FileNotFoundError ,How i get it *.nii.gz? about cnn-ms-lesion-segmentation

[mengluffy]

yes , me too. I don't know how to options folder, my nii.gz file all in one folder.

[sergivalverde]

Hi @masmai @mengluffy

Following the provided example notebook, each training image is inside a folder:

train_folder = '/mnt/DATA/w/CNN/images/train_images'
train_x_data = {}
train_y_data = {}

# TRAIN X DATA
train_x_data['im1'] = {'T1': os.path.join(train_folder,'im1', 'T1.nii.gz'), 
                       'FLAIR': os.path.join(train_folder,'im1', 'T1.nii.gz')}
train_x_data['im2'] = {'T1': os.path.join(train_folder,'im2', 'T1.nii.gz'), 
                       'FLAIR': os.path.join(train_folder,'im2', 'T1.nii.gz')}
train_x_data['im3'] = {'T1': os.path.join(train_folder,'im3', 'T1.nii.gz'), 
                       'FLAIR': os.path.join(train_folder,'im3', 'T1.nii.gz')}

# TRAIN LABELS 
train_y_data['im1'] = os.path.join(train_folder,'im1', 'lesion_bin.nii.gz')
train_y_data['im2'] = os.path.join(train_folder,'im2', 'lesion_bin.nii.gz')
train_y_data['im3'] = os.path.join(train_folder,'im3', 'lesion_bin.nii.gz')

So, your input folder should be something like:

[training_folder] 
                        [im1 folder]
                                            flair.nii.gz
                                            t1.nii.gz
                                            lesion_bin.nii.gz
                        [im2 folder]
                                            flair.nii.gz
                                            t1.nii.gz
                                            lesion_bin.nii.gz
                        
                       ...
                       
                        [im_n folder]
                                           flair.nii.gz
                                           t1.nii.gz
                                           lesion_bin.nii.gz

where image folders names are used as keys in the dictionary train_x_data

Take into account that you can call the training images differently simply filling the dictionary in other ways. For instance, if you have all training mages inside the same folder, then:

train_folder = '/path/to/training/images/'
train_x_data = {}
train_y_data = {}

# TRAIN X DATA
train_x_data['im1_id'] = {'T1': os.path.join(train_folder, 'T1_im1.nii.gz'), 
                       'FLAIR': os.path.join(train_folder, 'FLAIR_im1.nii.gz')}
train_x_data['im2_id'] = {'T1': os.path.join(train_folder, 'T1_im2.nii.gz'), 
                       'FLAIR': os.path.join(train_folder, 'FLAIR_im2.nii.gz')}
train_x_data['im3_id'] = {'T1': os.path.join(train_folder, 'T1_im3.nii.gz'), 
                       'FLAIR': os.path.join(train_folder,  'FLAIR_im3.nii.gz')}

# TRAIN LABELS 
train_y_data['im1_id'] = os.path.join(train_folder, 'lesion_bin_im1.nii.gz')
train_y_data['im2_id'] = os.path.join(train_folder,'lesion_bin_im2.nii.gz')
train_y_data['im3_id'] = os.path.join(train_folder,'lesion_bin_im3.nii.gz')

hope it helps,

[masmai]

I'm sorry to make you misunderstand. Actually, I've already put "traning_folder" Follow the instructions.
But I do not know where to find Traning images?

[sergivalverde]

I'm sorry but the repository do not provide training images.

[mengluffy]

I change folder in my computer, don't in code, than have another error
Traceback (most recent call last):
File "/home/kmust/cnnsg/train_leave_one_out.py", line 79, in
model = train_cascaded_model(model, train_x_data, train_y_data, options)
File "/home/kmust/cnnsg/base.py", line 42, in train_cascaded_model
X, Y = load_training_data(train_x_data, train_y_data, options)
File "/home/kmust/cnnsg/base.py", line 140, in load_training_data
y_data = [train_y_data[s] for s in scans]
KeyError: 'im1'

[sergivalverde]

And what's the question?

It's obvious that something is wrong with the dictionary, isn't it? The dictionary keys should correspond to the image folder names. Just with this information is very difficult to help:
Look at this:

https://github.com/sergivalverde/cnn-ms-lesion-segmentation/blob/master/train_leave_one_out.py#L44

[melrobin]

It appears that the data is not in .nii.gz format, but it seems in NRRD. Can someone confirm this?

[sergivalverde]

Original images are in nrrd format. You have to convert them. see here <https://stackoverflow.com/questions/47761353/nrrd-to-nifti-file-conversion?rq=1>for instance on how to do it.

…

On Tue, 15 May 2018 at 22:33 melrobin ***@***.***> wrote: It appears that the data is not in .nii.gz format, but it seems in NRRD. Can someone confirm this? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#3 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AFCmQmTtQqzRitDzqEfUhKpJaqryfU2aks5tyzusgaJpZM4PTaCb> .

[melrobin]

OK, thanks! I was able to perform the conversion.