Skip to content

Gallery

What SMINT output looks like on real tissue. Each figure links to the guide or tutorial that produces it.

Segmentation and registration

  • Whole-slide segmentation

    Nuclei, GFAP and phalloidin with cell and nucleus outlines

    Sample 3: cells (white) and nuclei (yellow) segmented with a fine-tuned Cellpose-SAM model on nuclei (cyan), GFAP (green) and phalloidin (magenta).

    Segmentation guide

  • Same-section registration

    Post-stain nuclei registered onto Xenium cells, before and after

    Sample 2, section z5: post-stain nuclei (pink) registered onto Xenium cells (blue) with multiround centroid registration.

    Same-section tutorial

  • How multiround registration works

    Displacement voting, block peaks and the resulting fit

    Displacement votes within a tissue block, the field of block peaks across the section, and the alignment before and after.

    Choosing a registration method

  • Snapping nuclei onto cells

    Before registration, after the smooth transform, and after snapping

    After the smooth transform, matched nuclei are snapped exactly onto their Xenium cells; 88% of cells receive a distinct nucleus on Sample 2 z5.

    Alignment guide

Spatial metabolomics registered onto spatial transcriptomics with STalign LDDMM

Sequential-section registration on Sample 2: MALDI spatial metabolomics (pink) onto Xenium spatial transcriptomics (blue) from a neighbouring section, before registration, with 11 paired landmarks, and after STalign LDDMM. Alignment guide.

3D cell-type maps of three samples

Registered serial sections stacked into 3D. Each point is a cell, coloured by its annotated cell type; up to 3,000 cells per cell type are shown for each sample so the page stays responsive.

Using the 3D view

  • Choose a sample with the Sample A / B / C buttons above the plot.
  • Rotate by dragging, zoom with the scroll wheel, a pinch or the
  • / − buttons, and reset the view with ↻.
  • Open the legend with the Legend button to see which colour is which cell type.
Sample Data Shown as
Sample A Spatial transcriptomics, cell centroids with a z coordinate 3D scatter, coloured by cell type
Sample B Spatial transcriptomics, eight registered serial sections Sections stacked by z position
Sample C Spatial transcriptomics, eight sections (layers z1–z8) Sections stacked by layer

Cell-type labels follow the glioblastoma tumour-microenvironment taxonomy (AC-like, MES-like, OPC-like, NPC-like) plus normal brain and immune populations. The page is generated by scripts/visualization/generate_3d_comparison.py with Plotly.