Optimizing data processing to improve the reproducibility of single-subject functional magnetic resonance imaging

Area of Science:

• Neuroimaging
• Medical Physics

Background:

• High reproducibility of functional magnetic resonance imaging (fMRI) activation maps is essential for reliable presurgical planning.
• Variability in fMRI data processing can impact the consistency of activation maps.

Purpose of the Study:

• To evaluate how different data processing methods affect the reproducibility of single-subject fMRI activation.
• To identify optimal processing strategies for enhancing fMRI reproducibility.

Main Methods:

• Compared various motion correction, spatial smoothing, regression, and thresholding methods.
• Assessed reproducibility using three metrics: percent difference in activation volume (PDAV), difference in center of mass (DCM), and Dice Similarity Coefficient (DSC).

Main Results:

• Minimal PDAV achieved with minimal spatial smoothing and AMPLE thresholding.
• Minimal DCM observed with affine motion correction and minimal spatial smoothing.
• Improved DSC with affine motion correction and generous spatial smoothing, though DSC's suitability for reproducibility is questioned.

Conclusions:

• Specific data processing methods can improve fMRI reproducibility.
• The optimal processing pipeline is dependent on the chosen reproducibility metric (e.g., PDAV, DCM, DSC).