DIPY

Download

dipy@python.org

April 2020

2 participants
2 discussions

[DIPY] FWDTI and DTI recon questions including if default output is upper or lower triangle
by dr.ajay.kurani＠gmail.com May 24, 2020

May 24, 2020

Hi DIPY Experts, I am interested in saving the diffusion tensor as a NIFTI file created as part of the free water implementation. I ran the following commands: import sys import os import bz2file import numpy as np import nibabel as nib import dipy.reconst.fwdti as fwdti import dipy.reconst.dti as dti import matplotlib.pyplot as plt fdwi="dwi_tpl_max2000.nii.gz" fbval="dwi_tpl_max2000.bvals" fbvec="dwi_tpl_max2000.bvecs" mask_wb="mask.tpl.final.nii.gz" img = nib.load(fdwi) data = img.get_data() affine = img.affine mask = nib.load(mask_wb) maskdata = mask.get_data() fwdtimodel = fwdti.FreeWaterTensorModel(gtab,fit_method='NLS') fwdtifit = fwdtimodel.fit(data, mask=maskdata) 1) The following errors were displayed even though the code ran after running the fwidtifit command above: .../python_local/scipy/optimize/minpack.py:447: RuntimeWarning: Number of calls to function has reached maxfev = 1800. warnings.warn(errors[info][0], RuntimeWarning) ...python_local/dipy/reconst/fwdti.py:454: RuntimeWarning: overflow encountered in exp y = (1-f) * np.exp(np.dot(design_matrix, tensor[:7])) + \ I wanted to verify if this is just due to imperfect masking near the edge or if this means the code did not fully run and that there are errors to fix. 2) I ran the following code to save the tensor, as from my understanding fwdtifit uses the DTI tensorfit module: img = nib.Nifti1Image(fwdtifit.quadratic_form, affine) nib.save(img, 'FWms_quadratic_form.nii.gz') Is this the upper or lower triangle? The NIFTI standard is lower triangle, but given that in the DTI module has a lower_triangular function I am guessing DIPY default is the upper triangle. I am having trouble using the lower_triangular function so if you can send me the correct way to interface with this command that would be much appreciated. Thanks, Ajay

2 1

Using dipy_recobundles
by Steven Meisler April 21, 2020

April 21, 2020

Hello, I am new to Dipy and this email list. I am working on post-processing streamlines created with the Tractoflow <https://tractoflow-documentation.readthedocs.io/en/latest/> pipeline. The pipeline’s final output is a whole brain tractogram .trk file created with Dipy code. As a next step, I would like to separate the whole brain tractogram into different white matter bundles, particularly the 80 bundles included in bundle_atlas_hcp482. Ultimately, I would like to use these extracted bundles to produce the AFQ-like tract profiles. <https://dipy.org/documentation/1.1.0./examples_built/afq_tract_profiles/> Since I am trying to script as much of this as possible, I would like to use the command line version of RecoBundles (dipy_recobundles <https://dipy.org/documentation/1.1.1./reference_cmd/dipy_recobundles/>). I have three questions about running this command. 1) dipy_recobundles takes as required inputs a whole brain tractogram and one .trk model file. Is it possible to, without looping the command, segment the whole brain tractogram with all of the atlas bundles? 2) The RecoBundles tutorial listed on the Dipy website <https://dipy.org/documentation/0.16.0./examples_built/bundle_extraction/> includes a registration step to make sure the target and model brains are in the same space. Is this step included in the command-line version, and if not is there a command that will do this (I also have the whole brain model from the hcp482 atlas for this purpose)? 3) Am I using the right command to do what I described in the first paragraph? I am not sure I understand the difference between dipy_recobundles and dipy_labelsbundles <https://dipy.org/documentation/1.1.1./reference_cmd/dipy_labelsbundles/>. I am running the latest version of Dipy (1.1.1) from the Brainlife docker container <https://hub.docker.com/r/brainlife/dipy> on a linux HPC. I would be happy to provide more information as needed. Thank you, Steven -- Steven Meisler, MSE <https://scholar.harvard.edu/steven-meisler>PhD Student in Speech and Hearing Bioscience and Technology <https://www.hms.harvard.edu/dms/shbt/> Harvard Medical School / MIT smeisler(a)g.harvard.edu <mailto:smeisler@g.harvard.edu> | (216) 215-3805

2 1