nuclearP

This project aims to create a comprehensive set of Fiji/ImageJ command-line utilities for nuclear segmentation and fluorescence intensity measurements in Drosophila eye discs (or other complex tissue). It uses Deconvolution Lab for Tikhonov-Mueller deconvolution, WEKA for machine learning-based probability map calculation and 3D ImageJ Suite for final segmentation and 3D object management.

The macros are implemented in Jython and use Java APIs of plugins listed above.

In short, these macros can be used to transform a 3D confocal image like this (only a single slice is shown):

into a list of objects that can be visualized like this:

License

nuclearP is dual licensed under GPL and CC BY 4.0 with the following restriction on scientific publishing:

This is a scientific piece of software, currently unpublished. You are free to use it in your project, however you should NOT publish your paper before we publish ours, unless you obtain permission from us to publish your paper on callaboration basis.

The above restriction will be removed once our paper is accepted.

How to use these scripts

Image segmentation

To run the segmentation script from console using default options follow this example:

fiji --headless ./src/NuclearSegmentation.py \
    --input-dir=./examples/input/ \
    --class-model=./examples/nuclei-test.model \
    --deconv-psf=./examples/psf-test.tif \
    --output-dir=../output

You can set the following segmentation parameters:

fiji --headless ./src/NuclearSegmentation.py \
    --input-dir=../Input \
    --output-dir=../Output \
    --img-channel=1 \
    --deconv-psf=../Parameters/PSF-SP8-63x-1024-Venus.tif \
    --deconv-lambda=0.01 \
    --deconv-k=50 \
    --class-model=../Parameters/SP8-63x-1024-Venus.model \
    --gauss-xy=0.4 \
    --gauss-z=0.4 \
    --seg-bkgd-thr=8192 \
    --seg-seed-thr=127 \
    --seg-seed-r=4 \
    --seg-watershed=true \
    --seg-vol-min=50 \
    --seg-vol-max=10000

Visualization of results

To generate a pseudoimage from the raw image and the object list CSV:

fiji --headless ./src/Plotter.py \
    --input-csv=../Output/Results/OBJ_mCh-disc-1.csv
    --input-image=../Input/mCh-disc-1.tif
    --output-image=../Output/Pseudoimage/mCh-disc-1.tif

Parameter optimization

To generate parameters file for optimizer:

fiji --headless ./src/SegmentationOptimizer.py \
    --params-out='parameters.csv' \
    --gauss-xy-min=0 \
    --gauss-xy-max=5.2 \
    --gauss-xy-step=0.2 \
    --gauss-z-min=0 \
    --gauss-z-steps=5 \
    --seg-bkgd-thr-min=4096 \
    --seg-bkgd-thr-max=36864 \
    --seg-bkgd-thr-step=4096 \
    --seg-seed-r-min=2 \
    --seg-seed-r-max=13 \
    --seg-seed-r-step=2

To run simulation for single paramater set:

fiji --headless ./src/SegmentationOptimizer.py \
    --one-shot=true \
    --gauss-xy=2 \
    --gauss-z=1 \
    --seg-bkgd-thr=12288 \
    --seg-seed-r=2 \
    --input-file=./examples/PM1_C0-sample.tif \
    --output-dir=../output/ \
    --output-file=simulation-2-1-12288-2.csv

To run simulations for the whole paramater array:

fiji --headless ./src/SegmentationOptimizer.py \
    --gauss-xy-min=0 \
    --gauss-xy-max=3 \
    --gauss-xy-step=1 \
    --gauss-z-min=0 \
    --gauss-z-steps=1 \
    --seg-bkgd-thr-min=12288 \
    --seg-bkgd-thr-max=16384 \
    --seg-bkgd-thr-step=4096 \
    --seg-seed-r-min=2 \
    --seg-seed-r-max=4 \
    --seg-seed-r-step=2 \
    --input-file=./examples/PM1_C0-sample.tif \
    --output-dir=../output/ \
    --output-file=simulation-results.csv

High performance computing

These are two PBS scripts for use on clusters/supercomputers in the nuclearP/src/hpc-scripts/ subdirectory. optimizer-worker.pbs is an example job for optimizing segmentation parameters. fullscale-worker.pbs is an example of segmentation job. These scripts are using worker framework and should be submitted using wsub:

wsub -threaded -batch fullscale-worker.pbs -data image-list.csv