Optimization of seed density in DTI tractography for structural networks

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Network Science

Background:

• Diffusion Tensor Imaging (DTI) enables mapping of human brain structural networks.
• Brain networks are formed by brain regions (nodes) and connecting fiber tracts (links).
• DTI-derived networks are susceptible to scan noise and tractography algorithm choices.

Purpose of the Study:

• To investigate the impact of seed number in tractography on structural network variance.
• To characterize network variance using a method analogous to NEMA standards for image noise measurement.
• To understand how seed density influences the stability and accuracy of brain network metrics.

Main Methods:

• Utilized Diffusion Tensor Imaging (DTI) data for structural network construction.
• Employed tractography with varying seed densities to generate brain networks.
• Applied a NEMA-like approach to quantify network variance.
• Analyzed the relationship between seed density and network stability metrics.

Main Results:

• Network variance was found to be inversely proportional to the square root of seed density.
• Increased seed numbers led to greater stability in structural network metrics.
• Higher seed counts can introduce spurious fibers, affecting nodal degrees and edge weights.
• Further increases in seed number showed diminishing returns in reducing network variance due to other imaging factors.

Conclusions:

• Seed number significantly impacts local network characteristics and overall architecture.
• A balance is needed between reducing network variance and managing computational resources.
• Proper thresholding is crucial for creating accurate weighted networks when using numerous seeds.
• Optimizing seed selection is key for reliable DTI-based brain network analysis.