Hemodynamic-informed parcellation of fMRI data in a joint detection estimation framework.

Area of Science:

• Neuroimaging
• Biophysics
• Computational Neuroscience

Background:

• Functional magnetic resonance imaging (fMRI) measures brain activity indirectly through the Blood-Oxygen-Level-Dependent (BOLD) signal.
• Disentangling vascular and neuronal contributions to the BOLD signal requires accurate estimation of the hemodynamic response function (HRF).
• Current methods often rely on predefined spatial supports (voxels or fixed parcels) for hemodynamic territories, limiting adaptability.

Purpose of the Study:

• To introduce a novel joint parcellation-detection-estimation (JPDE) procedure for analyzing brain hemodynamics in fMRI.
• To develop a method that adaptively identifies hemodynamic territories based on local properties.
• To improve the inference of evoked activity, HRF shapes, and their spatial supports.

Main Methods:

• The proposed JPDE procedure integrates an adaptive parcel identification step.
• Variational approximations are employed for efficient inference of evoked activity, HRF, and spatial supports.
• The method was validated on both synthetic and real fMRI datasets.

Main Results:

• The JPDE procedure demonstrated superior performance compared to standard detection-estimation schemes.
• Adaptive parcel identification based on local hemodynamic properties enhances analysis accuracy.
• The method successfully infers evoked activity, HRF shapes, and their corresponding spatial supports.

Conclusions:

• The JPDE procedure offers a more flexible and accurate approach to analyzing brain hemodynamics in fMRI data.
• This method advances the ability to disentangle vascular and neuronal signals in the BOLD response.
• JPDE serves as a promising new tool for brain exploration and neuroimaging research.