Merge pull request #287 from BodenmillerGroup/cellpose_sam

Cellpose sam

Author: victorjonathanibanez

AUTHORS.md

@@ -2,4 +2,4 @@
 
 Created and maintained by [Jonas Windhager](mailto:jonas@windhager.io) until February 2023.
 
-Maintained by [Milad Adibi](mailto:milad.adibi@uzh.ch) from February 2023.
+Maintained by [Victor Ibañez](mailto:victor.ibanez@uzh.ch) from February 2023.

CHANGELOG.md

@@ -5,6 +5,13 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.16.4] - 2026-02-26
+
+### Changed
+
+  - Updated readimc to 0.9.2
+  - Changed Cellpose container and description to v4.0.8, using CellposeSAM
+
 ## [0.16.3] - 2024-05-29
 
 ### Changed

Dockerfile

@@ -15,7 +15,7 @@ ARG FIXUID_VERSION="0.5.1"
 ARG ILASTIK_BINARY="ilastik-1.3.3post3-Linux.tar.bz2"
 ARG CELLPROFILER_VERSION="4.2.5"
 ARG CELLPROFILER_PLUGINS_VERSION="4.2.1"
-ARG CELLPOSE_VERSION="2.2"
+ARG CELLPOSE_VERSION="4.0.8"
 
 ########## TENSORFLOW ##########
 
@@ -186,22 +186,8 @@ RUN python -m pip install "cellpose==${CELLPOSE_VERSION}"
 USER steinbock:steinbock
 
 RUN mkdir -p /home/steinbock/.cellpose/models && \
-    cd /home/steinbock/.cellpose/models && \
-    curl -SsO https://www.cellpose.org/models/cytotorch_0 && \
-    curl -SsO https://www.cellpose.org/models/cytotorch_1 && \
-    curl -SsO https://www.cellpose.org/models/cytotorch_2 && \
-    curl -SsO https://www.cellpose.org/models/cytotorch_3 && \
-    curl -SsO https://www.cellpose.org/models/size_cytotorch_0.npy && \
-    curl -SsO https://www.cellpose.org/models/nucleitorch_0 && \
-    curl -SsO https://www.cellpose.org/models/nucleitorch_1 && \
-    curl -SsO https://www.cellpose.org/models/nucleitorch_2 && \
-    curl -SsO https://www.cellpose.org/models/nucleitorch_3 && \
-    curl -SsO https://www.cellpose.org/models/size_nucleitorch_0.npy && \
-    curl -SsO https://www.cellpose.org/models/cyto2torch_0 && \
-    curl -SsO https://www.cellpose.org/models/cyto2torch_1 && \
-    curl -SsO https://www.cellpose.org/models/cyto2torch_2 && \
-    curl -SsO https://www.cellpose.org/models/cyto2torch_3 && \
-    curl -SsO https://www.cellpose.org/models/size_cyto2torch_0.npy
+    curl -L -sS -o /home/steinbock/.cellpose/models/cpsam \
+      https://huggingface.co/mouseland/cellpose-sam/resolve/main/cpsam
 
 
 

docs/cli/segmentation.md

@@ -96,8 +96,8 @@ This will create grayscale cell/nuclear masks of the same x and y dimensions as
 
 ## Cellpose
 
-!!! danger "Experimental feature"
-    This is an experimental feature and is only available in the `-cellpose` flavors of the *steinbock* Docker container.
+
+!!! This feature is only available in the `-cellpose` flavors of the *steinbock* Docker container.
 
     Segmentation using cellpose likely requires fine-tuning of parameters, e.g. using steinbock command-line interface options.
 
@@ -106,22 +106,16 @@ This will create grayscale cell/nuclear masks of the same x and y dimensions as
 !!! note "End-to-end cell segmentation"
     This approach operates directly on image intensities and does not require a preceding pixel classification step.
 
-To segment cells using the default `cyto2` model:
-
-    steinbock segment cellpose --minmax
+To segment nuclei or cells using the default `cpsam` model:
 
-To segment nuclei using the `nuclei` model:
-
-    steinbock segment cellpose --minmax --model nuclei
+    steinbock segment cellpose
 
 !!! note "Cellpose image data"
-    Cellpose expects two-channel images as input, where the first channel must be a nuclear channel (e.g. DAPI) and the second channel must be a cytoplasmic channel (e.g. E-Cadherin). The nuclear channel is optional and only the cytoplasmic channel ("channel to segment") is required. Note that - compared to the original cellpose implementation - the channel order is reversed for compatibility with DeepCell/Mesmer.
+    For cell segmentation, Cellpose expects maximum 3-channel images as input, where one channel should be a nuclear channel (e.g. DAPI) and the second channel should be a cytoplasmic channel (e.g. E-Cadherin), the order does not matter anymore in v4+. For nuclei segmentation, only a nuclei channel should be provided and the parameter `--diameter` has to be adjusted accordingly.
 
-    If a `cellpose` column is present in the *steinbock* panel file, channels are sorted and grouped according to values in that column to generate the required input for DeepCell: For each image, each group of channels is aggregated by computing the mean along the channel axis (use the `--aggr` option to specify a different aggregation strategy). The resulting images consist of one channel per group; channels without a group label are ignored.
+    If a `cellpose` column is present in the *steinbock* panel file, channels are sorted and grouped according to values in that column to generate the required input for Cellpose: For each image, each group of channels is aggregated by computing the mean along the channel axis (use the `--aggr` option to specify a different aggregation strategy). The resulting images consist of one channel per group; channels without a group label are ignored.
 
     If no `cellpose` column is present, images are expected to be in the correct format already.
 
 !!! note "GPU support"
-    Currently. steinbock does not support cellpose segmentation with GPU support.
-
-    If GPU support is required, consider running cellpose on your host system independently.
+    In Cellpose v4+ GPU usage is supported inherently - there is a check whether GPU devices are available, otherwise it will fall back to CPU usage.

setup.cfg

@@ -3,8 +3,8 @@ name = steinbock
 url = https://github.com/BodenmillerGroup/steinbock
 author = Jonas Windhager
 author_email = jonas@windhager.io
-maintainer = Milad Adibi
-maintainer_email = milad.adibi@uzh.ch
+maintainer = Victor Ibañez
+maintainer_email = victor.ibanez@uzh.ch
 classifiers =
     Operating System :: OS Independent
     Programming Language :: Python :: 3

steinbock/segmentation/_cli/cellpose.py

@@ -13,14 +13,8 @@
 cellpose_cli_available = cellpose.cellpose_available
 
 
-@click.command(name="cellpose", help="Run an object segmentation batch using Cellpose")
-@click.option(
-    "--model",
-    "model_name",
-    type=click.Choice(["nuclei", "cyto", "cyto2"]),
-    default="cyto2",
-    show_default=True,
-    help="Name of the Cellpose model",
+@click.command(
+    name="cellpose", help="Run an object segmentation batch using CellposeSAM"
 )
 @click.option(
     "--img",
@@ -60,13 +54,6 @@
     show_default=True,
     help="Numpy function for aggregating channel pixels",
 )
-@click.option(
-    "--net-avg/--no-net-avg",
-    "net_avg",
-    default=True,
-    show_default=True,
-    help="See Cellpose documentation",
-)
 @click.option(
     "--batch-size",
     "batch_size",
@@ -76,45 +63,48 @@
     help="See Cellpose documentation",
 )
 @click.option(
-    "--normalize/--no-normalize",
-    "normalize",
-    default=True,
+    "--resample/--no-resample",
+    "resample",
+    default=False,
     show_default=True,
     help="See Cellpose documentation",
 )
 @click.option(
-    "--diameter",
-    "diameter",
-    type=click.FLOAT,
-    help="See Cellpose documentation",
+    "--channel-axis",
+    "channel_axis",
+    type=click.INT,
+    default=0,
+    show_default=True,
+    help="See Cellpose documentation.",
 )
 @click.option(
-    "--tile/--no-tile",
-    "tile",
-    default=False,
+    "--normalize/--no-normalize",
+    "normalize",
+    default=True,
     show_default=True,
     help="See Cellpose documentation",
 )
 @click.option(
-    "--tile-overlap",
-    "tile_overlap",
-    type=click.FLOAT,
-    default=0.1,
+    "--invert/--no-invert",
+    "invert",
+    default=False,
     show_default=True,
     help="See Cellpose documentation",
 )
 @click.option(
-    "--resample/--no-resample",
-    "resample",
-    default=True,
-    show_default=True,
+    "--rescale",
+    "rescale",
+    type=click.FLOAT,
+    default=None,
+    show_default=False,
     help="See Cellpose documentation",
 )
 @click.option(
-    "--interp/--no-interp",
-    "interp",
-    default=True,
-    show_default=True,
+    "--diameter",
+    "diameter",
+    type=click.FLOAT,
+    default=None,
+    show_default=False,
     help="See Cellpose documentation",
 )
 @click.option(
@@ -126,8 +116,8 @@
     help="See Cellpose documentation",
 )
 @click.option(
-    "--cellprobab-threshold",
-    "cellprobab_threshold",
+    "--cellprob-threshold",
+    "cellprob_threshold",
     type=click.FLOAT,
     default=0.0,
     show_default=True,
@@ -141,6 +131,36 @@
     show_default=True,
     help="See Cellpose documentation",
 )
+@click.option(
+    "--max-size-fraction",
+    "max_size_fraction",
+    type=click.FLOAT,
+    default=0.4,
+    show_default=True,
+    help="See Cellpose documentation",
+)
+@click.option(
+    "--niter",
+    "niter",
+    type=click.INT,
+    default=None,
+    help="See Cellpose documentation",
+)
+@click.option(
+    "--augment/--no-augment",
+    "augment",
+    default=False,
+    show_default=True,
+    help="See Cellpose documentation",
+)
+@click.option(
+    "--tile-overlap",
+    "tile_overlap",
+    type=click.FLOAT,
+    default=0.1,
+    show_default=True,
+    help="See Cellpose documentation",
+)
 @click.option(
     "-o",
     "mask_dir",
@@ -152,23 +172,25 @@
 @click_log.simple_verbosity_option(logger=steinbock_logger)
 @catch_exception(handle=SteinbockException)
 def cellpose_cmd(
-    model_name: str,
     img_dir,
     channelwise_minmax,
     channelwise_zscore,
     panel_file,
     aggr_func_name,
-    net_avg,
     batch_size,
+    resample,
+    channel_axis,
     normalize,
+    invert,
+    rescale,
     diameter,
-    tile,
-    tile_overlap,
-    resample,
-    interp,
     flow_threshold,
-    cellprobab_threshold,
+    cellprob_threshold,
     min_size,
+    max_size_fraction,
+    niter,
+    augment,
+    tile_overlap,
     mask_dir,
 ):
     channel_groups = None
@@ -179,24 +201,26 @@ def cellpose_cmd(
     aggr_func = getattr(np, aggr_func_name)
     img_files = io.list_image_files(img_dir)
     Path(mask_dir).mkdir(exist_ok=True)
-    for img_file, mask, flow, style, diam in cellpose.try_segment_objects(
-        model_name,
+    for img_file, mask, _, _ in cellpose.try_segment_objects(
         img_files,
         channelwise_minmax=channelwise_minmax,
         channelwise_zscore=channelwise_zscore,
         channel_groups=channel_groups,
         aggr_func=aggr_func,
-        net_avg=net_avg,
         batch_size=batch_size,
+        resample=resample,
+        channel_axis=channel_axis,
         normalize=normalize,
+        invert=invert,
+        rescale=rescale,
         diameter=diameter,
-        tile=tile,
-        tile_overlap=tile_overlap,
-        resample=resample,
-        interp=interp,
         flow_threshold=flow_threshold,
-        cellprob_threshold=cellprobab_threshold,
+        cellprob_threshold=cellprob_threshold,
         min_size=min_size,
+        max_size_fraction=max_size_fraction,
+        niter=niter,
+        augment=augment,
+        tile_overlap=tile_overlap,
     ):
         mask_file = io._as_path_with_suffix(Path(mask_dir) / img_file.name, ".tiff")
         io.write_mask(mask, mask_file)