Related Experiment Video
Updated: Jun 19, 2026
17:06
Co-analysis of Brain Structure and Function using fMRI and Diffusion-weighted Imaging
Published on: November 8, 2012
26.2K
Untamed: Unconstrained Tensor Decomposition and Graph Node Embedding for Cortical Parcellation
Yijun Liu1, Jian Li2,3, Jessica L Wisnowski4,5
1Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, USA.
Biorxiv : the Preprint Server for Biology
|January 23, 2024
Summary
This study introduces Untamed, a novel framework integrating tensor decomposition and graph learning for brain parcellation. It creates more functionally homogeneous cortical regions aligned with brain networks, improving upon existing methods.
Area of Science:
- Neuroscience
- Brain Imaging
- Computational Biology
Background:
- Cortical parcellation is crucial for neuroscience, often using resting-state fMRI (rsfMRI).
- Independent components analysis identifies overlapping functional networks from rsfMRI.
- A gap exists in relating disjoint cortical parcellations to brain-wide networks.
Purpose of the Study:
- To develop a novel framework linking cortical parcellations with brain-wide functional networks.
- To create near-homogeneous functional regions consistent with identified brain networks.
- To overcome limitations of traditional methods assuming statistical independence.
Main Methods:
- Integrated NASCAR (3D tensor decomposition) with graph representation learning.
- Produced cortical parcellations representing homogeneous functional regions.
- Utilized rsfMRI data for network identification and parcellation generation.
Main Results:
- Generated parcellations comparable or superior to established atlases in functional connectivity homogeneity.
- Demonstrated improved alignment with task contrasts and architectonic maps.
- Achieved high automation for rapid adaptation and custom parcellation generation.
Conclusions:
- The Untamed framework successfully integrates tensor decomposition and graph learning for advanced cortical parcellation.
- This approach yields functionally coherent parcels aligned with brain networks.
- Untamed offers an automated, adaptable tool for cognitive and clinical neuroscientific research, with publicly available resources.
