Hi all! I don’t work with MRI images but with confocal microscopy 3D images. I have images that I need to align in 3D, and I think that an approach such as the one used in DIPY affine registration should work well. However, I’m struggling to use DIPY with my images, since the format of my images is different (they're just regular 3D (x,y,z) numpy arrays). In particular, when I try to call functions such as transform_centers_of_mass, I’m lacking the static_grid2world and moving_grid2world parameters. Is there any way to obtain the grid2world parameters from the raw, non-MRI images? Thanks a lot!
@venkatbits Dear All, I am unable to find a tutorial or a documentation page for reconstructing fibers tracts based on deterministic approach for a simple 12 directional diffusion 'tensor' fit data. Can anyone kindly help me in giving the steps for deterministic tractography after 'tensor fit' to the DWI images.
Hello everyone,
I just started exploring the dipy library, starting from the tutorial.
So far I understand every step that is explained, except for the white matter mask part.
As I tried to use my own data to perform the tractography, I can not use the labels and the white matter mask provided by the example in the tutorial (the one from Stanford), as the dimensions are very different.
I have tried also just to use a whole brain mask produced using the median_otsu step, however the resulting figure was not satisfactory for a healthy participant (mostly only fibers in the right hemisphere were tracked).I suspect this mask could be the problem, but I'm aware too that these two problems could be independent of one another, so any help would be much appreciated!
Thank you
I just started exploring the dipy library, starting from the tutorial.
So far I understand every step that is explained, except for the white matter mask part.
As I tried to use my own data to perform the tractography, I can not use the labels and the white matter mask provided by the example in the tutorial (the one from Stanford), as the dimensions are very different.
I have tried also just to use a whole brain mask produced using the median_otsu step, however the resulting figure was not satisfactory for a healthy participant (mostly only fibers in the right hemisphere were tracked).I suspect this mask could be the problem, but I'm aware too that these two problems could be independent of one another, so any help would be much appreciated!
Thank you
Dear All, I am unable to find a tutorial or a documentation page for reconstructing fibers tracts based on deterministic approach for a simple 12 directional diffusion 'tensor' fit data. Can anyone kindly help me in giving the steps for deterministic tractography after 'tensor fit' to the DWI images.
Dear all, I'm trying to develop a pipeline that uses dipy functions (that have been very useful so far). But I am a bit confused by the adc function in dti.py since it uses the sphere362 in the data folder, instead of the given acquisition b values and vector directions. Is there any way to get the ADC values from a monoexponential model fitting as in https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4709925/ with bvecs provided as in during acquisition instead of sphere directions (nor b values converted to 1 as in the function) and thus end up with Xaxisvoxels x Yaxisvoxels x Slices X NumberofAcquisitionDirections nifty file instead of Xaxisvoxels x Yaxisvoxels x Slices X 362 due to sphere as happens by default? Thanks in advance and hope the question isn't confusing. I was just surprised since MD, FA etc. parametric maps in dti tutorial for example do use the bval bvec from your scan, unlike the adc function
@arokem first off, thanks for the very quick response. hope my reply isn't overly long or any remaining problems down to the fact it's been a few years since I last routinely did linear algebra :-) I did what you advised. strangely enough the code exactly as you sent me left me with a gtab object with the True and False in b0s_mask inverted (so the sphere generated was a nonsense 0 0 due to two basal images) but once I reverted that it was indeed an sphere with the xyz taking our directions. apparent_diffusion_coef(tenfit.quadratic_form, my_sphere) now produces way more reasonable voxel values than my clumsy attempt yesterday to "fix" the function to force our gtab in. however it's still Xaxisvoxels x Yaxisvoxels x Slices x 14 (since we do preclinical research, it's two b values for each of our 7 directions) instead of Xaxisvoxels x Yaxisvoxels x Slices x 7 directions, so I'm afraid the information we get from using two b values is lost. Additionaly I can't help but notice that apparent_diffusion_coef does away with the original bvals using np.ones to get a fully 1 vector. Trying to document myself I see design matrixes equal a column full of 1s and xn multiplied by intercept and regression line slope, but return D or return D.T gives the column/row of 1s in the 'wrong' place (for last column or a row) and in any case doing design_matrix(gtab)[:, :6] means the 1s row/column is dropped anyway before doing np.dot. Am I doing something wrong, understanding something wrong in how diffusion analysis is usually done or how dipy functions work? Just still a bit puzzled at b values being converted to 1s and then 'dropped'
lastly, now I also get a "Vertices are not on the unit sphere." warning, but maybe that's normal due to our handpicked values and 7 vectors not being much of a 'sphere'
Hi everyone! I have an issue when trying to perform cross-validation for DKI model on 7T data from HCP release. Specifically, the following error appears: File "goodness_of_fit.py", line 46, in <module>
dki_r2 = stats.pearsonr(data_slice[i, j, k, :], dki_slice[i, j, k, :])[0]**2
File "/home/rosella/anaconda3/lib/python3.8/site-packages/scipy/stats/stats.py", line 3868, in pearsonr
normym = linalg.norm(ym)
File "/home/rosella/anaconda3/lib/python3.8/site-packages/scipy/linalg/misc.py", line 140, in norm
a = np.asarray_chkfinite(a)
File "/home/rosella/anaconda3/lib/python3.8/site-packages/numpy/lib/function_base.py", line 485, in asarray_chkfinite
raise ValueError(
ValueError: array must not contain infs or NaNs
I do not know how to solve this and why it just happens for 7T data, unlike 3T . thanks
dki_r2 = stats.pearsonr(data_slice[i, j, k, :], dki_slice[i, j, k, :])[0]**2
File "/home/rosella/anaconda3/lib/python3.8/site-packages/scipy/stats/stats.py", line 3868, in pearsonr
normym = linalg.norm(ym)
File "/home/rosella/anaconda3/lib/python3.8/site-packages/scipy/linalg/misc.py", line 140, in norm
a = np.asarray_chkfinite(a)
File "/home/rosella/anaconda3/lib/python3.8/site-packages/numpy/lib/function_base.py", line 485, in asarray_chkfinite
raise ValueError(
ValueError: array must not contain infs or NaNs
I do not know how to solve this and why it just happens for 7T data, unlike 3T . thanks
@willi3by DBSI fitting is not available. However, if it is clear that DBSI is more advantageous than other implemented models for x reasons it makes sense to implement it. We may need to contact the authors. Be happy to start a discussion topic here https://github.com/dipy/dipy/discussions Please start the discussion by providing a summary of reasons that justify the time commitment.
@jarcosh : oh yes, that should have been
my_sphere = Sphere(xyz=gtab.bvecs[~gtab.b0s_mask])
But it looks like you've solved that already. Let me see if I can address the other issues:
Regarding the 7 vs. 14 directions: Don't you get the same ADC value for the two vectors representing the same direction?
I think that should be the case (and should incorporate information from both b-values, although you should be careful with that, especially if the higher b-value is higher than b=1,000, because the signal is probably not a Gaussian in many places.
I am not sure that I follow the question about the design matrix. In case it helps, take a look at equations 6 - 9 in this paper: https://www.sciencedirect.com/science/article/pii/S1053811906007403.
@arokem I feel silly now: you are completely correct that the ADC values for the two vectors representing the same direction are exactly the same. Being happy the voxels values now seemed reasonable/annoyed that the output still seemed in the wrong dimensions and having other issues to solve I didn't compare ROIs thoroughly as I usually do and didn't notice.
So moved on to something else and noticed just now. Will compare output from our other software to dipy values but as far as using dipy correctly goes I managed thanks to your advice. Your reference seems a better explainer than the one I linked, so thanks for that too, will read carefully edit: (and yes, the formula 9 clarifies how it actually works/how dipy manages and clears my confusion there too)
I am have a tck tracogram and want to use
voxel2streamline to see which streamlines pass through which voxels. I feed in the streamlines and the inverse of the DWI affine to map from streamline to voxel. I am having trouble interpreting the output, as I do not know how to map voxel index to location. Any guidance would be appreciated, thanks!
Hi Everyone! We are very happy to announce that a new educational course is available online: https://youtube.com/playlist?list=PLRZ9VSqV-6soOC0rUEAOV-QiSa_Qxk8JM :rocket: This is a complete course on Diffusion MRI with Theory and Practice :snake: This is your chance to go from zero-to-hero in using :star: DIPY :star: Topics include -
:white_check_mark: dMRI Basics
:white_check_mark: Microstructure Modeling
:white_check_mark: Denoising
:white_check_mark: Tractometry
:white_check_mark: Dictionary Learning
:white_check_mark: Advanced Multidimensional Diffusion Encodings
:white_check_mark: Registration
:white_check_mark: Tissue Classification
:white_check_mark: Deep Learning methodologies in MRI
:white_check_mark: High Field MRI
:white_check_mark: GPU Acceleration
:white_check_mark: Data Harmonization
:white_check_mark: ML in the Clinic and much more.....
Do not miss out on this golden opportunity! If you do like and use DIPY please do give us a star and cite the piece of software that you are using! It encourages us to keep going! May your data live a new life!!
fbval = f'{file2_path}.bval'
fbvec = f'{file2_path}.bvec'
fbvec = f'{file2_path}.bvec'
data, affine, img = load_nifti(file2_path, return_img=True)
bvals, bvecs = read_bvals_bvecs(fbval, fbvec)
print(bvals)
gtab = gradient_table(bvals, bvecs)
print(gtab)
tenmodel = TensorModel(gtab)
print(len(bvals))
new_img = img.get_fdata()
tenfit = tenmodel.fit(new_img)
The error I am getting is this
Traceback (most recent call last):
File "main.py", line 197, in <module>
convert(args.f1t, args.f2t, args.f1p, args.f2p, args.s, args.diff)
File "main.py", line 171, in convert
tenfit = tenmodel.fit(new_img)
File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/dipy/reconst/dti.py", line 793, in fit
self.kwargs)
File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/dipy/reconst/dti.py", line 1303, in wrapped_fit_tensor
*args, kwargs)
File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/dipy/reconst/dti.py", line 1381, in wls_fit_tensor
w = np.exp(np.einsum('...ij,...j', ols_fit, log_s))
File "<__array_function__ internals>", line 6, in einsum
File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/numpy/core/einsumfunc.py", line 1356, in einsum
return c_einsum(operands, *kwargs)
ValueError: operands could not be broadcast together with remapped shapes [original->remapped]: (10,10)->(10,10) (10000,11)->(10000,newaxis,11)
File "main.py", line 197, in <module>
convert(args.f1t, args.f2t, args.f1p, args.f2p, args.s, args.diff)
File "main.py", line 171, in convert
tenfit = tenmodel.fit(new_img)
File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/dipy/reconst/dti.py", line 793, in fit
self.kwargs)
File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/dipy/reconst/dti.py", line 1303, in wrapped_fit_tensor
*args, kwargs)
File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/dipy/reconst/dti.py", line 1381, in wls_fit_tensor
w = np.exp(np.einsum('...ij,...j', ols_fit, log_s))
File "<__array_function__ internals>", line 6, in einsum
File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/numpy/core/einsumfunc.py", line 1356, in einsum
return c_einsum(operands, *kwargs)
ValueError: operands could not be broadcast together with remapped shapes [original->remapped]: (10,10)->(10,10) (10000,11)->(10000,newaxis,11)
I have tried everything including this link right here, dipy/dipy#1790, but it still does not work for me
I have registered two volumes using
dipy.align._public.SymmetricDiffeomorphicRegistration. Using the .update() method, I have interpolated the displacement to a new 3D shape of (S,R,C,3). I would now like to split this 3D array into a list of 2D arrays and then apply each deformation field on a new 2D image. Is there a way of doing that?
# these are your original bvals and bvecs
bvals = gtab.bvals
bvecs = gtab.bvecs
# here we select only 0 and 1000 bvals
selelected_bvals = np.logical_or(bvals == 0, bvals == 1000)
# here we pick only the 3D volumes from the 4D dMRI data pertaining to 0 and 1000 that we did above.
data_selected = data[:, :, :, selelected_bvals]
# now we also take only the bvals and bvecs of the same 0 and 1000 shells.
gtab_selected = gradient_table(bvals[selelected_bvals], bvecs[selelected_bvals])
# now you can use these to fit the DTI model
tenmodel = dti.TensorModel(gtab_selected)
tenfit = tenmodel.fit(data_selected)
Does this answer your question?
import numpy as np; np.savetxt('bval_filename.bval', bvals); np.savetxt('bvec_filename.bvec, bvecs)
hi, I was trying to get the residuals of my data fit on dipy.reconst.dti, to derive R^2 fit for each voxel with a separate homemade function. I'm using fit_method = 'NLLS' so I checked _nlls_err_func. I keep getting ValueError: shapes not aligned so I know I must be doing something wrong. so I have two questions: a) where should I best get the 3x3 tensor from? the tenfit object has several variables with dimensions (inputXaxis, inputYaxis, inputSlices, 3, 3) such as tenfit.evecs.T or (3, 3, inputXaxis, inputYaxis, inputslices) b) how to best use it if I simply want an array of the residuals for each voxel? I assume what I want is residuals = nlls_err_func(tensor, tenmodel.design_matrix, data) but I must be fetching something wrong
As an example of what I mean, here's a test case I tried to do with a specific voxel in the first slice to see if I could fill the array iteratively later. As far as I can tell I'm following the instructions: 3x3 matrix, design matrix, array of signal during acquisition over all acquisition directions and b values, but I get an error 