Optimal design of multi-subject blocked fMRI experiments

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Experimental Design

Background:

• Functional magnetic resonance imaging (fMRI) experiment design requires careful selection of the number of subjects and scanning duration.
• Optimizing these parameters is crucial for maximizing statistical power and efficiency in multi-subject fMRI studies.
• Previous methods often relied on power analyses, which may not fully account for experimental costs.

Purpose of the Study:

• To present a novel method for determining the optimal number of subjects and scanning cycles in blocked-design fMRI experiments.
• To apply the A-optimality criterion and a linear cost function to optimize fMRI design parameters.
• To consider both individual stimulus effect estimation and contrast estimation between stimulus effects.

Main Methods:

• Utilized a mixed-effects model to analyze fMRI data.
• Derived analytical results for A-optimal number of subjects and cycles under uncorrelated error assumptions.
• Presented numerical results for correlated errors with a first-order autoregressive (AR1) structure.
• Incorporated a linear cost function to constrain the number of cycles and subjects.

Main Results:

• The optimal number of subjects and cycles are determined based on the A-optimality criterion and cost function.
• Results demonstrate the dependency of optimal parameters on the within- to between-subject variance ratio.
• Analytical solutions were obtained for uncorrelated errors, while numerical solutions were provided for AR1 error structures.

Conclusions:

• The proposed method offers a new approach to optimize scanning time and subject numbers in multi-subject fMRI.
• This method provides an analytical framework for optimizing fMRI design, considering experimental costs.
• The findings highlight the importance of variance ratios in determining optimal experimental parameters for fMRI studies.