This repository contains the custom computational pipeline used for template-based reconstruction of the optic radiation (OR) and optic tract (OT) and extraction of subject-level diffusion tensor imaging (DTI) measures.
The pipeline was developed for the visual pathway analysis performed in an Alzheimer's Disease Neuroimaging Initiative (ADNI) study.
The analysis consists of five main stages:
- DWI preprocessing
- FOD estimation and generation of a study-specific template
- Preparation of anatomical masks in the study-specific template space
- Template-based OR and OT tractography
- Transformation of tract masks to individual diffusion space and extraction of diffusion measures
The main diffusion measures extracted from the OR and OT are:
- Fractional anisotropy (FA)
- Mean diffusivity (MD)
- Axial diffusivity (AD; referred to as AxD in the manuscript)
- Radial diffusivity (RD)
Performs DWI preprocessing, including:
- DWI denoising
- Gibbs-ringing correction
- Eddy-current and head-motion correction
- Bias-field correction
- Brain masking
- Rigid registration to MNI space
- Gradient-vector rotation
The resulting dwi_preprocessed.mif file is used as the input for the
subsequent FBA-based processing.
This script was adapted and modified from the DWI preprocessing utility script provided by TractSeg. The original TractSeg project is distributed under the Apache License 2.0.
Performs FOD estimation and generates the study-specific template using MRtrix3.
The main steps include:
- Tissue response-function estimation
- Group-average response-function estimation
- DWI upsampling to 1.3-mm isotropic resolution
- Brain-mask generation
- Multi-tissue constrained spherical deconvolution (MSMT-CSD)
- Multi-tissue intensity normalization
- Study-specific unbiased FOD template generation
- Subject-to-template and template-to-subject nonlinear registration
- Study-specific FA template generation
The study-specific FOD template was generated from a balanced subset of 30 participants: 10 cognitively normal controls, 10 participants with mild cognitive impairment, and 10 participants with Alzheimer's disease.
Prepares the anatomical reference masks for visual pathway tractography.
An optic-radiation anatomical mask derived from the HCP-1065 Young Adult Fiber Template is transformed to the study-specific template space using nonlinear registration between the HCP-1065 FA template and the study-specific FA template.
The transformed left and right OR masks are used as anatomical constraints for subsequent OR tractography.
Performs probabilistic tractography of the OR and OT in the study-specific FOD template space.
The bilateral lateral geniculate nuclei (LGN) and optic chiasm (OC) were manually identified in the study-specific template space based on anatomical landmarks.
For the OR:
- A 4-mm-radius spherical LGN seed is used.
- The HCP-1065-derived OR anatomical mask is used as the tract constraint.
- 5,000 streamlines are generated for each hemisphere.
For the OT:
- A 4-mm-radius spherical LGN seed is used.
- The manually delineated OC is used as an inclusion region.
- 500 streamlines are generated for each hemisphere.
Probabilistic tractography is performed using the iFOD2 algorithm.
A two-stage tractography procedure is used to refine the spatial extent of the reconstructed tracts. First-stage tractograms are converted to track-density images (TDI).
For the OR, the TDI is directly binarized without an additional streamline-density threshold.
For the OT, voxels below 10% of the maximum TDI value are removed before binarization.
The resulting binary masks are used as spatial constraints for the second-stage tractography.
Transforms the final template-space OR and OT masks to each participant's 1.3-mm upsampled diffusion space using the template-to-subject nonlinear transform generated during the FBA workflow.
Diffusion tensors are fitted to the 1.3-mm upsampled DWI data, and FA, MD, AD, and RD maps are generated.
Mean FA, MD, AD, and RD values are then calculated within the left and right OR and OT masks.
The final subject-level diffusion measures are written to a CSV file.
The original analysis was performed in a Linux environment (Ubuntu 20.04 LTS) using the following software:
- MRtrix3
- FSL
- ANTs
- TractSeg
- ITK-SNAP
Software versions that could be verified from the original analysis records are documented in:
docs/software_versions.md
The optic-radiation anatomical masks were derived from the HCP-1065 Young Adult Fiber Template:
https://brain.labsolver.org/hcp_template.html
The FOD-template generation workflow was based on the MRtrix3 fixel-based analysis framework.
The DWI preprocessing script was adapted from the preprocessing utility distributed with TractSeg:
https://github.com/MIC-DKFZ/TractSeg
Imaging data are not distributed with this repository.
The imaging data used in the study were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and remain subject to the applicable ADNI data-use and access requirements.
The repository contains analysis scripts only and does not contain participant-level imaging data or identifying information.
The pipeline reflects the computational workflow used for the visual pathway analysis in the associated study.
Manual identification of the bilateral LGN and delineation of the optic chiasm are required before visual pathway tractography.
Users should verify image orientation, spatial registration, brain masks, tract reconstruction, and transformed tract masks when applying the pipeline to other datasets.
Citation information for this software is provided in CITATION.cff.
A permanent DOI for the archived release will be provided through Zenodo.
See the LICENSE file for licensing information and attribution of
third-party code.