Measuring Dependencies between Biological Signals with Temporal Self-supervision, and its Limitations

2025-07-29

AI Summary

How to robustly quantify nonlinear dependencies among multimodal biosignals (e.g., fMRI, physiological, behavioral) without prior assumptions or manual hyperparameter tuning, while distinguishing genuine associations from spurious correlations induced by confounds? Method: We propose *concurrence*, a general-purpose, time-structured self-supervised dependency measurement framework. It models statistical dependence by contrasting the feature separability of temporally aligned (positive) versus misaligned (negative) signal segments—requiring no parametric assumptions, domain-specific knowledge, or handcrafted features. Contribution/Results: This work is the first to leverage intrinsic temporal structure as a self-supervised signal for cross-modal dependency detection. On real-world neuroimaging and physiological datasets, concurrence automatically uncovers scientifically interpretable functional connectivity patterns. It significantly improves sensitivity and robustness to nonlinear, non-Gaussian dependencies compared to conventional measures. Moreover, its contrastive formulation provides an interpretable basis for identifying and mitigating confound-induced spurious correlations.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Security and Privacy: Large-scale security measurements

Abstract

Measuring the statistical dependence between observed signals is a primary tool for scientific discovery. However, biological systems often exhibit complex non-linear interactions that currently cannot be captured without a priori knowledge regarding the nature of dependence. We introduce a self-supervised approach, concurrence, which is inspired by the observation that if two signals are dependent, then one should be able to distinguish between temporally aligned vs. misaligned segments extracted from them. Experiments with fMRI, physiological and behavioral signals show that, to our knowledge, concurrence is the first approach that can expose relationships across such a wide spectrum of signals and extract scientifically relevant differences without ad-hoc parameter tuning or reliance on a priori information, providing a potent tool for scientific discoveries across fields. However, depencencies caused by extraneous factors remain an open problem, thus researchers should validate that exposed relationships truely pertain to the question(s) of interest.

Problem

• Measure complex non-linear dependencies in biological signals
• Develop self-supervised method without a priori knowledge
• Address extraneous factors affecting dependency validation

Innovation

• Self-supervised temporal alignment for dependency detection
• No a priori knowledge or parameter tuning required
• Validates dependencies across diverse biological signals