ACE Segmentation Function

Cutting edge vision transformer and CNN-based DL architectures trained on very large LSFM datasets to map cFos brain-wide.

CLI

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

$ miracl seg ace -h

The following menu will be printed to the terminal:

AI-based Cartography of Ensembles (ACE) segmentation method
usage: miracl ace [-h]
                  -sai SINGLE
                  -sao SA_OUTPUT_FOLDER
                  -sam {unet,unetr,ensemble}
                  [-sas height width depth]
                  [-sar X-res Y-res Z-res]
                  [-saw SA_NR_WORKERS]
                  [-sac SA_CACHE_RATE]
                  [-sab SA_BATCH_SIZE]
                  [-samc SA_MONTE_CARLO]
                  [-sav]
                  [-sau]
                  [-sag SA_GPU_INDEX]
                  [-sat SA_BINARIZATION_THRESHOLD]
                  [-sap SA_PERCENTAGE_BRAIN_PATCH_SKIP]

AI-based Cartography of Ensembles (ACE) segmentation method

optional arguments:
    -h, --help            show this help message and exit
    -sai SINGLE, --single SINGLE
                          path to raw tif/tiff data folder
    -sao SA_OUTPUT_FOLDER, --sa_output_folder SA_OUTPUT_FOLDER
                          path to output file folder
    -sam {unet,unetr,ensemble}, --sa_model_type {unet,unetr,ensemble}
                          model architecture
    -sas height width depth, --sa_image_size height width depth
                          image size (type: int; default: fetched from image
                          header)
    -sar X-res Y-res Z-res, --sa_resolution X-res Y-res Z-res
                          voxel size (type: float)
    -saw SA_NR_WORKERS, --sa_nr_workers SA_NR_WORKERS
                          number of cpu cores deployed to pre-process image
                          patches in parallel (type: int; default: 4)
    -sac SA_CACHE_RATE, --sa_cache_rate SA_CACHE_RATE
                          percentage of raw data that is loaded into cpu during
                          segmentation (type: float; default: 0.0)
    -sab SA_BATCH_SIZE, --sa_batch_size SA_BATCH_SIZE
                          number of image patches being processed by the model
                          in parallel on gpu (type: int; default: 4)
    -samc SA_MONTE_CARLO, --sa_monte_carlo SA_MONTE_CARLO
                          use Monte Carlo dropout (default: 0)
    -sav, --sa_visualize_results
                          visualizing model output after predictions (default:
                          False)
    -sau, --sa_uncertainty_map
                          enable map (default: False)
    -sag SA_GPU_INDEX, --sa_gpu_index SA_GPU_INDEX
                          index of the GPU to use (type: int; default: 0)
    -sat SA_BINARIZATION_THRESHOLD, --sa_binarization_threshold SA_BINARIZATION_THRESHOLD
                          threshold value for binarization (type: float;
                          default: 0.5)
    -sap SA_PERCENTAGE_BRAIN_PATCH_SKIP, --sa_percentage_brain_patch_skip SA_PERCENTAGE_BRAIN_PATCH_SKIP
                          percentage threshold of patch that is brain to skip
                          during segmentation (type: float; default: 0.0)

Flag	Parameter	Type	Description	Default
-sai, --single	SINGLE	str	path to raw tif/tiff data folder	None (required)
-sao, --sa_output_folder	SA_OUTPUT_FOLDER	str	path to output file folder	None (required)
-sam, --sa_model_type	{unet,unetr,ensemble}	str	model architecture	None (required)
-sar, --sa_resolution	X-res Y-res Z-res	float float float	voxel size	None (required)
-saw, --sa_nr_workers	SA_NR_WORKERS	int	number of cpu cores deployed to pre-process image patches in parallel	4
-sac, --sa_cache_rate	SA_CACHE_RATE	float	percentage of raw data that is loaded into cpu during segmentation	0.0
-sasw,--sa_sw_batch_size	SA_SW_BATCH_SIZE	int	number of image patches being processed by the model in parallel on gpu	4
-samc,--sa_monte_dropout	SA_MONTE_CARLO	int	use Monte Carlo dropout	0
-sav, --sa_visualize_results	True/False	bool	visualizing model output after predictions	False
-sau, --sa_uncertainty_map	True/False	bool	enable map	False
-sag, --sa_gpu_index	SA_GPU_INDEX	int	index of the GPU to use	0
-sat, --sa_binarization_threshold	SA_BINARIZATION_THRESHOLD	float	threshold value for binarization	0.5
-sap, --sa_percentage_brain_patch_skip	SA_PERCENTAGE_BRAIN_PATCH_SKIP	float	percentage threshold of patch that is brain to skip during segmentation	0.0

Note

The -sa in the flag part stands for segmentation ACE.

Example usage:

$ miracl seg ace \
   -sai ./walking/subject_01/cells/ \
   -sao ./output_dir \
   -sam unet