How to save CT or X-ray data to HDF5 files?

[anxingle]

It is really great work of this repository.

I just wonder could you provide the detail code which convert dicom or nifti file( or any file format) to HDF5 format. It is really helpful for me to convert my X-rays to CT scans. Many thanks.

[Flaick]

I run this code on my own dataset and think this will help you.
Firstly, you need to convert the CT images (.dicom or .nii) into the numpy array. This step you can found many materials online. For example, I use simpleITK to extract numpy array from .nii file.
Then you can create a .h5 file to store that. Note that the .h5 file is just a dictionary type and you can simply save that with
f = h5py.File('HDF5_FILE.h5','w') f['ct'] = your_ct_numpy_data f['xray1'] = your_xray_numpy_data f.close()

If you want to discuss more, my email is [email protected]. I am also studying from scratch.

[anxingle]

@Flaick Thanks for your kindly reply. I just wonder how to deal with spacing and origin, or just extract raw Numpy array from nifti to h5 file?
Most importantly, how to extract the raw NumPy array from the HDF5 file and save to Nifti file (so I can display in 3D slicer or ITK-SNAP).

[Flaick]

Hi, it is simple to extract raw numpy array from .h5 file. Use the code in the following,
hdf5 = h5py.File(file_path, 'r')
ct_data = np.asarray(hdf5['ct'])
x_ray1 = np.asarray(hdf5['xray1'])
x_ray2 = np.asarray(hdf5['xray2'])
To save the ct_data above into nii, you can use this code
out = sitk.GetImageFromArray(ct_data)
sitk.WriteImage(out,'simpleitk_save.nii.gz')

[sainatarajan]

@Flaick Can you provide the structure of the dataset directory that you used to train this model? Do I need to have a separate hdf5 file for each patient data?

[yinyin-llll]

Why i use those command but the computer don't correspond.

[malonzia]

Hi, I am a freshman in this field. If you know how to adjust the width and center appropriately when we transfer the nii to numpy，please send to [email protected]. Thanks.

[WesleyWu42]

Why i use those command but the computer don't correspond.

use python instead of python3