Computing moment-to-moment BOLD activation for real-time neurofeedback

Area of Science:

• Neuroimaging
• Signal Processing
• Computational Neuroscience

Background:

• Real-time functional magnetic resonance imaging (fMRI) is crucial for applications like brain-machine interfaces.
• Accurate and rapid estimation of blood oxygenation level dependent (BOLD) signal changes is challenging due to low signal-to-noise ratio in fMRI data.
• Existing real-time fMRI methods often compromise between temporal resolution and accuracy by averaging data or failing to account for noise.

Purpose of the Study:

• To develop a new method for estimating moment-to-moment BOLD activation changes from single fMRI acquisitions.
• To improve the accuracy of real-time fMRI feedback signals by distinguishing neural activity from noise.
• To provide a more reliable signal for applications requiring rapid BOLD signal estimation.

Main Methods:

• An incremental general linear model (GLM) fit to the fMRI time series was computed.
• The method calculates the expected signal intensity for each new acquisition.
• The difference between measured and expected intensity is scaled by estimator variance to create a statistic reflecting neural activation.

Main Results:

• The new method successfully separates moment-to-moment BOLD signal changes attributable to neural sources from noise.
• Validation using synthetic and real fMRI data demonstrated the method's effectiveness.
• Comparison with the only other published real-time fMRI method indicated improved performance.

Conclusions:

• The developed method offers a more accurate and rapid estimation of BOLD signal changes in real-time fMRI.
• This advancement has significant implications for learned regulation of brain activation, brain state monitoring, and brain-machine interfaces.
• The technique provides a feedback signal more reflective of neural activation, overcoming limitations of previous approaches.