Related Experiment Video
Updated: Dec 19, 2025
13:26
Measuring Connectivity in the Primary Visual Pathway in Human Albinism Using Diffusion Tensor Imaging and Tractography
Published on: August 11, 2016
12.6K
Neuro4Neuro: A neural network approach for neural tract segmentation using large-scale population-based diffusion
Bo Li1, Marius de Groot2, Rebecca M E Steketee3
1Sino-Dutch Biomedical and Information Engineering School, Northeastern University, Shenyang, China; Department of Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, the Netherlands.
Neuroimage
|June 4, 2020
Summary
Neuro4Neuro, a novel deep learning method, accurately segments white matter tracts from brain MRI scans. This fast and reproducible technique aids in studying aging and neurodegeneration in large populations and clinical settings.
Area of Science:
- Neuroimaging
- Computational Neuroscience
- Medical Image Analysis
Background:
- White matter (WM) microstructure changes are linked to normal aging and neurodegeneration.
- Accurate and reproducible WM tract characterization from diffusion MRI is crucial for studying these changes.
- Existing methods for WM tract segmentation can be slow and complex, limiting large-scale application.
Purpose of the Study:
- To introduce Neuro4Neuro, a novel convolutional neural network (CNN) based approach for direct WM tract segmentation from diffusion tensor images.
- To evaluate the segmentation performance, reproducibility, and speed of Neuro4Neuro.
- To demonstrate the utility of Neuro4Neuro in associating WM microstructure with aging and dementia subtypes.
Main Methods:
- Developed a 3D end-to-end CNN (Neuro4Neuro) for segmenting 25 WM tracts directly from diffusion tensor images.
- Trained Neuro4Neuro on a large population-based dataset (N=9752, 1.5T MRI) of aging individuals.
- Validated the method's generalization on an external dementia dataset (N=58, 3T MRI) and assessed reproducibility using scan-rescan data.
Main Results:
- Neuro4Neuro achieved good segmentation performance with high spatial agreement (Cohen's kappa, κ=0.72-0.83).
- Demonstrated high reproducibility with low scan-rescan errors in diffusion measures (e.g., fractional anisotropy: ε=1%-5%), outperforming a tractography-based method.
- The method was significantly faster (0.5s per tract) and generalized to external data.
- Proof-of-principle studies showed WM microstructure changes associated with aging and differences between dementia subtypes.
Conclusions:
- Neuro4Neuro offers a highly reproducible and fast method for WM tract segmentation.
- The technique shows potential for large-scale neuroimaging studies and clinical applications in aging and neurodegenerative diseases.
- This automated approach facilitates detailed analysis of WM microstructure in diverse populations.
