AutoNeuro: An Open-Source fMRI Toolbox for Real-Time Neuroadaptive Task Design

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biomedical Engineering

Background:

• Real-time functional magnetic resonance imaging (fMRI) allows adaptive study of brain function.
• Current real-time fMRI systems face limitations in infrastructure, flexibility, and closed-loop control, often focusing narrowly on neurofeedback.

Purpose of the Study:

• To introduce AutoNeuro, an open-source framework for real-time fMRI acquisition, analysis, and adaptive experimental control.
• To provide a flexible platform for closed-loop neuroimaging, moving beyond traditional neurofeedback designs.

Main Methods:

• AutoNeuro connects directly to MRI scanners for low-latency data streaming and processing.
• A modular pipeline performs real-time preprocessing and feature extraction.
• A Bayesian optimization agent adaptively selects task conditions within a defined 'experiment space' to maximize a user-defined objective function.

Main Results:

• The framework successfully operated within real-time processing constraints.
• Stable model estimates were maintained across iterations, demonstrating reliable performance.
• The system converged on experimental conditions that were most relevant to the measured brain metric, effectively mapping brain responses.

Conclusions:

• AutoNeuro offers a flexible and efficient platform for closed-loop neuroimaging.
• It supports both hypothesis-driven optimization and exploratory mapping of brain metrics.
• This framework advances the adaptive study of brain function using real-time fMRI.