Isolating Nonlinear Independent Sources in fMRI with $β$-TCVAE Models

2026-05-15

AI Summary

Traditional independent component analysis is constrained by the assumption of linear mixing, limiting its ability to capture the nonlinear dynamics inherent in fMRI data. This work proposes the first application of β-TCVAE—a total correlation variational autoencoder requiring no additional hyperparameters—to real fMRI analysis, leveraging nonlinear disentangled learning to recover spatiotemporal source signals and yield biologically interpretable latent representations. By integrating functional network connectivity analysis, the method successfully reproduces canonical intrinsic connectivity networks such as the default mode network. The extracted latent components exhibit coherent and neuroscientifically meaningful brain organization, establishing a novel paradigm for nonlinear separation of brain signals.

Abstract

Learning meaningful latent representations from nonlinear fMRI data remains a fundamental challenge in neuroimaging analysis. Traditional independent component analysis, widely used due to its ability to estimate interpretable functional brain networks, relies on a linear mixing assumption for latent sources, limiting its ability to capture the inherently nonlinear and complex organization of brain dynamics. More recently, deep representation learning methods have emerged as promising alternatives for modeling nonlinear latent structure. However, many of these approaches have been evaluated primarily on simulated datasets or natural image benchmarks, with comparatively limited validation on real-world neuroimaging data such as fMRI. In this work, we are motivated by the $β$-TCVAE (Total Correlation Variational Autoencoder), a refinement of the $β$-VAE framework for learning latent representations without introducing additional hyperparameters during training. We adapt and modify this model to fMRI data for nonlinear source disentanglement, aiming to separate mixed spatial and temporal brain signals into interpretable components. We show that the $β$-TCVAE framework can recover meaningful nonlinear spatial components with biological relevance, including well-established intrinsic connectivity networks such as the default mode network. Furthermore, we evaluate the learned representations using functional network connectivity, showing that the latent structure captures coherent and interpretable brain organization patterns. This study provides a pilot investigation that bridges nonlinear representation learning and fMRI analysis.

Problem

• fMRI
• nonlinear independent sources
• latent representations
• brain dynamics
• source separation

Innovation

• β-TCVAE
• nonlinear source separation
• fMRI
• disentangled representation
• functional brain networks