Cellpose-based Segmenter parameters

Channel

Sets the channel where the cells will be segmented.

Optional channel

Sets a second optional channel (e.g., nuclei channel) to support the segmentation.

Cellpose model

Select a cellpose model for inference. You can load and run any custom trained or pre-trained cellpose model (e.g. cellpose2, cellpose3 or cellposeSAM) in the operation.
Note: By default, all available downlaoded models in the Models folder (configured in Cellpose preferences) are listed in the dropdown. Use refresh button to always see the most recent available models. Alternatively, you can browse and load a model from any other directory by selecting Load from a file option from the dropdown. Use [...] to interactively select a cellpose compatible model in ONNX format.
Once the model is loaded, the operation will validate the model for matching criterion and if successful, will use it. Otherwise, the operation will show an error message or a warning.

Note: For instructions on how to convert cellpose models to ONNX refer to: Cellpose to ONNX Conversion Script for arivis

–

CP

Sets a model, which is mainly trained on fluorescent cell images .

–

CPx

Sets a model, which is trained on wide variety of images.

–

Cyto

Sets a cytoplasm model, which is trained on one channel cell images (if the image had a nuclear channel).

–

Cyto2

Sets a cytoplasm model, which is trained on two channels cell images (if the image had a nuclear channel).

–

Nuclei

Sets a model, which is trained on two-channel images. The first channel is the channel to segment, and the second channel is always set to an array of zeros.

–

Load external model

Sets a custom model, which allows you to load and run your trained Cellpose models or any pre-trained Cellpose models not included in the operation.
Note: Refer to the Help (F1) for more information.

Cell diameter

Sets value of the cell diameter in your data. Alternatively, use the Measure tool to determine it.
If you deactivate the cell diameter, the operation will assume the cell diameter matches the trained model diameter.
Note: Image data will be re-scaled according to the diameter and Model diameter to match your selected model.
Note: The Cellpose-SAM model has a good amount of size-invariance, it has been trained on images with different cell diameters, from 7.5µm to 120µm. So this parameter is optional for the CPSAM model. However, if your cells are very big, you may want to use the diameter to down-sample the data, which will increase inference speed.

Min area

Sets the reference objects diameter. Select a value to filter out the objects with smaller area, volume, or diameter.

Mask threshold

Sets the cell probability threshold between -6 and +6 to determine what is a cell and what is background.

Mask quality threshold

Sets the maximum allowable errors (mean square error) of the flows for each detected cell. The filter range is between 0 and 1 to exclude masks with lower quality values.

Smooth Cellpose model results

Sets the Gaussian smoothing process, to refine the output, reduce noise and potentially improve the overall performance of your cell detection application.

Normalization percentile

Sets the normalization range. The default values are 1% to 99%, where everything below the lower threshold and everything above the upper threshold is mapped out as outliers.

Note: Increase the range if you have data with big background areas and very little area of actual bright signal