konopczynski/vessel3ddl

Automated Multiscale 3D Feature Learning for Vessels Segmentation in Thorax CT Images

vessel3ddl's Issues

3d Patch Generation

Hi,

I am trying to use the "patches_3d.py" for my 3d parch generation from MRI/CT images. The code working with a patch size of (4 by 4 by 4) and not working with (32 by 32 by 32). I am getting below error:

Query: ct_patches_3d = extract_patches_3d(ct_new, (32,32,32), max_patches=None, random_state=None)

MemoryError Traceback (most recent call last)
in
----> 1 ct_patches_3d = extract_patches_3d(ct_new, (5,5,5), max_patches=None, random_state=None)
2
3 # Created a 3d patch

in extract_patches_3d(volume, patch_size, max_patches, random_state)
58 patches = extracted_patches
59
---> 60 patches = patches.reshape(-1, p_x, p_y, p_z, n_colors)
61 # remove the color dimension if useless
62 if patches.shape[-1] == 1:

MemoryError: Unable to allocate array with shape (643, 508, 508, 1, 32, 32, 32, 1) and data type int16

About the time consume?

When i run the 'UseClassifier.py', the script consume about 30 minutes. That's too long for just predict for a volume? Is there anywhere to accelerate the script? And can you provide the source code of the 2D algorithm corresponding to your comparison?

Implementation Error

When I implement ExtractPatches file, I got an error

mask.data[:] = open(param.paths2masks_unannotated[v_index]).read()

NotImplementedError: memoryview slice assignments are currently restricted to ndim = 1

Do you know how to fix it? Thanks!

Cannot read the data

There is a error in reading data when I run the ExtractPatches.py. The error is below:
ExtractPatches.py: Line 35, in extract_patches
HelpFunctions.py: Line 27, in ReadVolume
TypeError: right operand length must match slice length.
I hope that you can help me fix it. Thanks!