Related Experiment Video
Updated: Jun 23, 2025
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
15.6K
SpaDE: Semantic Locality Preserving Biclustering for Neuroimaging Data
Md Abdur Rahaman1,2, Zening Fu1, Armin Iraji1
1Center for Translational Research in Neuroimaging and Data Science (TReNDS).
Biorxiv : the Preprint Server for Biology
|June 25, 2024
Summary
This study introduces SpaDE, a novel deep learning method for analyzing brain connectome data in schizophrenia. SpaDE effectively identifies subgroups and neural features, improving interpretability and revealing group differences in brain networks and cognitive measures.
Area of Science:
- Neuroimaging
- Computational Neuroscience
- Machine Learning
Background:
- Neuroimaging studies often miss subtle patterns in smaller subgroups, limiting specificity and interpretability, especially in neuropsychiatric conditions.
- Subject heterogeneity in conditions like schizophrenia complicates traditional clustering methods.
- Analyzing high-dimensional, sparse neuroimaging data presents challenges in feature grouping and post hoc analysis.
Purpose of the Study:
- To develop a deep neural network, SpaDE, for unsupervised feature learning and biclustering of neuroimaging data.
- To enhance neurobiological interpretability by preserving semantic locality in subject and feature subgroups.
- To regularize for sparsity in representation learning for improved analysis.
Main Methods:
- Proposed a deep neural network architecture named semantic locality preserving auto decoder (SpaDE).
- Employed SpaDE for unsupervised biclustering on human brain connectome data from schizophrenia (SZ) and healthy control (HC) subjects.
- Compared SpaDE's performance against state-of-the-art biclustering methods.
Main Results:
- SpaDE successfully identified coherent subgroups of subjects and neural features, outperforming existing biclustering techniques.
- The method revealed modular neural communities with significant differences between HC and SZ groups in visual, sensorimotor, and subcortical brain networks.
- Bi-clustered connectivity substructures demonstrated strong correlations with cognitive functions, including attention, working memory, and visual learning.
Conclusions:
- SpaDE offers a powerful approach for unsupervised learning and biclustering in neuroimaging, enhancing interpretability and revealing biologically relevant patterns.
- The findings highlight the potential of SpaDE in understanding the neural underpinnings of schizophrenia and related cognitive deficits.
- This method advances the analysis of complex brain connectome data for both research and clinical applications.
