Diffuse optical reconstructions of functional near infrared spectroscopy data using maximum entropy on the mean

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Optical Physics

Background:

• Functional near-infrared spectroscopy (fNIRS) measures hemodynamic responses.
• Diffuse optical tomography (DOT) reconstructs cortical activity from scalp measurements.
• Existing DOT methods face challenges with depth sensitivity and temporal accuracy.

Purpose of the Study:

• To adapt and evaluate the Maximum Entropy on the Mean (MEM) source localization method for DOT reconstruction.
• To improve DOT accuracy by introducing depth weighting and a novel MEM initialization.
• To compare MEM performance against the Minimum Norm Estimate (MNE) for DOT.

Main Methods:

• Adapted nonlinear Maximum Entropy on the Mean (MEM) method for DOT.
• Implemented depth weighting within the MEM framework.
• Utilized a large-scale simulation scheme (4000 simulations) for performance evaluation.
• Validated results with real fNIRS data during a finger-tapping task, comparing with fMRI.

Main Results:

• MEM demonstrated more accurate DOT reconstructions compared to MNE.
• MEM showed robustness in low signal-to-noise ratio (SNR) conditions.
• Real-data results showed MEM reconstructions spatially agreed with fMRI activation maps.

Conclusions:

• The adapted MEM method offers improved accuracy and robustness for DOT reconstruction.
• MEM provides a valuable alternative to MNE for analyzing fNIRS data.
• This approach enhances the utility of DOT for non-invasive brain imaging.