3D Multiplexed Cell Detection Recipe

The 3D Multiplexed Cell Detection recipe in Aivia is based on Cellpose, a generalist deep-learning based algorithm for cellular segmentation. This recipe is designed for 3D data.

The recipe detects the cell membrane or nuclear membrane using one or multiple membrane marker channels. It provides options for two cell/nucleus detection methods: (1) the standard method that reimplemented original Cellpose segmentation method with speed optimization, (2) the faster flow-based method that only uses the two Cellpose gradient maps to detect cells/nuclei.

The recipe allows user to choose from four image smoothing options: (1) skip smoothing, (2) apply average filter, (3) apply Gaussian filter, and (4) apply median filter. The selected image smoothing option is applied to the input channel(s.) If multiple input channels are selected, they will be averaged to generate a combined image for cell/nucleus detection. 3D meshes and 2D slices of detected cells are created as the results. Previews are available for the combined image, probability map, and 2D slices of detected cells.

On this page:

Inputs and Outputs

Recipe inputs and outputs for Multiplexed Cell Detection and their descriptions are summarized in the table below.

	Name	Description
Inputs	Deep Learning Model	Select a deep learning model to apply before this recipe is applied
Inputs	Input Image(s)	Select the cell membrane or nuclear membrane marker channel(s) to be used while detecting cells in this recipe
Outputs	Cells	Select the object set to which the results of the detection are outputted. Both 3D meshes and 2D slices of detected cells are created. Defaults to creating a new object set.

Parameters and Presets

Parameters

Recipe parameters for Multiplexed Cell Detection and their descriptions are summarized in the table below.

Preset Group	Parameter Name	Min Value	Max Value	Description
Detection	Model Type	1	3	Specifies the model type that can be applied The available model types are the following: 1: ‘cyto’; 2: ‘nuclei’, 3: ‘cyto2’ cyto2 yields good results in most tested cases including nuclei and cytoplasm and is the recommended first step.
	Typical Cell Diameter	0	1,000	Specifies the typical diameter of objects (e.g. cells / nuclei) on the image. This parameter is used for rescaling the image input to the Cellpose detection algorithm
	Probability Threshold	0	100	Adjusts the sensitivity of cell/nucleus detection. The threshold is applied to the probability map shown in the Detection preview. Higher values mean lesser number of detected cells or nuclei
	Image Smoothing Filter Size (excluding Skip Smooth Image only)	0	100	Specifies the diameter of the filter that is used to smooth the input channel before further processing The available smoothing types are the following: Skip Smoothing - recommended setting for Cellpose as the first step before trying additional smoothing below Average Filter Smoothing applies an average filter with a filter kernel width and height specified by the value of the Image Smoothing Filter Size. The averaging filter is applied in 2D to each Z-stack of the input channel(s). Gaussian Filter Smoothing applies a Gaussian filter with a filter kernel width and height specified by the value of the Image Smoothing Filter Size. The Gaussian filter is applied in 2D to each Z-stack of the input channel(s). Median Filter Smoothing applies a median filter with a filter kernel width and height specified by the value of the Image Smoothing Filter Size. The Median filter is applied in 2D to each Z-stack of the input channel(s).
Partition	Cell Diameter	0	10,000	Specifies the range of detected cells or nuclei to keep based on their diameters
	Mesh Smoothing Factor	1	10	Adjusts the amount of smoothing applied to the surface reconstructions of the detected Cells; a lower value will generate surfaces with greater similarity to the input image
	Z-Stack Area Overlap Threshold (For the “Use standard method” option)	0.0	1.0	Specifies the intersection over union (IoU) threshold to determine whether to merge masks across Z stacks. A higher threshold will result in stricter merging criteria, potentially leading to more detected cells but introducing the risk of splitting a cell into partial cells. A lower threshold may result in more merged cells, potentially capturing more detail but introducing the risk of merging unrelated structures.
	Max Z-Stack Displacement (For the “Use flow-based method” option)	0	1,000	Specifies the displacement limit between mask centers to determine whether to merge masks across Z stacks. A lower value will result in stricter merging criteria, potentially leading to more detected cells but introducing the risk of splitting a cell into partial cells. A higher value may result in more merged cells, potentially capturing more detail but introducing the risk of merging unrelated structures.

Presets

There are two preset groups in the recipe: (1) Detection and (2) Partition. Each group has three pre-configured parameter groupings to help you get started on the analysis. The default preset values are in the sections to follow.

Partition

Parameter Name	Low	Medium	High
Cell Diameter	1~30 px or µm	2~100 px or µm	20~200 px or µm
Mesh Smoothing Factor	0	0	0
Z-Stack Area Overlap Threshold (For the “Use standard method” option)	0.2	0.5	0.8
Max Z-Stack Displacement (For the “Use flow-based method” option)	2	10	40

User Interface and Previews

User Interface

Standard method

Flow-based method

Detection Preview

Two intermediate images and 2D/3D view of detected cells or nuclei are available for detection preview

(a) Combined Image - Preview
- Preview of the combined image used for cells or nuclei detection
(b) Probability Map - Preview
- Preview of the probability map representing the likelihood of cells/nuclei to be detected
  - Values of the probability map range from 0 to 100
    - Standard method probability map
    - Flow-based method probability map
(c) 3D Meshes - Preview
- Preview of 3D meshes of detected cells or nuclei
(d) 2D Slices - Preview
- Preview of 2D slices of detected cells or nuclei

Partition preview

2D/3D view of detected cells or nuclei are available for partition preview
- Same as “3D Meshes - Preview” and “2D Slices - Preview” in detection preview

Tutorial (written steps)

Before beginning the tutorial, please download the 3D Multiplexed Cell Detection Demo image. For information on how to select presets or modify parameter values, please refer to the tutorial on how to use the Recipe Console.

Unzip the demo file and load the demo image, “3D Multiplexed Cell Detection Demo.aivia.tif” into Aivia.
In the Recipe Console, click on the Recipe selection dropdown menu and select the 3D Multiplexed Cell Detection recipe.
Select following membrane marker channels in the dropdown menu for Input Images:
- OPAL 570 Gating
- OPAL 680 Gating
- OPAL 7800 Gating
Click on the Show Advanced Interface icon to expand the Recipe Console and show parameter options for the recipe.
Modify the parameter values as the follows:
- Detection
  - Model Type: 3
  - Typical Cell Diameter (µm) : 7
  - Probability Threshold: 50
  - Select “Average Filter Smooth” and set
    - Image Smoothing Filter Size: 3
- Partition
  - Cell Diameter (µm) : 4~50
  - Smoothing Factor: 2
  - Select “Use standard method” and set
    - Z-Stack Area Overlap Threshold: 0.5
Click the Start button or press the F4 key on your keyboard to begin applying the recipe to the image.

The detected cell or nuclear membrane outlines will be overlaid on the image for viewing in 2D or 3D.

Input image

Parameters

Detection result

Detection result in 2D view

Detection result in 3D view