File import - Hierarchical Data Format (HDF) Files (.h5)

[alhadjil]

Hi there,

thanks for this interactive tool! Looking forward to explore all of its potential.

I have stacked pics taken with LSFM saved as a hierarchical data format (HDF) files (.h5) and I'm wondering if VesselVio can be tunned to take such files directly as import. If not, how would you recommend to proceed?

Thanks!

[JacobBumgarner]

Hi @alhadjil! Thanks for checking out VesselVio.

Honestly, I think it would be best to implement compatibility with .h5 files. This seems simple enough to me, as there are numerous packages available that are compatible with HDF files (h5py and even pandas).

Two quick questions:

1. Do you have a small .h5 file that you'd be willing to share with me for testing purposes?
2. Are you using VesselVio from the terminal? If so, you will be able to use .h5 input as soon as I update the image_processing.py code. The 1.1.2 release is still a week or two away, and this feature will likely be included in that update.

[JacobBumgarner]

It also seems like .h5 files store header information, and the exact name of the image in the .h5 file is needed in order to access the file from the hf data frame. See docs here.

This means that if users are planning on loading .h5 files, I'll probably have to place a specific call to ask them what the name of the header is that points to the dataset.

Does this sound accurate to you?

[alhadjil]

Hey,

thanks for the prompt reply!

Unfortunately, my files are too big to upload. I'm a newbie in programming, so I was excited to test the interactive version .exe of VesselVio. I could add the updated image_processing.py in my folder for VesselVio.exe use, no?

[JacobBumgarner]

If you'd be willing to slice a small volume of your dataset (<25mb) and then share it with me for testing purposes, that would be great. If not, no worries. Ideally it would be nice for me to have a 'real world' dataset for testing rather than something that I just generate locally.

I could add the updated image_processing.py in my folder for VesselVio.exe use, no?

Unfortunately, that's not quite how the bundled application works. The library folder is only included because I explicitly add it during bundling, otherwise those files wouldn't be present.

I will add this to the list of features that I would like to include in the v1.1.2 update :) In the meantime, see below for our temporary solution (.h5 -> .nii conversion).

[alhadjil]

What about converting .h5 in .nii?

[alhadjil]

or fiji ^^

[JacobBumgarner]

Cool. As soon as I'm out of this meeting I'll get something written for you ! ~1h

[alhadjil]

I'm working it out with Fiji for the moment, let's see how it goes...

[JacobBumgarner]

Sounds good. Just shoot me a note if you need help!

[JacobBumgarner]

Ok - maybe we can start from here so that I can figure out the shape of your volume.

You'll need to make sure that you have numpy and h5py installed. Since you're using a JupiterLab notebook, the following should work to make sure that they're installed:

import sys
!{sys.executable} -m pip install numpy h5py

After they're installed, try running this. Find the keys of your h5 file, then replace 'KEY_HERE' with the appropriate key that points to your volume.

import h5py
import numpy as np

filename = "/Users/jacobbumgarner/Desktop/NEONDSTowerTemperatureData.hdf5"

hfile = h5py.File(filename, 'r')

file_keys = list(hfile.keys())
print (f"H5 file keys: {file_keys}")

volume = np.asarray(hfile['KEY_HERE'])

print (f"Volume shape: {volume.shape}")
print (f"Volume data type: {volume.dtype}")

Could you then share the print output of your file keys, the volume shape, and the volume data type with me?