DICOM loaded with wrong orientation

[JumpLK]

Describe the bug
Hi, I'm loading some DICOM from this dataset : https://www.ircad.fr/fr/recherche/3d-ircadb-01-fr/
The way I load it is with LoadImaged

The problem is that the orientation of the data is wrong (see screenshot below)

To Reproduce
Download the dataset mentioned above
Run this code (replace ircad_dir by the folder where you downloaded the data)

from monai.transforms import (Compose, LoadImaged, SaveImaged, Spacingd, Orientationd)
from monai.data import DataLoader, Dataset
from monai.transforms.utility.dictionary import AddChanneld
import os

root_dir = r'..\datas\\'
ircad_dir = root_dir + r'data_ircad\\'


train_images_ircad = [os.path.join(ircad_dir, x, 'PATIENT_DICOM')  for x in os.listdir(ircad_dir)]
train_labels_ircad = [os.path.join(ircad_dir, x, 'LABELLED_DICOM')  for x in os.listdir(ircad_dir)]
data_dict = [{'image':train_images_ircad[0], 'label':train_labels_ircad[0]}]

transforms = Compose([
    LoadImaged(keys=["image","label"]),
    AddChanneld(keys=["image", "label"]),
    Spacingd(keys=["image", "label"], pixdim=(1.5, 1.5, 2.0), mode=("bilinear", "nearest")),
    Orientationd(keys=["image", "label"], axcodes="RAS"),
    SaveImaged(keys=["image", "label"], output_postfix='_after_trans')
    ])


ds = Dataset(data=data_dict, transform=transforms)
dl = DataLoader(ds, batch_size=1, num_workers=0)

for data in dl:
    continue

Expected behavior
the data should have the correct orientation

Screenshots

note : the ground truth is obtained by opening the DICOM from 3D Slicer
Environment

================================
Printing MONAI config...

MONAI version: 0.4.0+127.g380f042
Numpy version: 1.19.2
Pytorch version: 1.7.1
MONAI flags: HAS_EXT = False, USE_COMPILED = False
MONAI rev id: 380f042

Optional dependencies:
Pytorch Ignite version: 0.4.2
Nibabel version: 3.2.1
scikit-image version: 0.18.1
Pillow version: 8.1.1
Tensorboard version: 2.4.1
gdown version: 3.12.2
TorchVision version: 0.8.2
ITK version: 5.1.2
tqdm version: 4.56.0
lmdb version: 1.1.1
psutil version: 5.8.0

For details about installing the optional dependencies, please visit:
https://docs.monai.io/en/latest/installation.html#installing-the-recommended-dependencies

================================
Printing system config...

System: Windows
Win32 version: ('10', '10.0.18362', 'SP0', 'Multiprocessor Free')
Win32 edition: Core
Platform: Windows-10-10.0.18362-SP0
Processor: Intel64 Family 6 Model 158 Stepping 10, GenuineIntel
Machine: AMD64
Python version: 3.8.8
Process name: python.exe
Command: ['C:\Users\Camille\anaconda3\envs\kitware\python.exe', 'c:\Users\Camille\Desktop\important\Kitware\liver-segmentation\segmentation\tmp.py']
Open files: [popenfile(path='C:\Program Files\WindowsApps\Microsoft.LanguageExperiencePackfr-FR_18362.35.108.0_neutral__8wekyb3d8bbwe\Windows\System32\fr-FR\kernel32.dll.mui', fd=-1), popenfile(path='C:\Program Files\WindowsApps\Microsoft.LanguageExperiencePackfr-FR_18362.35.108.0_neutral__8wekyb3d8bbwe\Windows\System32\fr-FR\KernelBase.dll.mui', fd=-1)]
Num physical CPUs: 6
Num logical CPUs: 12
Num usable CPUs: 12
CPU usage (%): [9.7, 1.4, 16.7, 1.4, 12.7, 2.8, 9.9, 9.9, 8.5, 1.4, 5.6, 56.9]
CPU freq. (MHz): 2208
Load avg. in last 1, 5, 15 mins (%): [0.0, 0.0, 0.0]
Disk usage (%): 90.1
Avg. sensor temp. (Celsius): UNKNOWN for given OS
Total physical memory (GB): 15.9
Available memory (GB): 8.2
Used memory (GB): 7.7

================================
Printing GPU config...

Num GPUs: 1
Has CUDA: True
CUDA version: 10.2
cuDNN enabled: True
cuDNN version: 7605
Current device: 0
Library compiled for CUDA architectures: ['sm_37', 'sm_50', 'sm_60', 'sm_61', 'sm_70', 'sm_75', 'compute_37']
GPU 0 Name: GeForce GTX 1070 with Max-Q Design
GPU 0 Is integrated: False
GPU 0 Is multi GPU board: False
GPU 0 Multi processor count: 16
GPU 0 Total memory (GB): 8.0
GPU 0 Cached memory (GB): 0.0
GPU 0 Allocated memory (GB): 0.0
GPU 0 CUDA capability (maj.min): 6.1

[rijobro]

Hi, thanks for raising this issue. I think the error might be in the image reader. Using an example DICOM image from the link you provided:

In [41]: data = LoadImaged(keys=["image"])({"image": "PATIENT_DICOM"})

In [42]: data["image"].shape
Out[42]: (91, 512, 512)

In [43]: data["image_meta_dict"]
Out[43]: 
{'origin': array([0., 0., 0.]),
 'spacing': array([0.74218798, 0.74218798, 2.        ]),
 'direction': array([[1., 0., 0.],
        [0., 1., 0.],
        [0., 0., 1.]]),
 'original_affine': array([[0.74218798, 0.        , 0.        , 0.        ],
        [0.        , 0.74218798, 0.        , 0.        ],
        [0.        , 0.        , 2.        , 0.        ],
        [0.        , 0.        , 0.        , 1.        ]]),
 'affine': array([[0.74218798, 0.        , 0.        , 0.        ],
        [0.        , 0.74218798, 0.        , 0.        ],
        [0.        , 0.        , 2.        , 0.        ],
        [0.        , 0.        , 0.        , 1.        ]]),
 'spatial_shape': array([ 91, 512, 512]),
 'original_channel_dim': 'no_channel',
 'filename_or_obj': 'PATIENT_DICOM'}

So image size is [91, 512, 512] with direction matrix as [1, 1, 1] along the leading diagonal and 0 elsewhere. In slicer, the same image has size [512, 512, 91] with direction matrix as [-1, -1, 1] along the leading diagonal and 0 elsewhere. So there is no reported permutation between the two direction matrices yet there has been a shuffle of dimensions.

If you simply open the image and save it again, it no longer occupies the same 3D space according to Slicer and ITKSnap:

Compose([
    LoadImaged(keys=["image"]),
    AddChanneld(keys=["image"]),
    SaveImaged(keys=["image"], meta_key_postfix="meta_dict")
])({"image": "PATIENT_DICOM"})

If the bug is in the reader, I guess that means the problem is in the interface between the ITK reader and MONAI, since Slicer will also use the same IO. Thoughts? @thewtex @aylward @Nic-Ma @ericspod

[rijobro]

Ah, related: #1357 (comment). From what I understand, the original_affine value is wrong which causes all the observed problems.

[aylward]

Should try updating to ITK 5.2 rc2 or newer.

[rijobro]

Hi @aylward, I tried this morning using 5.2 rc3. From the other issue, I get the impression that something also needs to be updated on the MONAI side.

[JumpLK]

Hi, thank you for your answers, are you working on the issue or should I try to make a workaround on my side ?

[thewtex]

Hi @JumpLK, I will work on a patch for this.