This paper presents a novel approach for filtering
functional Magnetic Resonance Imaging (fMRI) data streams
using Bayesian techniques, specifically designed for real-time
Radio Frequency (RF) control loops. We implement and compare
two primary filtering methods: causal Kalman filtering for realtime applications and non-causal Gaussian smoothing for optimal
post-processing analysis. Our results demonstrate that Bayesian
filtering techniques can significantly improve the signal-to-noise
ratio (SNR) of fMRI data while maintaining critical temporal
features necessary for RF control systems. Performance metrics
including filter latency, computational efficiency, and filtering
efficacy are analyzed across different noise conditions. The
proposed approach enables more robust RF control systems that
can adapt to the inherently noisy nature of fMRI signals.
Index Terms—fMRI, Bayesian filtering, Kalman filter, Gaussian smoothing, RF control loops, real-time signal processing,
neuroimaging | bgilbert2
