MultiChannel

[oscarramirezs]

hello all
can someone help me please.

• I have few CT scan to make a segmentation model, I would like to segment two bones with this model.

• the problem is, I have a file .nii for the CT, one file .nii for bone trainer based on components and ignite #1 and one file .nii for bone NifTI image reader 2D/3D #2.

• is there a tutorial to convert these two files into a single multichanel file? or what solusion can I use,

• I want to use the Brain tumor 3D segmentation example for the segmentation

thank you
OB

[rijobro]

Are the images of the same dimension? If yes, you can simply read them and then concatenate the arrays. If not, you can use Orientationd (to make sure they're similarly orientated) and Spacingd (to ensure they have the same voxel spacing).

[oscarramirezs]

Hi @rijobro thank you for your help
the images are the same size 512x512 but different deep ex ( 512, 512, 200) and (512, 512, 350)
-I have ( train images - train_labels_humerus -train_labels_scapula)

[{'image': 'Images_256/DEAB621C050A_A_mimics.nii',
  'label': {'label_scapula': 'Masks_scapula_256/DEAB621C050A_A_mimics_scapula.nii',
   'label_humerus': 'Masks_humerus_256/DEAB621C050A_A_mimics_humerus.nii'}}]

I did the concanate and I have the label in to channels ex (2 , 512, 512, 200)

loader = LoadImaged(keys=("label_scapula", "label_humerus"))
for i in range(len(data_dict_l)):
   data_dict = loader(data_dict_l[i])
   label_scapula, label_humerus = data_dict["label_scapula"], data_dict["label_humerus"]
   add_channel = AddChanneld(keys=["label_scapula", "label_humerus"])
   datac_dict = add_channel(data_dict)
   mask= np.concatenate((datac_dict['label_scapula'], datac_dict['label_humerus']), axis=0)
   label_bones.append(mask)
   i+=1
label_bones[0].shape

(2, 512, 512, 256)

### > but` I'm tryin do doing into the train_transforms = Compose(..

using the tutorial (Brain tumor 3D segmentation with MONAI), it crashes
Do you have an example of a class that I can use for this? I will appreciate it very much

like

class ConvertToMultiChannelBasedOnBratsClassesd(MapTransform):
    """

    """

    def __call__(self, data):
        d = dict(data)
        for key in self.keys:
            result = []
            # merge label 2 and label 3 to construct TC
            result.append(np.logical_or(d[key] == 2, d[key] == 3))
            # merge labels 1, 2 and 3 to construct WT
            result.append(
                np.logical_or(
                    np.logical_or(d[key] == 2, d[key] == 3), d[key] == 1
                )
            )
            # label 2 is ET
            result.append(d[key] == 2)
            d[key] = np.stack(result, axis=0).astype(np.float32)
        return d

thanks Oscar

[rijobro]

You're going to have to be more specific about the error you're getting. Also:

• if you concatenate an image of size 512, 512, 200 with an image of size 512, 512, 350, that will cause an error.
• you have an i+=1 which is unnecessary inside of a for loop.