BMC-SSA method stabilizes brain dynamics from noisy fMRI data

Reconstructing latent state-space geometry from time series is key to studying nonlinear brain dynamics, but standard delay-coordinate embedding assumes long, noise-free recordings—a condition rarely met in real-world fMRI. fMRI data is typically short, noisy, and contaminated with autocorrelated background structure that can obscure oscillatory components and destabilize embeddings. To address this, Sir-Lord Wiafe, Carter Hinsley, and Vince D. Calhoun introduce Bootstrap Monte Carlo Singular Spectrum Analysis (BMC-SSA). This method fuses Monte Carlo SSA with bootstrap resampling to identify and retain only those oscillatory modes that are statistically supported and reproducible across resampled data. The result is a reconstruction that emphasizes reliable oscillatory structure, enhancing determinism and stabilizing subsequent delay-coordinate embeddings.

In their experiments on fMRI data, BMC-SSA significantly improved the reliability of functional connectivity measures and uncovered subtle differences in state-space dynamics that were previously masked by noise. The method provides a general framework for robust embedding of noisy, finite signals, with immediate applications in neuroscience and potentially other fields like climate science or finance where short, noisy time series are common. The work, presented as a 6-page conference paper on arXiv, demonstrates that BMC-SSA can extract meaningful brain dynamics even under challenging recording conditions, opening new avenues for studying brain function in clinical and experimental settings.