zpincus / celltool Goto Github PK

Hi @zpincus,

I am trying to measure the intensity along a centerline contour, which I have in the pathway Mg1655/Mg1655_Centerlines. Within the same pathway I also have my image files to be analyzed. They are named appropriately (ie. contours: Mg1655-01-01.contour images: Mg1655-01.tif, etc.) and I was able to run this image swath code:
celltool measure_contours Mg1655/Mg1655_Centerlines/Mg1655*.contour --output-file="LineScansMg1655.csv" --image-swath Mg1655/Mg1655_Centerlines/Mg1655-*.tif

However, when I replace the option "--image-swath" with "--integrate-image", I receive "NotImplementedError"

celltool measure_contours Mg1655/Mg1655_Centerlines/Mg1655*.contour --integrate-image Mg1655/Mg1655_Centerlines/Mg1655*.tif

Hey

In numpy images are represented by the ij indexing that is top left corner is the origin. But i assume that warping should consider cartesian coordinates. How do you ensure that. In np.mgrid isnt the origin at the top left too. I am confused about the issue of origin and the coordinate system in numpy and in biomedical imaging because i believe in itk the origin is specified as 0, 0. How do we take care of this issue here

Hi @zpincus ,
I'm trying to run your warp_images function and I'm confused what to pass as "output region". I'm passing the following tuple but getting a "invalid shape for coordinate array" error.
tuple - (0, 0, image.shape[0], image.shape[1])
Error -
Traceback (most recent call last):
File "wrap2.py", line 117, in
print(warp_images(from_points, to_points, [image], (0, 0, image.shape[0], image.shape[1])))
File "wrap2.py", line 26, in warp_images
return ndimage.map_coordinates(numpy.asarray(image), transform, order=interpolation_order)
File "/home/ritika/anaconda3/envs/shipmnts/lib/python3.7/site-packages/scipy/ndimage/interpolation.py", line 341, in map_coordinates
raise RuntimeError('invalid shape for coordinate array')
RuntimeError: invalid shape for coordinate array

In Python 3.x, image_warp yields the following error:

.../python3.6/site-packages/celltool/numerics/image_warp.py in _make_inverse_warp(from_points, to_points, output_region, approximate_grid)
     55         iy1 = (y_indices+1).clip(0, y_steps-1)
     56         t00 = transform[0][(x_indices, y_indices)]
---> 57         t01 = transform[0][(x_indices, iy1)]
     58         t10 = transform[0][(ix1, y_indices)]
     59         t11 = transform[0][(ix1, iy1)]

IndexError: arrays used as indices must be of integer (or boolean) type

This is an unfortunate consequence of PEP 238 applied to the following location: https://github.com/zpincus/celltool/blob/master/celltool/numerics/image_warp.py#L37
I will submit a pull request shortly.

Hi @zpincus, I discovered your unwarping code (warp_images) through this forum post by Stéfan van der Walt today. It saved me days of struggle!

I had a follow-up question (just below that post) which I thought I'd just bring up with you directly here too. Having defined a transform through your function I now want to transform an array of x-y coordinates (nx2) from the original space into the warped space. How do I do that?