood Enrichment. This gives you a pairwise co-localization score for all your cell populations.
To identify recurring spatial microenvironments (e.g. immune niches, tumor-stroma interfaces):
→ Run Recurrent Cellular Neighborhoods. This clusters cells based on their spatial context, revealing the tissue microenvironments present in your data.
Goal: determine whether two populations of interest are spatially associated in your tissue, and identify exactly where this co-localization occurs.
Typical use cases: quantifying immune cell infiltration into tumour tissue, measuring proximity between macrophages and necrotic regions, or identifying tertiary lymphoid structures in inflammatory disease.
Prerequisites: complete Workflow 1 through Step 5. You need a Mapping and at least two defined Selections or Categorical Properties before starting here.
Make sure you have clearly named Selections for both populations of interest.
For two cell types (e.g. CD8⁺ T cells and tumour cells):
Use manual gating, UMAP/t-SNE clustering, or ACD to create a Selection for each. Name them clearly (e.g. CD8_T_cells, Tumor_cells).
For a cell type and a pathology region (e.g. macrophages near necrotic tissue):
Neighborhood Enrichment gives you a single score that summarizes whether two populations tend to be spatially close or not, across the entire tissue section.
Interpreting the result:
To compare co-localization across patients or conditions, use the Plotting job with Cohorts to visualize enrichment score distributions as a box plot.
For a detailed description of the method, see Statistical analysis.
Neighborhood Enrichment tells you whether two populations are associated, but not where in the tissue this happens or what the surrounding cellular context looks like. Recurrent Cellular Neighborhoods answers these questions.
Interpreting the result:
After the job completes, open the RCN results in the right panel. Look for an RCN that is enriched for both populations of interest — for example, a neighborhood where CD8⁺ T cells and tumour cells are both over-represented. This is your immune-tumour interface microenvironment.
Enable the RCN overlay in the viewer: cells belonging to each neighborhood are color-coded, so you can see immediately where in the tissue each microenvironment occurs.
For a detailed description of the method, see Statistical analysis.
| Question | Job |
|---|---|
| Are my two populations spatially associated? | Neighborhood Enrichment |
| Where do they co-localize, and what is the surrounding context? | Recurrent Cellular Neighborhoods |
| How does co-localization differ across patients or conditions? | Plotting with Cohorts |
| Which cells belong to each population? | Manual gating, t-SNE/UMAP, or ACD |