Trouble writing and saving .nii files

I have a working MATLAB code that produces 5 T1 maps for 5 different angles from an MRI data set of (512 512 170 5). When I look at them in MATLAB Figures they show the right values and look fine (see attached). However, when I try and save them using the code below, they save like the images shown in FSLeyes, where they are being split into different slices (5 x 34 slices = 170). I can't understand why?

% Loop to save each flip angle as an individual NIfTI file
for i = 1:size(T1, 4)
    % Convert the T1 data to double
    Data = double(T1(:, :, :, i));
    % Define voxel size
    VoxSize = [0.6, 0.6, 1];
    % Create a name for the NIfTI file
    NameForSaving = sprintf('T1_Map_Angle%d.nii', i);
    % Define the file location for saving
    filelocation = 'location_hidden';
    fullpath = fullfile(filelocation, NameForSaving);
    % Save the data as a NIfTI file
    Nii_Saver(Data, VoxSize, fullpath);
end

[Voss]

What is Nii_Saver?

[Walter Roberson]

I suspect that you need to permute Data before passing it into Nii_Saver

[Grace]

Hi @Walter Roberson,

Yes, i have tried to permute the dat afterwards but it doesnt make a difference. So at the beginning of my code when i enter it into the program i use IMG0 = permute(reshape(squeeze(double(Data)), [512, 512, 5, 170]), [1, 2, 4, 3]); and this is what is causing my data to be handled in the wrong way. However, if i dont use them then my T1 values come out way off. When using montage, i can see that it has split my 170 slices into 5 sets so instead of having a 1 x 170 slices its created an array i belive of 34 x 5 which adds up to 170 (i.e set 1 has slice 1, 6, 11, 16 etc, set 2 has slice 2, ,7 12, 17 etc, set 3 has slice 3, 8, 13, 18 etc) but I can't figure out a way to reverse this process after calculating T1.

[Umar]

Hi @Grace,

In order to resolve this issue, I will suggest a revised approach:

When extracting each flip angle’s data for saving, directly access it without reshaping unnecessarily and modify your saving loop, here is modified code snippet,

       for i = 1:size(T1, 4)

       % Directly convert the T1 data for this angle

       Data = double(T1(:, :, :, i));

       % Define voxel size

       VoxSize = [0.6, 0.6, 1];

       % Create a name for the NIfTI file

       NameForSaving = sprintf('T1_Map_Angle%d.nii', i);

       % Define the file location for saving

       filelocation = 'location_hidden';

       fullpath = fullfile(filelocation, NameForSaving);

       % Save the data as a NIfTI file using Nii_Saver

       Nii_Saver(Data, VoxSize, fullpath);

       end

Make sure that `Nii_Saver` correctly interprets the dimensions of `Data`. If it expects a specific orientation or arrangement of dimensions (e.g., [X,Y,Z]), make sure that you are providing it in that format.After saving each file, it might be beneficial to load them back into MATLAB using a tool like `niftiread` to verify their integrity and correct dimensionality before moving on to visualization in FSLeyes. Hope this helps.

[Grace]

Hi @Walter Roberson,

Thanks for the suggestion, and although it gives me the right layout, my T1 values are very large. The only thing i can think of is to re-order it in the right way but I am not sure how?

[Cris LaPierre]

Your slices don't makes sense to me

(i.e set 1 has slice 1, 6, 1, 16 etc, set 2 has slice 2, 12, 17 etc, set 3 has slice 3, 13, 18 etc)

Do you mean (i.e set 1 has slice 1, 6, 11, 16 etc, set 2 has slice 2, 7, 12, 17 etc, set 3 has slice 3, 8, 13, 18 etc)?

[Grace]

@Cris LaPierre yes, sorry for the typo!

[Grace]

@Cris LaPierre so the order of my ouput slices looks like this:

The size of my data is [512 512 170 5] and here there are five angles, 2, 10, 11, 12, 13. Where the first angle is a proton density image so will appear black. I believe my slices have been organised into 1,6,11,16...2,7,12,17...3,8,13,18... upto 170 slices for each angle. Whereas what i want it to show is 1,2,3,4,5,6...170 for each angle.

[Umar]

Hi @Grace,

Here’s an adjusted code snippet that keeps the original structure while saving:

    for i = 1:size(T1, 4)

    % Directly convert the T1 data for this angle

    Data = double(T1(:, :, :, i)); % This extracts the correct angle

    % Define voxel size

    VoxSize = [0.6, 0.6, 1];

    % Create a name for the NIfTI file

    NameForSaving = sprintf('T1_Map_Angle%d.nii', i);

    % Define the file location for saving

    filelocation = 'location_hidden';

    fullpath = fullfile(filelocation, NameForSaving);

    % Save the data as a NIfTI file using Nii_Saver

    Nii_Saver(Data, VoxSize, fullpath);

     end

If reordering is necessary (to get slices organized as 1,2,...170 for each angle), consider implementing a more explicit reordering function before saving:

    for i = 1:size(T1, 4)

    Data = double(T1(:, :, :, i));

    % Assuming you want to rearrange slices

    reorderedData = reshape(permute(Data, [1, 2, 4, 3]), [512, 512,     170]);

    VoxSize = [0.6, 0.6, 1];

    NameForSaving = sprintf('T1_Map_Angle%d.nii', i);

    fullpath = fullfile(filelocation, NameForSaving);

    Nii_Saver(reorderedData, VoxSize, fullpath);

    end

Before running on your complete dataset, consider testing with smaller sample arrays to confirm that your process works as expected without introducing errors into your T1 values. I will wait for @Cris LaPierre suggestion.

[Cris LaPierre]

@Grace, the ordering of your slices suggests to me that your reshape command is not correctly ordered.

First, what is your raw data size? I'm wondering if you need to perform a reshape or a permute.

The current slice order suggests to me that the '5' dimension is in the wrong place. What happens if you do the following? Are your slices now in the correct order?

IMG0 = reshape(squeeze(double(Data)), [512, 512, 170, 5]);

[Grace]

Hi @Umar

Unfortunately, I have tried this and played around with reshape and remute but i still dont get out the right layout. See below:

[Grace]

@Cris LaPierre Yes, they are now in the right order, but the next step which is iterative math calulcation to work out the T1 value of each pixel, does not work. So i was hoping to leave it how it is ordered to run through the math and then re-order it after.

For exmaple, i have just tried this on the data after the calculations:

% Loop over each angle

n_angle = 5;

for angle = 1:n_angle

a = T1(:, :, 1:34:end, :);

end

and it correctly orders the first 5 slices. As the data is being stored as slice 1,6,11,16 etc then the 34th slice is actually slice 2 and so on. Is there a way for my to repeat this for each slice? I.e instead of doing:

b = T1(:, :, 2:34:end, :);

where now it starts on slice 2 which is actually slice 6 etc and repeating manually to 34 i can do it all in one go? Does that make sense?

[Cris LaPierre]

I think we need more details. Why are you taking every 34th slice?

[Grace]

Okay @Cris LaPierre, @Umar.

So, I start with data that looks like this, size = [512 512 850]:

You can see i have 850 images and those images are actually 5 x 170 images each using a different angle. So they are stored as 1-170 for each angle one after the other. Then I permute and reshape the data to look llike this:

Here, it has split the 170 slices for each angle into 34 and has taken the first slice from each set and grouped them together for each angle. I've shown it visually in this table:

So, what we are seeing is:

Then the data undergoes calculations to find the T1 value for each pixel. This is how my data is laid out after the math iterations:

So here, it has re-shaped my angles back into 1,2,3,4,5 (as the first angle (2 degrees) is a proton density image so should appear black) but has kept the sets. So now displays my slices like this:

So the math is right but when I download the files as niftys (which i do seperately for each angle) and display them in FSL, they look like this:

So the above is one angle only. Becuase the data is laid out as size = [ 512 512 170 5 ], the far right image is my [512 512] but the other two images show that my 170 has been split into [5 34]. So the actual size of my data is [512 512 (34 5) 5] if that makes sense? When what i want to change it to show is instead of my angles split into [5 34] and saving like this 1,6,11,16...2,7,12,17...3,8,13,18...4,9,14,19...5,10,15,20...170, I want to display it 1,2,3,4,5...170.

This is where I have tried to rearrange the data back to it's original layout using reshape/permute but it hasnt worked. So I thought that the other way i could do it is creating a new arrayand selectively storing the slices by picking out the first slice for every 34 slices which would be 1,2,3,4,5 then going back and picking out the second slice for every 34 slices which wouold be 6,7,8,9,10 and so on and then lining them in order.

Hopefully, I have explained that better now.

[Umar]

Hi @Grace,

Based on your description, if I comprehend correctly, it seems like you are trying to rearrange your data in a specific order after performing calculations. You want to display the slices in the order of 1, 2, 3, 4, 5...170 instead of the current order of 1, 6, 11, 16...2, 7, 12, 17...3, 8, 13, 18...4, 9, 14, 19...5, 10, 15, 20...170. To achieve this task, you can use indexing and reshaping operations. Here's an example code that demonstrates how you can rearrange the data:

   % Assuming your data is stored in a variable called 'T1'

   % Get the size of your data

    [height, width, slices_per_angle, num_angles] = size(T1);

   % Reshape the data to [height x width x (slices_per_angle * 

   num_angles)]

   reshaped_data = reshape(T1, [height, width, slices_per_angle * 

   num_angles]);

   % Rearrange the slices to the desired order

   new_order = reshape(1:slices_per_angle*num_angles, 

   [num_angles,slices_per_angle]);

   new_order = new_order'; % Transpose to match desired order

   % Index and select slices according to the new order

   rearranged_data = reshaped_data(:, :, new_order(:), :);

So, as you can see, I reshaped the original data `T1` into a new shape where all the slices are concatenated together, then created a matrix `new_order` that represents the desired order of the slices. By indexing with `new_order(:)`, you can select and rearrange the slices accordingly.This should result in `rearranged_data` having the desired order of slices: 1, 2, 3, 4, 5...170. I assume your data is stored in a four-dimensional array with dimensions `[height x width x slices_per_angle x num_angles]` because adjustments may be needed if your data is organized differently.

I will assume @Cris LaPierre probably have a different approach. For now, this should help. Please let me know if you have any further questions.

[Cris LaPierre]

How are you saving each angle as a NIfTI? What is the code behind Nii_Saver?

[Umar]

Hi @ Cris LaPierre,

To save each angle from a 4D array as individual NIfTI files, here is the code snippet.

% Assuming 'data' is your 4D array with dimensions [height, width, 

slices_per_angle, num_angles]

[height, width, slices_per_angle, num_angles] = size(data);

for angle = 1:num_angles

    % Extract the data for the current angle

    angle_data = data(:, :, :, angle);

    % Define the filename for the NIfTI file

     filename = sprintf('angle_%d.nii', angle);

    % Save the angle data as a NIfTI file

      niftiwrite(angle_data, filename);

    end

So, basically you will loop through each angle, extract the corresponding 3D slice, and save it as a NIfTI file using niftiwrite. The filenames will be then dynamically generated to reflect the angle number. Hope, you have a better solution.

[Cris LaPierre]

Thank you, but I think we need to know how Nii_Saver is doing it.

[Umar]

Hi @ Cris LaPierre,

Well, in this case, we have to wait for @Gracec to provide the missing piece of puzzle.