Instance Segmentation Function

This module needs to be used in conjunction with the ACE segmentation module.

This module perfoms the following:

1. Load in the patches from ACE segmentation output folder

2. Perform connected component analysis on the patches to label neurons

3. Get region properties for each neuron

4. Save the results to a json file

5. Optionally stack the patches to form 2D image slices

Note

Stacking is needed to correct the neuron count since counting is done in parallel. The patches must be processed in sequence to get the correct neuron count, which is done by patch stacking.

Main outputs

1. Connected component patches of the neurons located at <input_folder>/cc_patches/

2. A json file containing the region properties of each neuron located at <input_folder>/cc_patches/neuron_info_final.json

3. Optionally, 2D image slices of the neuron labels located at <output_folder>

CLI

To look at the arguments that need to be provided to the function, invoke the help menu using:

$ miracl seg instance -h

The following menu will be printed to the terminal:

usage: miracl seg instance -i <input_seg_folder> [ -o <output_folder> ] [ -g <glob_pattern> ] [ -p <percentage_brain_patch_skip> ]

Instance segmentation step of ACE segmentation

This module perfoms the following:
    1. Load in the patches from ACE segmentation output folder
    2. Perform connected component analysis on the patches to label neurons
    3. Get region properties for each neuron
    4. Save the results to a json file
    5. Optionally stack the patches to form image slices

Notes: stacking is needed to correct the neuron count since counting is done
in parallel. The patches must be processed in sequence to get the correct
neuron count, which is done by patch stacking.

Required arguments:
-i INPUT_FOLDER, --input_folder INPUT_FOLDER
                        Path to ACE segmentation output folder
                        (generated_patches/)
-r RAW_INPUT_FOLDER, --raw_input_folder RAW_INPUT_FOLDER
                        Path to raw clarity tiff folder

Optional arguments:
-o OUTPUT_FOLDER, --output_folder OUTPUT_FOLDER
                        Path to output folder (default: <input_folder>.parent
                        /cc_slices/)
--properties PROPERTIES [PROPERTIES ...]
                        Properties to compute for each neurone (default:
                        ['area', 'centroid', 'bbox', 'label'])
-g GLOB_PATTERN, --glob_pattern GLOB_PATTERN
                        Glob pattern to match files in <input_folder>
                        (default: [A-Zo]*patch_*.tiff)
-p PERCENTAGE_BRAIN_PATCH_SKIP, --percentage_brain_patch_skip PERCENTAGE_BRAIN_PATCH_SKIP
                        Percentage of brain patch skip (default: 0.0)
--no-stack            Stack the patches to form image slices. Default is to
                        stack the patches.
-c CPU_LOAD, --cpu-load CPU_LOAD
                        Percentage of CPU load to use (default: 0.95)
-h, --help            Show this help message and exit

Flag	Parameter	Type	Description	Default
-i, --input_folder	INPUT_FOLDER	str	Path to ACE segmentation output folder (generated_patches/)	None (required)
-r, --raw_input_folder	RAW_INPUT_FOLDER	str	Path to raw clarity tiff folder	None (required)
-o, --output_folder	OUTPUT_FOLDER	str	Path to output folder (default: <input_folder>.parent /cc_slices/)	<input_folder>.parent /cc_slices/
--properties	PROPERTIES	str+	Properties to compute for each neuron	["area", "centroid", "bbox", "label"]
-g, --glob_pattern	GLOB_PATTERN	str	Glob pattern to match files in <input_folder>	"[A-Zo]*patch_*.tiff"
-p, --percentage_brain_patch_skip	PERCENTAGE_BRAIN_PATCH_SKIP	float (0.0 to 1.0)	Percentage of brain patch skip	0.0
--no-stack	N/A	bool	Stack the patches to form image slices. Default is to stack the patches.	False
-c, --cpu-load	CPU_LOAD	float	Percentage of CPU load to use	0.95

Example usage:

$ miracl seg instance \
     -i generated_patches/ \
     -r raw_data/ \
     -o output_dir \
     -p 0.2