Adhesions runs "process_imgages.py" as its main module
process_images.py should be run from the command line or in an IDE that allows entering arguments.
example run: python process_images.py '/home/exampledata/cell1/' cell1 _c .tif '/home/randomdir/csvs/'
LONGER EXPLANATION:
This code is setup to process an entire data set of time series images, for each time point of which a stack of size "num" was taken
Arguments:__
It takes five arguments:
-## directory ## -- where images should be -- that is, if there are images c1.tif, c2.tif, etc at /home/dir/images/cell1/t1/cell1t1_c1.tif,/home/dir/images/c2.tif..., home_dir = '/home/dir/images/cell1/'
-## dataset ##--which identifies the label for the data set being processed; it assumes that there is a naming scheme for each image that includes the dataset name, then the subfolder name under which the image is stored --in our example, dataset = 'cell1'
-## basename ## -- in our example, this would be '_c'; it is whatever identifier might be added onto dataset+subfolder, like NIS elements does when collecting stacks
an optional "## num ##" -- the default is 25 images in the stack
-## heightsdir ## -- directory where a set of csvs, corresponding to each time point, are stored. There's given to where this directory might be because it often happens that we process our adhesion heights in a completely separate workflow. so it's not tied to be under the same home directory as the images themselves.
Dependencies/Python Packages Needed:_________________
numpy scipy PIL (Image, matplotlib)
mahotas and pymorph-- available through easy_install or pip, on PyPI
--these are well documented computer vision libraries https://pypi.python.org/pypi/pymorph https://pypi.python.org/pypi/mahotas/1.0.3
can also be installed from source, see instructions: http://mahotas.readthedocs.org/en/latest/
Python version used: 2.7.3 (as long as it's not Python 3.x, it's fine) run with EPD (Enthought Python Distribution) v7.3.2--had buffer overflow issues with python2.7 development done in Ubuntu 12.04 'Precise Pangolin', IDE = Eclipse v3.7.2, pydev through Aptana
Modules:______________________
ImageStackgou.py sets up classes Image and ImageStack (with initials at the end for disambiguation), which represent single images and a stack of images, respectively.
zproject.py takes a stack of images and "z-projects" them into a single tif image, per ImageJ/FIJI's average intensity algorithm, where all values for a given pixel in a stack are added and averaged over the number of slices in the stack. scroll down to 'z project': http://rsb.info.nih.gov/ij/docs/menus/image.html#stacks
segmentation_watershed.py uses the implementation of the Watershed algorithm in pymorph to identify individual segments (in our case adhesions) in the zprojected image. It applies a Gaussian filter first to smooth out some features. It basically convolves the input signal with the Gaussian distribution and is used to reduce noise. See more information: http://en.wikipedia.org/wiki/Watershed_%28image_processing%29
http://en.wikipedia.org/wiki/Gaussian_filter and http://docs.scipy.org/doc/scipy-0.7.x/reference/generated/scipy.ndimage.filters.gaussian_filter.html
segmentation_watershed.py returns an ndarray with the same dimensions as the input image. At every location (corresponding to every pixel) there is either a '0' or a number corresponding to which segment the pixel is determined to belong to.
tabulate_adh_props.py compares the output of the segmentation with a heights csv that is formatted the same, but has values corresponding with adhesion protein height--calculated in another program. It populates a new ndarray with height values only at places where there was a height value and a segment number assignment. It also groups segment heights together to calculate the average adhesion height.
It returns the cross-referenced heights array as well as average adhesion height values and adhesion size (pixel count).
histogram.py plots the average adhesion heights in a histogram and saves it as a .png file in the corresponding time point folder.
draw_segment_image.py labeles each segment with a different color and number so the image can be cross referenced with output csvs with adhesion properties like size and height
readcsv.py and readwrite_file.py are utilities that read and write csvs and txt files.
compile_files.py is a utility that goes into a directory and assumes there are two layers of subfolders. It runs through all the files and collects all files that contain a given "keyword" (usually used to identify an extension type) into the topmost folder.