v2.2.2

Added

• Added CI for MacOS and Windows by @alihamraoui in #391
• Add support for local StarDist models by @alihamraoui in #392
• Add scene and timepoint support to bioio reader by @alihamraoui in #395
• Add tiffslide>=3.0.0 (which supports zarr v3) in the wsi extra, and use as a default backend for sopa.io.wsi

Changed

• Aggregation refactoring by @alihamraoui in #401. Channel aggregation now returns an AnnData object. Can choose not to drop cells in sopa.aggregate.

Fixed

• Fixed ome-zarr issue when writing to disk #400

New Contributors

• @alihamraoui made their first contribution in #391

Full Changelog: v2.2.1...v2.2.2

v2.2.0

[2.2.0] - 2026-02-24

This release unpins zarr, allowing for upgrading to the newer SpatialData version.

Everything should be backward compatible, except for some updates in the sopa.io.wsi reader (more details below).

Changed

• Moved to zarr>=3.0.0.

Fixed

• Fixed 'No module named "pkg_resources"' (#387) due to installing newer SpatialData version.

Added

• Python 3.14 should now be supported. More tests will be performed to ensure stability.
• Faster import sopa (more lazy loaded packages).
• Add spatial plot colored by cell annotation in the Sopa report (if any cell type or leiden)

Breaking changes

• Drop support for python==3.10. Sopa now depends on python>=3.11.
• sopa.io.wsi now uses openslide as a default backend, and uses wsi as a default image name (see all function arguments in the docs).
• Removed deprecated sopa.io.wsi_autoscale

v2.1.11

[2.1.11] - 2025-12-18

Added

• Use the new stardist version to remove numpy pinning (now, all extras can use numpy>=2.0.0).
• gene-column argument is now optional in Baysor and Comseg configs (inferred by default)

Fixed

• Also ensure shapes transformation is always 2D when the transformation is extracted from images
• Xenium v4 users: the Explorer should now open the correct new-format image
• Use the right gene column in comseg when no config is provided (i.e., inferred by Sopa).

Changed

• Remove support for the Xenium Explorer <4.0.0, please update it if not already done.
• Remove deprecated sopa crop command line

v2.1.10

[2.1.10] - 2025-12-05

Added

• proseg support for Visium HD data (see the related tutorial).
• Bypass spatial join when using only one patch to speed-up make_transcript_patches (#362)

Fixed

• Update attrs on disk when updating the latest Sopa boundaries (#363)
• Ensure shapes transformation is always 2D (#359)

v2.1.9

[2.1.9] - 2025-11-14

Fixed

• Fix assign_transcript_to_cell when the GeoDataFrame index name is already a column name (#346)
• Xenium v4 aggregation issue due to the presence of / in the channel names (#353)
• Ensure low-quality transcripts are not counted during aggregation (#349)

Changed

• Building Docker images using python 3.12 instead of python 3.10
• Lower the default tolerance during vectorization for more precise shapes (#340)

Added

• Setting sopa.settings.simplification_tolerance to change the default shapely tolerance. For instance, set it to 0.1 for low simplification, or 0 for no simplification (#340)
• Add an argument to load cells_boundaries and cells_table in sopa.io.merscope (False by default) (#346)
• Do not overwrite the prior shape key in the transcript dataframe (useful if running multiple times make_transcript_patches) (#354)

v2.1.8

Hot fix: pin pyarrow<22.0.0
Was fixed in scverse/spatialdata#1002 but we can't use the latest spatialdata version until we support zarr v3 (#347)

v2.1.7

[2.1.7] - 2025-10-04

Added

• Pin zarr<3.0.0 until we fully support it (see #347 for progress on this)
• WSI reader improvements: add slideio backend and faster embedding inference when not saving on disk @stergioc (#218)
• Argument to choose whether to run HVG in scanpy_preprocess (False by default)

Fixed

• Ensure the H&E embeddings are deterministic: using timm for resnet, and ensure always in eval mode (#218)

Minor

• CI improvements (#348)

v2.1.5

[2.1.5] - 2025-10-11

Added

• Use the default proseg presets if infer_presets = True (default) and if not yet provided in the command_line_suffix
• Exclude the CosMX control genes from the transcript patches and the aggregation
• Added optional only_excluded argument in count_transcripts to count the genes that are normally excluded from the counts (#329)

Fixed

• Ensure that cell expansion with no overlap preserves a polygon output (#318)
• Fixed proseg usage on CosMx data (#323)
• Don't perform segmentation on extremely small patches @Marius1311 (#282)
• Fixed incorrect prior cell ID assignment to float dtype instead of int in the CosMx reader when fov is not None (#335)
• Do not save the cache column in the transcript patches to allow moving the .zarr directory before segmentation

Broken changes

• The CosMX reader now stores the transcript coordinates in microns instead of pixels, so Baysor/Comseg config needs to be adjusted (#323)

v2.1.4

[2.1.4] - 2025-09-29

Added

• Visium HD: also store the spatial coordinates of the bins in microns (not just in pixels)
• Added an --overwrite option to the sopa convert command line (#306)
• Support python==3.13
• Add roi_key in image/transcript patches to decide which shapes to use for the region of interest (#309)
• Optional command line sopa scanpy-preprocess for basic preprocessing (see Snakemake config example)

Fixed

• Use csr matrices instead of coo in transcript and bins aggregation to support anndata>=0.12.0 (#305)
• Fix ValueError due to forward slash in morphology marker name @professor-sagittarius (#311)

v2.1.3

[2.1.3] - 2025-08-29

Added

• Add support for the new proseg version >=3.0.0
• Added sopa-comseg Docker image for ComSeg support in nf-core/sopa
• Can read CosMx polygons and cell labels (#285)

Fixed

• Overwrite the scale parameter when running baysor and providing both config and scale parameters (#294)
• Avoid losing prior cell 0 in prior assignment when unassigned_value != 0
• Fixed stardist dependencies, as it still doesn't support numpy>=2.0.0
• Fixed FOV shift in some CosMx versions (#286)