Real-Time fMRI Pipeline Decodes Single-Trial Visual Perception Within Seconds

Real-time closed-loop neurofeedback based on functional magnetic resonance imaging (fMRI) has led to important scientific and clinical advances, but real-time fMRI analysis lags behind state-of-the-art fMRI decoding largely due to computational factors, since most advanced decoding pipelines do not fit within real-time processing, where analysis must be conducted in a matter of seconds without leveraging data acquired later in the session. The study presents a real-time compatible adaptation of the computationally intensive MindEye2 pipeline for reconstructing perceived natural images, and uses simulated analyses to document the factors driving performance changes from offline to real-time analysis. The authors frame the work as a proof-of-concept for deploying powerful fMRI decoding pipelines in real time, paving the way for their use in brain-computer interfaces for scientific discovery and clinical treatment.